mbfinotti/partnerships-skillsチェック済み
SKILL DETAIL
alliance-prioritization
mbfinotti/partnerships-skills/alliance-prioritization
Rank named candidate strategic alliances and technology partnerships by expected value, effort, and risk, to decide where partner-team capacity goes - must-pass screens, a weighted alliance scorecard with sensitivity analysis, co-innovation and platform-dependency risk checks, a ranked shortlist, and a go/no-go recommendation memo. Covers B2B tech alliances and B2C brand and creator partnerships. Use whenever the user mentions alliance prioritization, partnership ranking, partner shortlisting, or asks which of several named partners deserves investment, even if they never say prioritization. Portfolio-level comparison only. Do NOT use for modeling one partner's P&L - use mbfinotti/partnerships-skills@partner-economics instead.
インストール · 518出典を見る
Installation
npx skills add https://github.com/mbfinotti/partnerships-skills --skill alliance-prioritization
スキルファイル
SKILL.md
最終同期 · 2026/10/23
evals/evals.json›
{
"skill_name": "alliance-prioritization",
"evals": [
{
"id": 1,
"prompt": "I run partnerships at Kelver, a workflow automation vendor doing about $38k ACV, sales-led, mid-market moving upmarket. We have nine technology alliance candidates on the list and two alliance managers, so realistically three alliances next year. I already built a scorecard: brand strength, market size, account overlap, exec relationship, and revenue potential, each at 20%, everyone scored 1-5. Top of the sheet right now is Northwind Cloud at 4.8 (that is the hyperscaler, our CEO golfs with their VP), then Piedmont Analytics at 4.4, then Solvexa Data at 3.2. Integration estimates I have: Northwind 7 manager-months, Piedmont 6, Solvexa 4, and the other six range from 2 to 5. Goal is partner-sourced enterprise pipeline. Can you turn this into a shortlist I can take to our VP of Sales next Tuesday?",
"expected_output": "A rebuilt ranking that rejects the equal-weight sheet, anchors weights on complementarity and compatibility with written justifications, divides value by effort, draws the capacity line on summed effort rather than the headcount of three, runs a sensitivity re-rank, and ends with explicit verdicts including at least one candidate left off.",
"files": [],
"expectations": [
"Rejects the equal 20% weight distribution outright rather than tuning the numbers inside it",
"States that every weight must carry a written one-line justification",
"Names resource complementarity and organizational compatibility as the two criteria that should carry the highest weight",
"Points out that neither complementarity nor compatibility appears anywhere in the user's five existing criteria",
"Flags brand strength and exec relationship, together 40% of the sheet, as logo halo and relationship criteria that crowd out fit",
"Caps account overlap as one input rather than leaving it as an equal share of the ranking spine",
"Introduces an effort column and ranks candidates on value divided by effort",
"States the ranking order explicitly as value divided by effort, then value composite, then logo attractiveness",
"Shows that Northwind Cloud's 7 manager-months move its position once effort divides value, rather than leaving it first",
"Draws the capacity line on summed effort estimates rather than on the headcount number of three",
"Runs a sensitivity check shifting each weight plus or minus 20% and reports which positions hold and which flip",
"Applies a tie band of roughly 5% and refuses to treat a sub-5% composite gap as decisive",
"Scores the CEO-connected Northwind Cloud on the same sheet as every other candidate rather than protecting or excluding it",
"Ends with at least one candidate not making the shortlist, each non-shortlisted candidate carrying an explicit verdict",
"Asks for the missing interview inputs (existing alliance portfolio, data available per candidate, sign-off owner) before finalizing the table"
]
},
{
"id": 2,
"prompt": "Follow-up on our alliance shortlist at Kelver. Piedmont Analytics came out highest on the scorecard at 4.4, but legal came back and said Piedmont has no SOC 2 Type II, their last attestation lapsed 14 months ago, and they had a customer data exposure incident in 2024 that they disclosed late. The alliance needs us to push customer records to them. Our VP wants to keep them in the shortlist and just add 'security posture' as a criterion at 15% weight so the sheet reflects it. Piedmont would still come out second even scoring a 1 there. Can you rework the scorecard that way?",
"expected_output": "A refusal to convert the security failure into a weighted criterion, removal of Piedmont from the ranked list on the gate, the evidence cited, conditional-go named as the only route back, and the shortlist and capacity line redrawn.",
"files": [],
"expectations": [
"Refuses to convert security posture into a weighted scorecard criterion",
"States that must-pass gates live outside the tradeable score",
"Removes Piedmont Analytics from the ranked list on the gate failure rather than re-ranking it lower",
"States explicitly that a high composite score cannot buy a candidate past a gate",
"Names the specific evidence behind the gate decision: the lapsed attestation, the late-disclosed incident, and the customer-record data-sharing requirement",
"Identifies the failed gate as the security and compliance posture gate within a defined must-pass gate set",
"Offers conditional-go as the only route back, tied to a named condition and a deadline, rather than to a scored position",
"Lists the other must-pass gates the remaining candidates must clear, including financial viability, legal or antitrust exposure, exit path, and the three-laws screen",
"Explains that scoring and socializing a candidate before its security or legal gate is what makes a gate tradeable in the first place",
"Records the gate decision with its evidence inside the memo rather than parking it in a compliance annex",
"Does not average or blend the security concern into any composite score",
"Redraws the shortlist and the capacity line after the removal rather than promoting the next-highest value score by default"
]
},
{
"id": 3,
"prompt": "Vantiq Systems here, we make field-service scheduling software. Our CPO says our biggest gap is predictive maintenance; we have no ML capability and no plan to build one. Six alliance candidates. Five are straightforward integration partners, 3 to 4 manager-months each, all would ship inside a quarter. The sixth is Orrin Labs, whose anomaly-detection engine would actually close the gap, but it needs them to expose a streaming inference endpoint they do not have yet, and we would need to build an event pipeline on our side. Their roadmap has the endpoint dated for Q3 next year; ours has the pipeline in the same half. Orrin would be about 9 manager-months. We have 14 manager-months total. My ranking has Orrin dead last on value-per-effort. Is that right?",
"expected_output": "An explanation that the ratio structurally starves the deep co-innovation candidate, confirmation that both dependencies are dated on the two parties' own roadmaps, promotion of Orrin by reserving capacity before scoring rather than by adding composite points, and co-innovation plus adoption-chain risk scored explicitly.",
"files": [],
"expectations": [
"Names the value-per-effort ratio as structurally underweighting a candidate that is both expensive and strong",
"Identifies Orrin Labs as the deep co-innovation candidate in this list",
"Ties the promotion decision to the stated goal being product-gap fill rather than pipeline",
"Checks whether the unshipped dependency sits on the two parties' own dated roadmaps rather than a third party's",
"Confirms both the Orrin endpoint and the Vantiq pipeline carry dates, and treats that as the condition making the candidate promotable",
"States that an undated dependency would earn a recycle rather than a reserved capacity slot",
"Promotes Orrin by reserving a slice of capacity before scoring starts, not by adding points to its composite score",
"States that adding points to the composite would be the false precision the sensitivity analysis exists to undo",
"Spends the reserved co-innovation slot inside the 14 manager-month budget rather than on top of it",
"Scores co-innovation risk for Orrin as a named risk category, measuring how much of the value depends on an unshipped roadmap",
"Scores adoption-chain risk, asking which intermediaries must act before the end customer sees the value",
"Keeps Orrin's raw value-per-effort position visible rather than rewriting the ratio to justify the promotion",
"States how much capacity is left for the five integration candidates after Orrin's reserved slot"
]
},
{
"id": 4,
"prompt": "We are Brackenly, a 40-person devtools company. Two candidates left after gates. One is Helixon, a big observability platform: we would list in their marketplace and build on their agent SDK, and about 70% of the joint value depends on that SDK. They shipped a first-party feature last quarter that overlaps maybe 60% with what we do, and their marketplace terms let them feature their own modules above third-party ones. The other is Tunstall Metrics: small, no product overlap with us, no prior relationship, roadmap heading straight at a segment we also want, and their CEO has said publicly that partnerships are 'a distraction'. But their tech fit is excellent and the integration is easy. Which do we back?",
"expected_output": "Platform envelopment named and scored for Helixon with staged commitment and a dependency ceiling rather than a veto, relational and performance risk scored separately for both, Tunstall flagged as high relational risk, and the risk composite shown beside value and effort.",
"files": [],
"expectations": [
"Names platform envelopment as the specific risk Helixon carries",
"Cites the warning signs present: adjacent first-party features shipped, marketplace terms permitting self-preferencing, and value sitting in a thin replicable layer",
"States that high envelopment risk does not forbid the alliance",
"Requires staged commitment, contractual safeguards, and a stated dependency ceiling in the memo if Helixon proceeds",
"Scores relational risk and performance risk separately rather than collapsing them into one risk number",
"Identifies Tunstall Metrics as high relational risk, citing the competing roadmap, the absence of prior ties, and the publicly stated posture against partnerships",
"States that relational risk is mitigated by governance and staged commitment, not by enthusiasm or tech fit",
"States that performance risk is mitigated by pilots and milestone gates",
"States that a candidate high on both relational and performance risk is a kill or a small staged pilot, never a flagship bet",
"Shows the risk composite beside the value and effort columns rather than in an annex read after the decision",
"Names which risk drives the gap between the two candidates rather than presenting them as tied",
"States what mitigation would close that gap",
"Refuses to rank the two on tech fit and integration ease alone"
]
},
{
"id": 5,
"prompt": "I head brand partnerships at Solane, a premium skincare brand. We are the high-end one in our category, average basket 95 euros. I have eleven partnership candidates for next year: four creator partnerships, three brand co-brands, two retail-media tie-ups, two podcast sponsorships. Budget covers maybe four. My agency sent audience overlap numbers and follower counts for all eleven and ranked them by reach. Top of their list is Marisse Bell, 2.1M followers, 48% audience overlap with us, but I have heard nothing about whether her numbers are real. Second is a co-brand with Halvern, a luxury fragrance house even more prestigious than us. Can you build me the ranking?",
"expected_output": "Fraud and brand-safety screens run pass/fail before scoring, brand fit named as the B2C fit construct, the high-esteem dilution warning applied to Solane in the Halvern pairing, reach capped as one input, and the same method spine as the B2B version with verdicts and a capacity line.",
"files": [],
"expectations": [
"Runs creator fraud vetting as a pass/fail screen before any scoring, not as a scored criterion",
"Treats Marisse Bell's unverified follower base as unscreened rather than pre-screened",
"Requires an audience audit from a fraud-screening tool as the evidence behind that gate",
"Names brand safety as a second pass/fail gate for this candidate list",
"Names brand fit or attitude congruence as the core fit construct on the B2C side",
"States that attitude spillover between paired brands runs in both directions and not symmetrically",
"Warns that as the high-esteem brand Solane is the more exposed party in the Halvern co-brand, rather than the automatic beneficiary",
"Models at least one negative-spillover scenario per shortlisted candidate and what it would cost Solane",
"Caps audience overlap as one input rather than letting the agency's reach ranking be the ranking spine",
"States that the method spine, including staged gates, justified weights, sensitivity analysis, the capacity line, and the memo shape, is identical to the B2B version rather than inventing a separate B2C method",
"Demands incrementality-tested metrics rather than vanity reach figures",
"Scores effort per candidate and ranks on value against effort rather than on reach",
"Issues a verdict per candidate drawn from go, conditional-go, recycle, hold, and kill",
"Draws a capacity line and leaves below-the-line candidates with explicit re-entry triggers"
]
},
{
"id": 6,
"prompt": "Writing the exec summary for our alliance shortlist at Carrowmore Software. I want to open with something that lands. I have two numbers from a partner-ops platform's report: '70% of alliances fail' and 'deals are 53% more likely to close when a partner is involved'. I also found a McKinsey survey saying 70% of alliances underperform. They all point the same direction, so I am going to lead with the 70% failure rate and use the 53% to justify the headcount ask. Draft me that paragraph.",
"expected_output": "A paragraph that refuses the single settled failure rate, presents the published spread and the lower re-analysis figures, separates underperformance from failure, downgrades the 53% claim to the vendor's own network figure, flags provenance on every number, and substitutes the peer-reviewed alliance-function finding for the headcount argument.",
"files": [],
"expectations": [
"Refuses to state a single settled alliance failure rate",
"Presents the published spread of roughly 20% to 80% depending on how failure is defined",
"Cites the credible re-analysis putting failure nearer 25-30%",
"Notes the lower 18-20% figure for firms that have an alliance-management function",
"Traces the widely repeated 60-70% phrasing to the 2007 Harvard Business Review piece by Hughes and Weiss",
"States that this source cites no study, sample, or definition behind the figure",
"Distinguishes the McKinsey survey as measuring underperformance and need for restructuring, not outright failure",
"Notes the McKinsey survey's small base of roughly 30 companies",
"Identifies the 53% close-rate figure as a self-reported vendor survey of roughly 500 professionals",
"Contrasts it against the same vendor's own network data showing roughly an 11.7% win-rate lift",
"Recommends quoting the smaller, harder 11.7% figure if either is used at all",
"Attaches a provenance flag to every number kept in the memo, distinguishing peer-reviewed, analyst, vendor first-party, and vendor survey",
"Offers the Kale, Dyer and Singh finding of 63% versus 50% alliance success with a dedicated alliance function, across 1,572 alliances, as the defensible basis for the headcount ask"
]
},
{
"id": 7,
"prompt": "Board meeting is Thursday morning. It is Tuesday afternoon and I have maybe four hours on this, no analyst, and I cannot pull RevOps for an overlap report. Three candidates: Aldwyn Networks, Fennimore Cloud, Brightsea AI. The board wants to know which one we sign this quarter. We have one alliance manager. Give me the scorecard.",
"expected_output": "The scorecard dropped under the effort ceiling and replaced by gates plus effort to first value, the full gate set run on all three rather than staged, reach-led criteria deleted outright with the reason stated, a capacity line still fixed, verdicts still issued, and candidate facts flagged unverified.",
"files": [],
"expectations": [
"Removes the weighted scorecard entirely rather than producing one under the four-hour ceiling",
"Ranks instead on the must-pass gates plus effort to first value",
"States explicitly in the output that the scorecard was dropped and why",
"Runs the full gate set on all three candidates rather than staging desk checks ahead of expensive reviews",
"Explains that staging saves nothing on a longlist of three",
"Deletes reach-led criteria outright given no overlap dataset and no audience-audit budget",
"States the deletion reasoning: stripped of the dataset, reach-led leaves only market size and brand",
"Names each deleted option and its reason rather than demoting it to the bottom of a menu",
"Fixes a capacity line from the single alliance manager before recommending anything",
"Issues a verdict per candidate drawn from go, conditional-go, recycle, hold, and kill",
"Flags the candidate facts as user-supplied and unverified rather than presenting them as checked"
]
},
{
"id": 8,
"prompt": "Second time around on this. Last quarter I brought a ranked alliance list to our steering committee and they reopened it. The CRO said the weights were 'made up' and we spent 90 minutes arguing about whether integration depth should count more than account overlap. Nothing got signed. We are re-running it in six weeks for the FY plan, and this scorecard is going to get re-run at every portfolio review after that. Seven candidates, four alliance-manager slots. Same committee, same CRO. How do I keep this from happening again?",
"expected_output": "Formal weight elicitation promoted above the justified-weights default for this specific situation, best-worst method named and preferred over AHP, its cost and what it actually buys stated plainly, and the sensitivity re-rank named as the artifact that survives the argument.",
"files": [],
"expectations": [
"Promotes formal weight elicitation above the default of individually justified weights for this situation",
"Names the best-worst method as the elicitation technique to use",
"Prefers best-worst method over AHP, citing fewer pairwise comparisons and less rank-reversal exposure",
"Names the conditions that promote it: the weights themselves being what the room argues about, a sign-off owner who distrusts the sheet, a committee that has already reopened a ranking, and a scorecard that will be re-run at every portfolio review",
"States that it costs roughly half a day of structured elicitation per stakeholder plus the coordination to convene them",
"States that it buys agreement on the weights, never precision in them",
"Keeps individually justified weights as the written record regardless of which elicitation method is used",
"Rejects equal weights outright as the documented practitioner failure mode rather than ranking them as a cheap option",
"Names rank reversal as a documented weighted-scorecard defect, attributed to Belton and Gear",
"Names the arbitrary-weights critique of weighted scorecards, attributed to Dyer",
"Runs the plus or minus 20% sensitivity re-rank and reports stable versus flipping positions as the artifact that survives the argument",
"States plainly that no elicitation method substitutes for good criteria"
]
},
{
"id": 9,
"prompt": "Hanmere Group, we sell a consumer finance app and also license our risk engine to banks. Mixed candidate list, five names. Two are bank-side tech alliances, two are consumer co-brands with lifestyle brands, and one is Ostrander, a small analytics shop whose whole product is one model we could honestly rebuild in about five engineer-months, though partnering is faster. One of the consumer co-brands, Verrow Athletic, has an unresolved supply-chain labour story from last year, but their pipeline contribution model is the strongest of the five. My finance lead suggested we average the brand risk into the score as a 15% deduction so we are not being unfair to them. Nine manager-months of capacity.",
"expected_output": "An ally-versus-acquire-versus-build pre-filter that pulls Ostrander out before scoring, a refusal to average the Verrow brand-safety concern into any score, both evaluation logics run on the hybrid list and gated on the stricter, and the capacity line drawn on summed effort inside nine manager-months.",
"files": [],
"expectations": [
"Runs an ally-versus-acquire-versus-build test before scoring anything",
"Identifies Ostrander as a candidate where building rather than allying may be the right mode, citing the five engineer-month rebuild estimate",
"Removes wrong-mode candidates at that pre-filter step rather than carrying them into the scorecard",
"Refuses to average the Verrow Athletic brand-safety concern into a composite score as a percentage deduction",
"States that gates never average",
"Treats brand safety as a pass/fail gate for the consumer co-brand candidates",
"Identifies the candidate list as hybrid, carrying both B2B and B2C currencies",
"Runs both evaluation logics on the hybrid candidates: pipeline math and brand-fit or contagion screening",
"Gates on the stricter of the two logics rather than blending or averaging them",
"Prevents Verrow Athletic's strong pipeline contribution from buying it past the unresolved brand-safety flag",
"Draws the capacity line on summed effort estimates inside the nine manager-months",
"Names security and compliance plus financial viability as the B2B side's signature gates alongside brand safety on the consumer side"
]
},
{
"id": 10,
"prompt": "Re-running last year's alliance shortlist at Ledbetter Analytics. Of the four we greenlit: Corrowine shipped the integration and sourced 11 opportunities, Sable Point shipped nothing and their PM went quiet, Askerton got acquired by a competitor of ours in March, and Firthwell had a down round in May. Below the line last year we had three on hold and two we killed. One of the held ones, Pennard Tech, signed with our closest competitor in June and it is apparently going well. My new top three look completely different from last year's. Write me the re-ranking.",
"expected_output": "A re-ranking that names which documented triggers fired, re-runs the viability and ownership gates on the affected candidates, logs Pennard Tech as a regret-log entry, measures first-value milestone attainment, issues fresh verdicts, and dates the next review.",
"files": [],
"expectations": [
"Identifies the named re-ranking triggers that fired: the acquisition of Askerton, the funding event at Firthwell, and the slipped gate milestone at Sable Point",
"States that a reversed top three signals weak criteria rather than a changed world unless a named trigger fired",
"Confirms triggers did fire here and connects each one to the candidate it affects",
"Records Pennard Tech in a regret log as a passed-over candidate that partnered with a competitor and visibly performed",
"Treats the regret log as the honest test of where the cut line was drawn",
"Re-runs the financial viability gate on Firthwell after the down round rather than only re-scoring it",
"Re-runs the gate set on Askerton given the change of ownership by a competitor",
"Requires every held candidate to carry an explicit re-entry trigger and flags any hold that lacked one as a slow kill nobody admitted to",
"Measures the share of shortlisted alliances that hit their first-value milestone by the gate date",
"Issues fresh verdicts drawn from go, conditional-go, recycle, hold, and kill for every candidate",
"Sets the next portfolio review date and the triggers that would force an earlier one",
"Judges the prioritization decision itself rather than the partners' ongoing operational performance"
]
}
],
"trigger_queries": [
{ "query": "We have nine candidate technology alliances and capacity for three. Rank them and tell me which to drop.", "should_trigger": true },
{ "query": "Which of these five partners deserves our alliance manager's time next year?", "should_trigger": true },
{ "query": "Help me build an alliance scorecard for six named candidates.", "should_trigger": true },
{ "query": "I need to shortlist partnership candidates for FY planning.", "should_trigger": true },
{ "query": "How do I decide between three strategic partnerships when I can only do one?", "should_trigger": true },
{ "query": "partner shortlisting", "should_trigger": true },
{ "query": "alliance prioritization framework", "should_trigger": true },
{ "query": "We got inbound from four companies wanting to partner. Which ones are actually worth it?", "should_trigger": true },
{ "query": "Score these candidate ISVs against each other and give me a ranked list.", "should_trigger": true },
{ "query": "My CEO wants us to sign the hyperscaler. Can we compare it fairly against the four smaller candidates?", "should_trigger": true },
{ "query": "Re-run last year's partner shortlist - two shipped what they promised and one got acquired.", "should_trigger": true },
{ "query": "I have a list of possible partners and two people. Where do we spend the time?", "should_trigger": true },
{ "query": "Rank our candidate co-branding partners for next season.", "should_trigger": true },
{ "query": "Which creators should get the four partnership slots we can afford this year?", "should_trigger": true },
{ "query": "We cannot do all of these partnerships. Help me pick.", "should_trigger": true },
{ "query": "Build me a weighted scorecard to compare partnership candidates.", "should_trigger": true },
{ "query": "How should I weight the criteria when comparing alliance candidates?", "should_trigger": true },
{ "query": "Our partnership pipeline has 12 names on it and no way to choose between them.", "should_trigger": true },
{ "query": "Take a look at this partner scorecard and tell me if it holds up.", "should_trigger": true },
{ "query": "Draft a go/no-go memo for our candidate alliances.", "should_trigger": true },
{ "query": "which partners should we invest in", "should_trigger": true },
{ "query": "partnership ranking methodology", "should_trigger": true },
{ "query": "I need a defensible way to tell three of these five companies no.", "should_trigger": true },
{ "query": "Our steering committee reopened the partner ranking. How do I make it stick this time?", "should_trigger": true },
{ "query": "Compare these four integration partners and tell me which two we build with.", "should_trigger": true },
{ "query": "Capacity planning for the alliance team across all our candidate partnerships.", "should_trigger": true },
{ "query": "We have more partner opportunities than people. Where is the cut line?", "should_trigger": true },
{ "query": "Help me decide which technology partnerships to commit to this quarter.", "should_trigger": true },
{ "query": "Run a sensitivity check on the weights in our partner scorecard.", "should_trigger": true },
{ "query": "How do I stop the biggest logo from automatically winning our partner selection?", "should_trigger": true },
{ "query": "Everyone wants to partner with the big platform. Is that actually the right call versus the three smaller ones?", "should_trigger": true },
{ "query": "Draft the shortlist memo for our alliance candidates.", "should_trigger": true },
{ "query": "Screen these six candidate alliances before we score any of them.", "should_trigger": true },
{ "query": "What must-pass checks should candidate partners clear before we rank them?", "should_trigger": true },
{ "query": "We are picking two of seven possible brand collaborations. Walk me through it.", "should_trigger": true },
{ "query": "alliance candidate evaluation criteria", "should_trigger": true },
{ "query": "Rank partnership opportunities by value versus effort.", "should_trigger": true },
{ "query": "Which of these named companies should we actually build an alliance with?", "should_trigger": true },
{ "query": "An exec relationship is driving our partner choice and I think it is the wrong call.", "should_trigger": true },
{ "query": "I want to compare candidate partners on more than just account overlap.", "should_trigger": true },
{ "query": "Help me justify dropping a partner candidate the CRO likes.", "should_trigger": true },
{ "query": "Selecting between strategic partners for next year's plan.", "should_trigger": true },
{ "query": "partner selection scorecard with justified weights", "should_trigger": true },
{ "query": "Give me a ranked list with verdicts for our partnership candidates.", "should_trigger": true },
{ "query": "We shortlisted five partners and now leadership wants a sixth added. How do I handle that?", "should_trigger": true },
{ "query": "Should we do the deep co-build partnership or three easy integrations instead?", "should_trigger": true },
{ "query": "Our partner candidate list keeps growing and nothing ever gets killed.", "should_trigger": true },
{ "query": "Evaluate these three candidate alliances for risk before we commit to any.", "should_trigger": true },
{ "query": "How risky is building our alliance on that platform compared to the other candidates?", "should_trigger": true },
{ "query": "Portfolio review of our alliance candidates - what should change?", "should_trigger": true },
{ "query": "Which of these partnership options gives us the most for the hours we actually have?", "should_trigger": true },
{ "query": "Set up a repeatable process for choosing which partners we work with.", "should_trigger": true },
{ "query": "I have a spreadsheet of candidate partners with equal weights on everything.", "should_trigger": true },
{ "query": "Tell me which two of these six are worth a business case.", "should_trigger": true },
{ "query": "alliance shortlist", "should_trigger": true },
{ "query": "We need to choose between an analytics partner, a security partner, and a data platform partner.", "should_trigger": true },
{ "query": "Partnership investment decision across several named candidates.", "should_trigger": true },
{ "query": "Model the P&L for our reseller deal with Cavanti - margin, cost to serve, payback.", "should_trigger": false },
{ "query": "Is this one partner worth pursuing? Build me the single-partner business case.", "should_trigger": false },
{ "query": "What is our partner CAC versus our direct CAC?", "should_trigger": false },
{ "query": "Map our partner ecosystem - what partner types do we have and where are the coverage gaps?", "should_trigger": false },
{ "query": "Which partner category should we launch next, resellers or MSPs?", "should_trigger": false },
{ "query": "Sequence our partner category roadmap for the next 18 months.", "should_trigger": false },
{ "query": "Design the tier ladder for our reseller program - silver, gold, platinum criteria.", "should_trigger": false },
{ "query": "Our gold tier has 80% of partners in it. How do I fix tier compression?", "should_trigger": false },
{ "query": "Build a QBR structure for reviewing our existing channel partners.", "should_trigger": false },
{ "query": "One of our resellers has missed target three quarters running. What is the corrective ladder?", "should_trigger": false },
{ "query": "Should we list on AWS Marketplace or Azure Marketplace first?", "should_trigger": false },
{ "query": "Is an app store listing worth the engineering effort for us?", "should_trigger": false },
{ "query": "Write the deal registration policy for our co-sell motion.", "should_trigger": false },
{ "query": "Our AE and a partner both claim the same account. Who wins?", "should_trigger": false },
{ "query": "Draft the rules of engagement between direct sales and channel.", "should_trigger": false },
{ "query": "Design our partner program from scratch - are we even ready for partners?", "should_trigger": false },
{ "query": "What certification track should our resellers go through?", "should_trigger": false },
{ "query": "Plan the joint webinar and launch sequence with Tessamark now that we have signed.", "should_trigger": false },
{ "query": "Write the better-together story for our launch with Kelso.", "should_trigger": false },
{ "query": "We just signed Marisse. Write the campaign brief for her deliverables.", "should_trigger": false },
{ "query": "Which 6 of these 40 TikTok creators fit our spring campaign?", "should_trigger": false },
{ "query": "Vet this creator's audience quality before we approach her.", "should_trigger": false },
{ "query": "Draft the DM I send to this creator proposing a collab.", "should_trigger": false },
{ "query": "What should my opening position be on usage rights and exclusivity in this creator negotiation?", "should_trigger": false },
{ "query": "What KPIs should we set for this one creator campaign before launch?", "should_trigger": false },
{ "query": "Set the commission rate for our affiliate program.", "should_trigger": false },
{ "query": "Lifetime versus 12-month recurring affiliate commission - which should we offer?", "should_trigger": false },
{ "query": "Does this affiliate's Instagram caption meet FTC disclosure rules?", "should_trigger": false },
{ "query": "Build detection rules for cookie stuffing in our affiliate program.", "should_trigger": false },
{ "query": "Design the welcome email sequence for newly approved affiliates.", "should_trigger": false },
{ "query": "Audit this month's affiliate payout run before we disburse.", "should_trigger": false },
{ "query": "What metrics belong on our affiliate program dashboard?", "should_trigger": false },
{ "query": "Draft our affiliate program terms and conditions.", "should_trigger": false },
{ "query": "Write the outreach email to recruit coupon and cashback partners.", "should_trigger": false },
{ "query": "How big should our refer-a-friend reward be?", "should_trigger": false },
{ "query": "Customers are creating fake accounts to claim referral rewards. What guardrails do we need?", "should_trigger": false },
{ "query": "How do I break into partnerships from a sales role?", "should_trigger": false },
{ "query": "Write the job posting for a channel partner manager.", "should_trigger": false },
{ "query": "Design the interview loop for an alliance manager hire.", "should_trigger": false },
{ "query": "Which partnerships skill should I use for this task?", "should_trigger": false },
{ "query": "Give me a prioritization framework for our product backlog.", "should_trigger": false },
{ "query": "Should we use RICE or ICE to prioritize these features?", "should_trigger": false },
{ "query": "Help me prioritize my tasks for this week.", "should_trigger": false },
{ "query": "Rank these OKRs by impact for the quarter.", "should_trigger": false },
{ "query": "Score these acquisition targets and tell me which company we should buy.", "should_trigger": false },
{ "query": "Which of these three SaaS vendors should we buy - compare pricing and support.", "should_trigger": false },
{ "query": "Run a security review on this vendor before we sign the contract.", "should_trigger": false },
{ "query": "Build a scorecard for our marketing channels.", "should_trigger": false },
{ "query": "Rank our candidate markets for international expansion.", "should_trigger": false },
{ "query": "Which conferences should we sponsor next year with our event budget?", "should_trigger": false },
{ "query": "Prioritize these engineering projects against our roadmap capacity.", "should_trigger": false },
{ "query": "Score these job candidates against our hiring rubric.", "should_trigger": false },
{ "query": "Which suppliers should make our approved vendor list?", "should_trigger": false },
{ "query": "Help me choose between three investors for our Series B.", "should_trigger": false },
{ "query": "Rank our customer segments by expected lifetime value.", "should_trigger": false },
{ "query": "Which integrations should we build next on our own roadmap?", "should_trigger": false },
{ "query": "Compare these four project management tools and recommend one for our team.", "should_trigger": false }
]
}
references/b2b-vs-b2c-contrast.md›
# B2B vs B2C Candidate Evaluation - Sourced Contrast
## What is identical
The method's spine does not change between sides:
- Stage-gated funnel.
- Must-pass gates outside the tradeable score.
- Justified weights with sensitivity analysis.
- Effort and risk scored beside value.
- Capacity cut line.
- Verdict vocabulary.
- Portfolio review cadence.
- The memo shape.
Only the criteria content, data sources, and dominant risks change.
## What genuinely differs
| Dimension | B2B tech alliance | B2C brand/creator partnership |
| ------------------------ | ------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Currency | Pipeline, ACV, integration adoption, co-sell records | Brand equity, audience reach, incremental attributed revenue |
| Core fit construct | Resource complementarity + organizational compatibility (Sarkar et al.; Mindruta, Moeen & Agarwal 2016) | Brand fit / attitude congruence - consumer attitudes toward a pairing spill over onto each partner brand, moderated by familiarity, and not equally (Simonin & Ruth 1998) |
| Dominant risk | Channel conflict, platform envelopment, roadmap dependency | Reputational contagion, negative spillover, influencer fraud |
| Key data | Account overlap, closed-won influence samples, marketplace co-sell metrics | Audience overlap, brand-lift and incrementality tests, fraud-screening scores |
| Decision owners | Partnerships, product, sometimes sales engineering | Marketing and brand leadership |
| Measurement weakness | Overlap-as-proxy; vendor-funded benchmarks | Self-reported brand-lift surveys; incrementality tests that lack clean holdouts |
| Signature pass/fail gate | Security/compliance, financial viability | Brand-safety flags and fake-engagement share |
Additional sourced B2C notes:
- The co-branding evidence is fragmented: one meta-analysis (197 effect sizes) found success drivers context-dependent, and studies disagree on whether brand fit or product fit dominates. Another study found co-branding reduced brand-esteem on average - high-esteem brands should be the most cautious candidates-recipients, not the most eager.
- Creator vetting is chronically under-resourced in practice: surveyed brands widely report spending under half an hour vetting a single influencer, and a significant minority conduct no due diligence at all (industry surveys - directional, not audited). Treat a candidate roster that arrived unvetted as unscreened, not pre-screened.
## Hybrid cases
A consumer-facing ISV, a fintech co-brand, a retail-tech alliance: these carry both currencies. Run both evaluation logics - pipeline math and brand-fit/contagion screening - and gate on the stricter of the two. Do not average a brand-safety concern into a pipeline score; gates never average.
## Vendor-statistic hygiene
Almost every headline partnership statistic originates from a vendor selling the software the statistic justifies. Before letting a number into the memo:
- Demand sample size, recruitment method, and whether it measures audited outcomes or self-reported perception.
- The widely recycled "deals are 53% more likely to close with a partner involved" is a self-reported convenience survey of roughly 500 professionals; the same vendor's own network data showed roughly an 11.7% win-rate lift. Quote the smaller, harder number if either.
- The "50-70% of alliances fail" line spans 20-80% across sources depending on the definition of failure, and a credible re-analysis argues 25-30% (18-20% for firms with an alliance-management function). Present the spread, never a single settled rate. The most-repeated "60-70%" phrasing traces to Hughes and Weiss (Vantage Partners), "Simple Rules for Making Alliances Work," Harvard Business Review, November 2007 - which cites no study, sample, or definition behind the figure. A separate, more traceable data point measures something different: McKinsey's 2004 survey of roughly 30 companies found 70% had major alliances underperforming and needing restructuring, not outright failed.
- Upgrade a benchmark to decision-grade only on independent replication or the vendor's own outcome data - never its survey headline.
- Flag provenance in the memo for every number kept: peer-reviewed, analyst, vendor first-party, or vendor survey.
references/gates-and-governance.md›
# Stage Gates, Workshops, and Governance
## The funnel
Alliance teams run candidate selection as a stage-gated funnel, mirroring corporate-development screening:
1. **Universe → longlist** - every plausible named candidate, sourced from customer data, overlap reports, marketplace neighbors, inbound requests (inbound goes in, but never ranks on enthusiasm alone).
2. **Longlist → screened list** - the must-pass gates, staged: desk checks across the whole longlist, week-long reviews on survivors only. Never scorecard a candidate that fails a gate.
3. **Screened list → ranked shortlist** - scorecard, effort, risk composite, sensitivity analysis, capacity cut.
4. **Shortlist → business case** - light case per survivor: joint value hypothesis, first-value milestone, effort estimate, risk summary, staged-commitment plan.
5. **Business case → go/no-go** - a dated decision meeting with the sign-off owner, not an email thread that trails off.
## Verdict vocabulary
Use the standard gate verdicts, and treat each as a business decision - continue funding this investment or not:
- **Go** - fund, assign an owner, date the first-value gate.
- **Conditional-go** - fund only after a named condition clears (a security attestation, a signed data agreement, a pilot result). Name the condition and its deadline.
- **Recycle** - promising but the case is not ready; send back with the specific gap named.
- **Hold** - parked with an explicit re-entry trigger (funding event, roadmap ship, capacity freeing up). A hold with no trigger is a slow kill nobody admitted to.
- **Kill** - out, with the reason recorded so the same candidate does not relitigate itself next quarter.
## Joint value proposition workshop
The strongest pre-signature test for a top-ranked candidate, where both sides can be convened:
- Invite matched cross-functional stakeholders from both companies - sales with sales, product with product, marketing with marketing. The matched-persona structure is what makes the output stick internally on both sides.
- Agree the specific customer problem being solved jointly before anything else; misaligned target personas doom the joint value proposition and everything downstream.
- Then build the proposition from each side's concrete value components - what each party adds that the customer can verify.
- Exit test: can the room state, in one sentence, what a named customer gets from the combination that neither side delivers alone? A candidate that cannot pass this fails regardless of scorecard rank.
Two ways to test joint value before signing, and the choice is not close:
- **Convened workshop** (default): costs a week of scheduling, matched stakeholders on both sides, and agreement on what roadmap detail may be shared. Buys the only pre-signature evidence that the value story survives contact with the counterparty.
- **Internal hypothesis**: costs an hour and no coordination. Buys a testable claim and nothing more - mark it unvalidated and issue a conditional-go with the workshop as the condition.
- evidence strength, and efficiency alike: `workshop > internal hypothesis` - the cheap option cannot produce that evidence at any effort level, so its low cost never promotes it. Take it only when the counterparty genuinely cannot be convened before the decision.
- compliance cost: `workshop > internal hypothesis` - the workshop puts unreleased roadmap in front of a counterparty, so it runs under an NDA, and a competitor-adjacent candidate needs counsel to scope what may be discussed before the room is booked.
## Governance for survivors
- Stand the alliance up with a joint steering committee: co-chaired by one senior executive per side, a small even split of voting members, an agreed escalation path. Design it at go/no-go time - bolting governance on after the first conflict is the expensive order.
- Give every alliance - including the "strategic" ones - leading-indicator milestones with dates. Labeling an alliance strategic to exempt it from measurement is a documented failure pattern of the profession.
## Portfolio review cadence
- Re-rank the portfolio on a fixed cadence (quarterly or semi-annual) plus event triggers: a candidate or partner is acquired, a platform changes its terms, a gate milestone slips, a serious competitor alliance forms.
- Expect additions, substitutions, and removals - firms demonstrably reconfigure alliance portfolios in response to performance gaps, and pruning is a sign the process works, not that the original ranking failed.
- The strongest institutional argument for keeping this process alive: firms with a dedicated alliance function and codified process reported 63% alliance success versus 50% without one (Kale, Dyer & Singh, 2002; 200 firms, 1,572 alliances). That is the closest thing this field has to a validated causal driver. The ranking memo, verdict log, and review cadence are that function's working artifacts.
references/risk-and-dependency-screens.md›
# Risk and Dependency Screens
Risk feeds the ranking as scores and gates - never a downstream compliance annex read after the decision.
## Must-pass gates (pass/fail, outside the score)
A gate failure removes the candidate regardless of composite score. Each gate decision cites its evidence in the memo.
Gates are never ranked against each other on value - every one of them is pass/fail, and a gate that can be skipped for being low-value is a scorecard criterion in disguise. Only their run order is ranked, by what each costs to check:
- check effort: `legal/antitrust == security attestation == creator fraud > exit path == three-laws > financial viability` - the top three tie because each is a week on somebody else's calendar: counsel review, a TPRM questionnaire, a paid audience audit. Exit path and three-laws tie one rung below because each needs a short artifact drafted before it can be judged at all - a contract sketch, a JVP sketch - an hour or two of your own time. Financial viability sits lowest on its own: funding history, layoffs and filings read off public sources in minutes.
- Run them in the reverse of that order: desk gates across the whole longlist first, the week-long reviews only on survivors. The staged default and its compliance trade-off are in SKILL.md § Method choices.
| Gate | Fails when | Evidence to check |
| --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------- |
| Financial viability | Credible risk the candidate folds or is fire-sold inside the alliance horizon | Funding history, layoffs, filings, churn signals |
| Security/compliance posture | Candidate cannot meet the data-sharing bar the alliance needs (SOC 2 / ISO 27001-class attestations, TPRM questionnaire) | Attestation status, breach history |
| Legal/antitrust exposure | Exclusivity or data terms create regulatory or IP exposure | Counsel review of proposed terms |
| Brand safety (B2C) | Unresolved brand-safety flag, active controversy, values conflict | Vetting report, press scan |
| Creator fraud (B2C) | Fake-follower or engagement-fraud share above the stated tolerance | Audience audit from a fraud-screening tool |
| Three-laws screen | The combination fails any of Gomes-Casseres's laws: creates no value beyond standalone, cannot operationally act as one, or has no workable value split | JVP draft, integration feasibility, commercial term sketch |
| Exit path | No plausible wind-down without stranding customers | Contract sketch, data-portability check |
## Relational vs performance risk (Das & Teng)
Score both, separately, per candidate - they are different failure modes with different mitigations:
- **Relational risk** - the partner behaves opportunistically or stops cooperating. Raised by: competing roadmaps, asymmetric dependence, no prior ties, misaligned incentives. Mitigated by governance and staged commitment, not by enthusiasm.
- **Performance risk** - everyone cooperates and the alliance still misses its objectives. Raised by: unproven market, hard integration, weak joint value proposition. Mitigated by pilots and milestone gates.
A candidate high on both is a kill or a small staged pilot - never a flagship bet.
## Ecosystem dependency risk (Adner's Wide Lens)
Two risks standard scorecards miss entirely; ask both questions per candidate:
- **Co-innovation risk** - does joint success require the candidate (or a third party) to build something that does not exist yet? Score how much of the value depends on someone else's unshipped roadmap.
- **Adoption-chain risk** - must intermediaries (the candidate's sales force, resellers, retailers, platform reviewers) adopt before the end customer ever sees the value? Every unwilling link discounts the expected value.
Sketch the value blueprint for top candidates: every actor whose action is necessary for the customer to get the value. A blueprint with three dependent actors you cannot influence is a materially riskier alliance than its scorecard suggests.
## Platform-envelopment risk (Eisenmann, Parker & Van Alstyne)
For any candidate whose platform you would build on, score the risk of being commoditized by the partner you are joining: the platform owner can absorb your functionality by bundling it, or quietly re-rank and substitute you even while technical access stays open.
Warning signs:
- The platform ships adjacent first-party features.
- Its marketplace terms allow self-preferencing.
- Your value sits in a thin layer it could replicate.
High envelopment risk does not forbid the alliance. It demands staged commitment, contractual safeguards, and a stated dependency ceiling in the memo.
## B2C-specific screens
- **Reputational contagion** - consumer attitudes toward a brand pairing spill over onto each partner, and not symmetrically (Simonin & Ruth, 1998); a co-branding meta-analysis found high-esteem brands are the most exposed to dilution. Model at least one negative-spillover scenario per shortlisted candidate: partner scandal, product failure, values controversy - and what it costs you.
- **Creator fraud vetting** - run as pass/fail before any scoring. Screen for fake followers, engagement pods, bot activity, undisclosed past controversies. One investigative firm reports roughly one in four vetted influencers carried significant risk flags (vendor-reported figure - treat as directional). Demand incrementality-tested metrics, not vanity reach.
## Feeding risk back into the rank
Fold the scored (non-gate) risks into a risk composite per candidate and show it beside the value and effort columns. Two candidates with equal expected value and different risk composites are not tied - say which risk drives the gap and what mitigation would close it.
references/scoring-criteria-and-weights.md›
# Scoring Criteria and Weighting Mechanics
## Criteria menu
Pick 6-10 criteria; more dilutes the weights into noise. The recurring families across academic and practitioner treatments:
| Criterion | What it measures | Evidence anchor |
| --------------------------------------- | ---------------------------------------------------------------------------------- | --------------------------------------------------------------------------------- |
| Resource complementarity | Candidate brings resources/capabilities you lack; value only exists in combination | Sarkar et al.; Doz & Hamel's co-specialization |
| Organizational compatibility | Similar norms, decision speed, operating cadence, status tier | Mindruta, Moeen & Agarwal (2016): different in resources, similar in institutions |
| Strategic-goal fit | Direct line from this alliance to the stated goal of the ranking | Interview output; no goal, no score |
| Market/ICP overlap | Shared target accounts or audience, evidenced not assumed | Account-overlap data, closed-won sample, audience data |
| Product complementarity vs substitution | Adjacent in the customer workflow, not competing for the same budget | Substitution flips a candidate into a channel-conflict risk |
| Joint value proposition strength | A customer problem the combination solves that neither side solves alone | Gomes-Casseres's first law, tested in the JVP workshop |
| Relational capital / prior ties | Prior collaboration, existing champions on both sides | Prior ties lower expected governance cost (Reuer & Ariño) |
| Alliance readiness of the counterparty | Dedicated partner function, named owner, history of shipped partnerships | Kale, Dyer & Singh: the dedicated function is the validated success driver |
| Effort to first value | Integration cost, enablement load, legal complexity, alliance-manager hours | The denominator; see below |
| Risk composite | Output of the risk and dependency screens, folded back in | Scored separately, then shown beside value and effort |
## Weighting mechanics
Weight priority across the families above, and the ordering to state in the memo:
- predictive value, and efficiency alike: `complementarity == compatibility > strategic-goal fit > JVP strength == alliance readiness > product complementarity > prior ties > market/ICP overlap`
- Complementarity and compatibility tie because they are one two-sided construct, not two criteria that happened to land level: the evidence says partners must be different in resources _and_ similar in institutions, and neither half predicts anything alone.
- JVP strength and alliance readiness tie because both are necessary and neither substitutes for the other - one says the alliance is worth doing, the other says the counterparty can actually do it, and a candidate failing either is worth nothing however well it scores on the other.
- Product complementarity outranks prior ties rather than tying with it: substitution flips a candidate into channel conflict and destroys value outright, while prior ties only lower expected governance cost - and are the same relationship halo the Failure Modes table warns about.
- data effort: `market/ICP overlap > alliance readiness == prior ties > complementarity == compatibility == goal fit` - overlap is the one family that usually needs a licensed dataset or a RevOps pull, and it still ranks last on what it predicts. Readiness and prior ties tie because one pull answers both: the counterparty's public partnership history beside your own CRM relationship records. The last three tie at near-zero because all three are scored from desk research against a goal the Interview already fixed - no external data, the same hour of work.
- Weight complementarity and compatibility highest - they are the two constructs with peer-reviewed support. Everything else is negotiable; these two anchoring the top is not.
- Justify every weight in one written line ("overlap 15% - our data covers only 40 deals"). An unjustified weight is an equal weight in disguise.
- Reject default-equal weights outright. Equal weights are the documented practitioner failure, not a neutral choice.
- Segment weights by alliance type when the candidate list mixes types - a technology integration and a co-sell alliance must never share one weight set.
- Cap account-overlap's weight. Overlap flags warm accounts; no independent study shows it predicts durable alliance success. The category's own network data showed roughly an 11.7% win-rate lift while its survey marketing claimed 53% - treat every overlap-derived number as a hypothesis.
- Score effort separately from value and divide: rank on expected value against effort. A dazzling candidate at triple the integration cost of the runner-up is often the wrong pick.
## Why the machinery needs fixing
Weighted scorecards inherit two documented defects from the decision-analysis literature:
- Adding or removing a near-duplicate candidate can reverse the ranking of the others (Belton & Gear, 1983).
- The weights themselves lack any axiomatic foundation - Dyer (1990) called the resulting rankings "arbitrary".
Practitioner scorecards with equal weights and subjective 1-5 scores are more exposed, not less. The scorecard stays useful only with three guardrails:
1. **Must-pass gates outside the score.** Viability, security, legal, brand safety are pass/fail, never tradeable points.
2. **Sensitivity analysis.** Shift each weight ±20%, one at a time, and re-rank. Report stable positions and flips in the memo. If the top two swap under a plausible shift, present both orderings and the judgment call - never hide the instability.
3. **Tie-bands.** Composite scores within ~5% of each other are a tie. Break ties with documented qualitative judgment (goal fit, portfolio balance, timing), not decimals.
The weight-distribution options and their ordering live in SKILL.md § Method choices; a formal multi-criteria method sits third there, below justified and type-segmented weights. Inside that rung, prefer best-worst method over AHP - fewer pairwise comparisons, less rank-reversal exposure. Say plainly either way that no method substitutes for good criteria.
## Scoring discipline
- Score 1-5 per criterion with a written anchor for each level ("5 = shipped a comparable integration in the last 18 months"), so two scorers land on the same number.
- Score from evidence: overlap reports, closed-won samples, public filings, marketplace listings, audience audits. Mark every evidence-free score as a guess, visibly.
- Score politically protected candidates on the same sheet. The point is not to block the executive's candidate - it is to make its true rank visible before capacity is spent.
- Record scorer and date; a scorecard nobody signed is a scorecard nobody defends at the gate review.
references/worked-examples.md›
# Worked Examples
Both examples are fictional companies; the mechanics are the point.
## Positive example - mid-market B2B SaaS, six named candidates
Scenario: a workflow-automation vendor (sales-led, ~$40k ACV) must decide where two alliance managers' capacity goes next year. Goal: partner-sourced pipeline in the enterprise segment. Capacity: roughly eight alliance-manager-months a year, which the team also describes as "two managed alliances".
**Gates.** Six candidates enter. One - a fast-growing point solution - fails the viability gate (two down-rounds, key-team departures, evidence cited). Five proceed.
**Scorecard.** Criteria and justified weights:
- Complementarity 25% - "value only exists combined - their data layer, our workflow engine".
- Compatibility 20% - "enterprise deals die on mismatched deal cadence".
- Goal fit 15%.
- Verified overlap 10% - "capped - overlap sample is 60 accounts".
- JVP strength 15%.
- Alliance readiness 15%.
| Candidate | Value composite | Effort (mgr-months/yr) | Value ÷ effort | Risk composite | Rank |
| ------------------------------------ | --------------- | ---------------------- | -------------- | ----------------------------------------------------- | ---- |
| DataFabric (data platform) | 4.3 | 4 | 1.08 | Medium | 1 |
| SecureID (identity vendor) | 3.1 | 3 | 1.03 | Low | 2 |
| FormaCRM (CRM vendor) | 3.6 | 5 | 0.72 | Medium | 3 |
| CloudSuite (hyperscaler marketplace) | 4.1 | 7 | 0.59 | Medium-high (envelopment) | 4 |
| BrightBI (analytics) | 2.9 | 6 | 0.48 | High (co-innovation: value needs their unshipped API) | 5 |
Read the CloudSuite row before anything else. It is second of five on value and fourth after dividing, because 7 manager-months buys 4.1 where SecureID's 3 buys 3.1. Ranking this table by the value column alone - the intuitive move, and the one the negative example below is built on - puts CloudSuite in a two-slot shortlist and leaves SecureID out.
**Capacity line.** Drawn on summed effort, not headcount: DataFabric (4) plus SecureID (3) is 7 of the 8 available manager-months; adding FormaCRM (5) would need 12. The headcount phrasing and the effort sum happen to agree here - they stop agreeing the moment one candidate costs double another.
**Sensitivity.** ±20% shifts on every weight leave the top two intact. DataFabric and SecureID sit 4% apart, inside the 5% tie-band, but both fit inside capacity, so this tie needs no breaking - say so rather than manufacturing a tiebreaker.
The live question is CloudSuite: it enters the shortlist only if a scoped first phase cuts its effort below ~4 manager-months. Positions 4-5 are 21% apart and are not a tie.
BrightBI is the co-innovation candidate the ratio structurally starves, and here the last place is the right answer twice over: the goal is pipeline rather than a product gap, and the API the joint value depends on carries no ship date. An undated dependency earns a recycle, never a reserved capacity slot.
**Verdicts.**
- DataFabric: go - first-value milestone: integration live plus three joint opportunities in two quarters.
- SecureID: go - first-value milestone: identity integration live plus two joint enterprise opportunities in two quarters.
- FormaCRM: hold - re-entry trigger: an alliance-manager hire.
- CloudSuite: hold - re-entry trigger: a scoped phase-one under 4 manager-months, with envelopment safeguards written into the marketplace agreement.
- BrightBI: recycle - re-score when the dependent API ships.
One kill, two holds, one recycle: the ranking made a real decision.
Why this works:
- The capacity line forced a choice.
- Every weight has a reason.
- Effort divided value instead of sitting in a column nobody used.
- The gate failure was not traded away.
- The instability that existed was disclosed.
- Every non-go candidate left with a trigger, not a shrug.
## Negative example - the logo-collecting scorecard
Same vendor, same six candidates, the way this goes wrong in practice:
| Candidate | Brand (20%) | Market size (20%) | Overlap (20%) | Exec relationship (20%) | Revenue potential (20%) | Total |
| ----------------- | ----------- | ----------------- | ------------- | ----------------------- | ----------------------- | ----- |
| CloudSuite | 5 | 5 | 4 | 5 | 5 | 4.8 |
| MegaCorp Platform | 5 | 5 | 3 | 4 | 5 | 4.4 |
| DataFabric | 3 | 3 | 4 | 2 | 4 | 3.2 |
What is wrong, line by line:
- **Equal weights, no justifications** - nobody decided anything; the format decided. This is the arbitrary-weights defect in its purest form.
- **Brand and executive relationship as 40% of the score** - logo halo and prior personal ties are the documented revealed criteria that crowd out fit; the evidence says complementarity and compatibility should dominate, and neither appears at all.
- **No effort column** - CloudSuite at 7 manager-months and DataFabric at 4 are scored as if capacity were free.
- **No gates** - MegaCorp Platform's pending security review would have been a conditional-go at best; here it ranks second on charm.
- **No risk scoring** - CloudSuite's envelopment exposure is invisible; the scorecard is recommending building on a platform that ships competing features.
- **No sensitivity check, no tie-band** - the 4.8 vs 4.4 gap reads as decisive and would not survive a modest re-weighting.
- **No capacity line, no verdicts** - all six candidates stay "in play", which means the loudest sponsor wins later, off the record.
- **The revenue-potential column is unevidenced** - TAM-flavored guesses scored as data.
The output looks rigorous - a table, percentages, a winner. It is a ratified gut feeling, and the two candidates it crowns are the famous one and the sponsored one. When a user brings a scorecard shaped like this, rebuild from the gates up rather than tuning its numbers.
SKILL.md›
---
name: alliance-prioritization
description: Rank named candidate strategic alliances and technology partnerships by expected value, effort, and risk, to decide where partner-team capacity goes - must-pass screens, a weighted alliance scorecard with sensitivity analysis, co-innovation and platform-dependency risk checks, a ranked shortlist, and a go/no-go recommendation memo. Covers B2B tech alliances and B2C brand and creator partnerships. Use whenever the user mentions alliance prioritization, partnership ranking, partner shortlisting, or asks which of several named partners deserves investment, even if they never say prioritization. Portfolio-level comparison only. Do NOT use for modeling one partner's P&L - use mbfinotti/partnerships-skills@partner-economics instead.
license: MIT
metadata:
author: Maya-Beth Finotti
version: "1.0.2"
---
# Alliance Prioritization
Rank specific, named candidate alliances and technology partnerships against each other, and decide which ones the team invests in. No single canonical ranking method owns this discipline - say so plainly rather than invent one. What the evidence actually supports:
- **A dedicated alliance function and structured process** is the field's best-validated success driver: 63% vs 50% alliance success across 1,572 alliances (Kale, Dyer & Singh, 2002). A repeatable prioritization process is part of that function.
- **Complementarity plus compatibility** is the best-validated fit construct: partners different in resources, similar in organizational institutions (Sarkar et al.; Mindruta, Moeen & Agarwal, 2016). Weight these highest.
- **Weighted scorecards are known-fragile machinery**: rank reversal (Belton & Gear, 1983) and arbitrary weights (Dyer, 1990). The fixes - justified weights, sensitivity analysis, must-pass gates kept outside the tradeable score - are mandatory here, not optional polish.
- The famous "50-70% of alliances fail" figure is unsettled: published estimates span 20-80% depending on how failure is defined, and one credible re-analysis puts it at 25-30%. Never quote it as settled fact.
## Interview
Ask before ranking anything. One question per message; offer multiple-choice options; skip whatever the user already answered.
- How many named candidates, and who are they? (Categories instead of names → the ecosystem sibling; a single candidate → the economics sibling.)
- What strategic goal must the ranking serve: pipeline, product-gap fill, market entry, retention, platform presence, brand reach?
- How many alliances can the team actually run - headcount, current load per alliance manager? This number becomes the shortlist cut line.
- What does the existing alliance portfolio look like: active alliances, concentration, redundancy or conflict with the candidates?
- What data exists per candidate: account-overlap data, closed-won history, integration usage, marketplace co-sell records, prior-relationship history?
- Any candidate that is politically protected - an executive relationship, board pressure, a prior tie? Name it now; it gets scored like the rest, visibly.
- Who signs off, and does a gate process or steering committee already exist?
- B2B tech alliances, B2C brand/creator partnerships, or a hybrid list?
- By what date must this ranking land - a board slide next week, a signed alliance this quarter, next year's plan?
- One-off win or compounding asset: a shortlist for this decision only, or a scorecard the team re-runs every portfolio review?
- What is the effort ceiling on the analysis itself - analyst hours available, whose data you can request, whose sign-off you can spend?
Those last three re-order the method choices in Brainstorming the ranking, so ask them before proposing any criteria set.
- A near-term date promotes the speed-to-first-value criteria set and staged gates.
- A compounding mandate promotes the complementarity-led set and segmented weights that survive re-use.
- An effort ceiling under about a day removes the scorecard entirely - rank on gates plus effort-to-first-value and say that is what you did.
- The goal answer carries the same weight: a product-gap or market-entry goal reserves capacity for a deep co-innovation candidate that value-per-effort would otherwise starve.
## Workflow
Every method choice - criteria set, weight distribution, gate strictness, where the capacity line falls - is ranked with its trade-off in Brainstorming the ranking. Open the relevant menu at the step that needs it, put its default and the runner-up to the user, and wait. Never let an assumption harden into the ranking unstated.
1. Run the Interview. Confirm the input is two or more named candidates and one stated strategic goal.
2. Fix the capacity number before scoring anything. A ranking without a cut line decides nothing - every candidate just gets a flattering ordinal.
3. Pre-filter each candidate with the ally-versus-acquire-versus-build test (Dyer, Kale & Singh, 2004): confirm an alliance is even the right mode for the synergy sought. Wrong-mode candidates leave the list here.
4. Run the must-pass gates, staged by default - desk checks across the whole longlist, expensive reviews on survivors only: financial viability, security/compliance posture, legal exposure, brand safety for B2C, and the Gomes-Casseres three-laws screen. A gate failure removes the candidate - a strong logo can never buy its way past a gate. See [references/risk-and-dependency-screens.md](references/risk-and-dependency-screens.md).
5. Build the scorecard for survivors: criteria anchored on resource complementarity and organizational compatibility, weights individually justified in writing, never default-equal. Use account-overlap data as one capped input, not the ranking spine. See [references/scoring-criteria-and-weights.md](references/scoring-criteria-and-weights.md).
6. Score effort alongside value: integration cost, enablement load, alliance-manager hours. Rank candidates on `value ÷ effort > value composite > logo attractiveness` - say that ordering in the memo, because the value column alone is what the room will read otherwise.
7. Score risk explicitly per candidate: relational vs performance risk (Das & Teng), co-innovation and adoption-chain risk (Adner's Wide Lens), and platform-envelopment risk for any candidate whose platform you would build on (Eisenmann, Parker & Van Alstyne).
8. Run the sensitivity analysis: shift each weight ±20% and re-rank. Report which rank positions survive and which flip - a ranking that only holds under one weight set is an opinion wearing a spreadsheet.
9. Draw the capacity line where the summed effort estimates exhaust available alliance-manager months, falling back to the raw headcount number when candidate efforts are comparable. Spend any co-innovation slot reserved before scoring inside that budget, never on top of it - see What the ordering starves. Everything above the line is the shortlist; everything below gets an explicit verdict - hold, recycle, or kill - with a re-entry trigger.
10. Draft a light business case per shortlisted candidate: joint value hypothesis, first-value milestone, effort estimate, risk summary. Deep unit economics hand off to the economics sibling.
11. Where both sides can be convened, pressure-test the top candidates in a joint value proposition workshop; a candidate that cannot articulate joint customer value fails here regardless of score. See [references/gates-and-governance.md](references/gates-and-governance.md).
12. Present the deliverable section by section - gates, scorecard, ranking, sensitivity, shortlist, verdicts - and validate each with the user before drafting the next. Revise on pushback.
13. Issue go/no-go verdicts per candidate - go, conditional-go, recycle, hold, kill - and require explicit user approval of the full memo before finalizing.
14. Check the pass threshold below; iterate the criteria, weights, or gates until every check holds.
15. Set the portfolio review cadence and re-ranking triggers: a candidate's acquisition, a platform policy change, a gate-milestone slip, a new entrant. Rankings decay; date the next review.
16. If your harness has persistent memory, memorize the approved ranking, the weights and their justifications, and each go/no-go verdict so later runs and portfolio reviews start from them.
17. If you can browse the web, verify each candidate's current state - ownership, funding, platform policies, recent incidents - before finalizing; stale candidate facts are a leading cause of dead-on-arrival shortlists. Otherwise flag candidate facts as user-supplied and unverified.
## Brainstorming the ranking
Enter brainstorming mode before scoring anything - never open with a finished table. Run it the way alliance teams actually run candidate selection: a gated debate, not a single answer.
- Ask one question per message throughout; never batch questions.
- Widen the longlist before narrowing it. Source candidates from customer data, overlap reports, marketplace neighbors, and competitor partner directories - inbound requests enter the list but never rank on enthusiasm alone.
### Method choices, ranked
The same discipline the scorecard applies to candidates applies to the method: rank by decision quality bought per analyst hour, never by what is cheapest to run. Open the relevant menu at the step that needs it, put the default and the runner-up to the user with the trade-off, then wait.
**Criteria set** - default: complementarity-led.
- **Complementarity-led**: resource complementarity and organizational compatibility anchor the weights. Costs an hour of criteria design, scored from desk research and closed-won samples the team already holds. Buys the only fit construct with peer-reviewed support, and the ranking most likely to survive the sensitivity re-run.
- **Speed-to-first-value-led**: weights integration effort, prior ties, counterparty alliance readiness. Costs near-zero - every input is already in the team's head. Buys the answer to which alliance ships first, not which one is worth having; the shortlist decays within two quarters.
- **Reach-led**: weights account or audience overlap, market size, brand. Costs a week plus a licensed overlap dataset or audience audit, and RevOps coordination to get it. Buys the least - overlap flags warm accounts with no evidence it predicts alliance success, and market size and brand are the door logo halo walks through. Keep these as capped inputs inside another set rather than a set of their own.
- analyst effort: `reach-led > complementarity-led > speed-to-first-value`
- decision quality, and efficiency alike: `complementarity-led > speed-to-first-value > reach-led`
**Weight distribution** - default: individually justified weights.
- **Individually justified weights**: one written line per weight. Costs an hour, and the justifications double as the memo's audit trail. Kills the arbitrary-weights defect at its source.
- **Type-segmented weight sets**: one set per alliance type when the list mixes technology integrations with co-sell alliances. Costs a few hours more plus a defensible split of the list; buys nothing on a homogeneous list except another sheet to maintain.
- **Formal best-worst method**: preferred over AHP - fewer pairwise comparisons, less rank-reversal exposure. Costs a half-day of structured elicitation per stakeholder and the coordination to convene them. Buys the most defensible weight set on the menu, which is exactly why efficiency starves it - see What the ordering starves for when to promote it anyway.
- **Equal weights**: rejected rather than ranked - near-zero effort, negative value. The format decides instead of the room; this is the documented practitioner failure mode.
- analyst effort: `best-worst > type-segmented > justified > equal`
- decision quality, and efficiency alike: `justified > type-segmented > best-worst > equal`
**Gate strictness** - default: staged.
- **Staged**: desk-checkable gates on the whole longlist (viability signals, public security posture, the three-laws screen), the expensive ones - counsel review of terms, TPRM questionnaire, audience audit - only on survivors. Costs hours across the longlist and a week on the few that reach the expensive tier.
- **Full gate set on every longlist candidate**: costs a week or more, most of it spent on candidates a five-minute check would have removed. Reaches the same shortlist, later. Justified only on a candidate list small enough that staging saves nothing.
- **Gates on finalists only**: costs near-zero up front and is worth less than nothing - scoring and socializing a candidate before its legal or security gate is exactly what makes a gate tradeable, and a candidate the room has already crowned is politically near-impossible to remove.
- analyst effort: `full set > staged > finalists-only`
- decision quality: `staged == full set > finalists-only` - tied because staging changes when the expensive review runs, never what it returns; both remove the same candidates and reach the same shortlist.
- efficiency: `staged > full set > finalists-only`
- compliance cost: `finalists-only > staged == full set` - gating last means roadmap and data have already been shared with a candidate counsel may reject, so the review it triggers arrives after the point where the decision could be reversed cheaply. The other two tie because both put counsel and security in front of a candidate before it is scored or socialized; only the number of candidates reviewed differs, and that is an effort difference, not a reversibility one.
**Where the capacity line falls** - default: effort-sum.
- **Effort-sum line**: cut where the candidates' summed effort estimates exhaust available alliance-manager months. Costs near-zero once step 6 has scored effort. The only line that notices two cheap alliances and one expensive one are not interchangeable.
- **Headcount line**: cut at the capacity number from the Interview. Costs near-zero, and is enough when candidate efforts are comparable - it assumes every alliance costs one manager the same, which is rarely true.
- **Score-gap line**: cut at the biggest break in composite scores. Costs near-zero and buys the least: scores within ~5% are a tie, and the gap moves under any re-weighting, so the line lands wherever the weights happened to fall. Use it to break a tie under an already-fixed capacity number, never as the line itself.
- decision quality, and efficiency alike: `effort-sum > headcount > score-gap` - effort is near-zero for all three, so only what each buys separates them.
Two things here stay deliberately unranked. The must-pass gates are pass/fail and non-tradeable, and the verdicts - go, conditional-go, recycle, hold, kill - are outcomes the evidence assigns, not options an analyst picks between. Ordering either would manufacture exactly the false precision the sensitivity analysis exists to undo.
### What the ordering starves
A value-per-effort ratio always underweights whatever is expensive _and_ strong. Name both instances out loud rather than letting the arithmetic quietly retire them.
- **In the method menu: formal best-worst elicitation.** It produces the most defensible weight set on offer - structured, attributable, least exposed to rank reversal - and loses every efficiency round to justified weights because it costs a half-day per stakeholder plus the coordination to convene them. Promote it when the weights themselves are what the room is arguing about: a sign-off owner who distrusts the sheet, a steering committee that has already reopened one ranking, or a scorecard the team will re-run every portfolio review. It buys agreement on weights, never precision in them.
- **In the candidate list: the deep co-innovation alliance.** A partnership whose value needs both sides to build something loses twice to a quick logo-swap integration - high effort in the denominator, plus a co-innovation discount on value it has not yet been allowed to prove. A portfolio ranked only on the ratio fills with cheap integrations and never places the one bet that changes the product. Promote it when the stated goal is product-gap fill or market entry rather than pipeline, and when the unshipped piece sits on the two parties' own dated roadmaps rather than a third party's. Promote it by reserving a slice of capacity before scoring starts - never by adding points to its composite, which is the false precision the sensitivity analysis exists to undo.
### Re-ranking and deleting against this user
Every ordering above is a default for a team with no unfair advantage, and it shifts with context and with who runs it. Re-rank against what you already know about this user: a licensed account-overlap dataset drops reach-led criteria from a week to an hour (still capped as an input, never the spine); an executive who has already picked a favourite raises the value of justified weights and the sensitivity re-run, because those are the artifacts that survive the argument.
Then delete outright whatever the user's constraints rule out, and say which option you deleted and why. A ruled-out option demoted to the bottom of a menu comes back later as scope somebody re-argues:
- No licensed account-overlap data and no audience-audit budget - delete reach-led criteria. Stripped of the dataset, its remaining inputs are market size and brand, the two the logo halo travels on.
- An executive who has already picked a favourite - delete gates-on-finalists-only. Its one defence is that nobody has committed yet, and somebody has.
- A longlist of two or three candidates - delete staged gates and run the full set on everyone. Staging exists to avoid spending a week on candidates a desk check removes; at that size there is nothing to save.
Then, before anything is presented:
- Name the assumptions out loud before scoring: which candidate the room already wants, which number nobody can evidence, which effort estimate is a guess.
- Argue the strongest case _against_ your own top-ranked candidate before presenting it; logo halo and executive relationships are the documented ways this exercise fools itself.
- Present the deliverable one section at a time - gates, then scorecard, then ranking, then sensitivity, then verdicts - and validate each before drafting the next.
- Gate finalization on the user's explicit approval of the assembled memo.
## The Ranked Shortlist Memo
Deliver the engagement as this artifact - the decision record the user takes to their sign-off owner:
```
ALLIANCE SHORTLIST - <scope>, <date>
Goal : strategic goal served + the capacity line and where it came from
Candidates : N considered; gate failures named, each with the failed gate and evidence
Scorecard : criteria and weights, one-line justification per weight
Ranking : full ranked table - composite score, effort, risk score per candidate
Sensitivity : which weight shifts reorder the top of the table, and which never do
Shortlist : candidates above the line, each with verdict and its reasoning
Next gates : per shortlisted candidate - first-value milestone, owner, gate date
Below the line: each candidate's verdict (hold/recycle/kill) and its re-entry trigger
Open questions: data that was missing, and what result would change the ranking
```
Three rules govern the memo:
- Every score traces to evidence a reviewer can check - "no shared closed-won accounts in a 40-deal sample" is reviewable, "weak overlap" is not.
- Every quoted benchmark carries a provenance flag - most published partnership statistics come from vendors that profit when the number looks good.
- The open-questions section is mandatory: a ranking with no stated uncertainty has stopped asking.
## Pass Threshold
Ship nothing until all of these hold; iterate until they do:
1. **One criteria set, justified weights.** Every candidate scored on the same criteria; every weight carries a written one-line justification; no equal-weight default anywhere.
2. **Sensitivity-tested.** The ±20% weight-shift re-rank was run; the memo states which positions are stable and which flip.
3. **Gates stayed non-tradeable.** No gate failure was overridden by a high composite score; every gate decision cites evidence.
4. **Shortlist fits capacity.** Shortlist length ≤ the stated capacity number, with effort estimates that sum inside it.
5. **Risk scored, not appended.** Relational, performance, and dependency risk scored per shortlisted candidate and reflected in the ranking, not parked in an annex.
6. **At least one candidate did not make it.** A ranking that greenlights every candidate decided nothing; if all genuinely pass, say so and defend it explicitly.
## KPIs
Judge the prioritization decision itself, six to twelve months out - not the partners' ongoing performance (that belongs to `mbfinotti/partnerships-skills@partner-performance`):
- Share of shortlisted alliances hitting their first-value milestone (first joint deal, integration live, first campaign) by the gate date.
- Partner-sourced and partner-influenced pipeline (B2B) or incremental attributed revenue and brand-lift (B2C) from shortlisted alliances, against the effort spent.
- Gate survival: share of shortlisted candidates that earn "go" at their next gate rather than recycle/kill.
- Ranking churn at the next portfolio review - a reversed top-3 signals weak criteria, not a changed world, unless a named trigger fired.
- Regret log: passed-over candidates that partnered with a competitor and visibly performed. Painful, and the only honest test of the cut line.
## B2B and B2C
Identical in both:
- The stage-gate funnel.
- Must-pass gates kept outside the tradeable score.
- Justified weights plus sensitivity analysis.
- The capacity cut line.
- The memo shape.
Say so when asked. Do not reinvent the method per side.
What genuinely changes - design for it:
| Dimension | B2B alliance/tech | B2C brand/creator |
| ------------------ | ---------------------------------------------------------- | --------------------------------------------------------------- |
| Ranking currency | Pipeline, ACV, integration adoption | Brand equity, audience, incremental attributed revenue |
| Core fit construct | Resource complementarity + org compatibility | Brand fit / attitude congruence (Simonin & Ruth) |
| Dominant risk | Channel conflict, platform envelopment, roadmap dependency | Reputational contagion, negative spillover, creator fraud |
| Key data | Account overlap, closed-won history, co-sell records | Audience overlap, brand-lift/incrementality tests, fraud scores |
| Signature gate | Security/compliance, viability | Brand-safety and fraud vetting as pass/fail |
Full sourced contrast, hybrid cases, and the evidence-hygiene notes in [references/b2b-vs-b2c-contrast.md](references/b2b-vs-b2c-contrast.md).
## Failure Modes
| Failure | Fix |
| -------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------ |
| Logo prestige outranks fit - the famous candidate wins on halo | Weight complementarity + compatibility highest; score the protected candidate visibly, like the rest |
| Equal weights and false precision | Justify every weight in writing; run the sensitivity re-rank |
| Account overlap used as the ranking spine | Cap it as one input; the vendor's own network data shows far smaller effects than its survey headlines |
| "Strategic" candidate exempted from measurement | Every candidate gets a first-value milestone and a gate date, however strategic |
| Ranking produced, no capacity line | Fix capacity in the Interview; shortlist length ≤ capacity, always |
| Gate failure traded away against a high score | Gates live outside the score; a SOC 2 failure is not worth any number of scorecard points |
| A 0.1 composite gap treated as decisive | Scores within ~5% are a tie; break ties with documented judgment, not decimal theater |
| Ranking never revisited | Portfolio review cadence plus named re-ranking triggers, dated in the memo |
## Invocation Examples
- "We have nine candidate technology alliances and capacity for three. Rank them and tell me which to drop."
- "Leadership wants us to sign a hyperscaler alliance. Score it against the four smaller candidates we already had on the list."
- "Re-run last year's alliance shortlist - two candidates shipped the integrations they promised and one had a down round."
## Reference
- [references/scoring-criteria-and-weights.md](references/scoring-criteria-and-weights.md) - criteria menu, weighting mechanics, the scorecard critique and its fixes, effort scoring, tie-bands.
- [references/risk-and-dependency-screens.md](references/risk-and-dependency-screens.md) - must-pass gate list, Das & Teng risk taxonomy, Adner's two ecosystem risks, platform envelopment, B2C contagion and fraud screens.
- [references/gates-and-governance.md](references/gates-and-governance.md) - stage-gate funnel, verdict vocabulary, joint value proposition workshop format, steering-committee shape, portfolio review cadence.
- [references/worked-examples.md](references/worked-examples.md) - a fully worked ranked shortlist, and a negative example of a logo-collecting scorecard annotated line by line.
- [references/b2b-vs-b2c-contrast.md](references/b2b-vs-b2c-contrast.md) - the sourced B2B/B2C contrast, hybrid cases, and vendor-statistic hygiene.
Sibling skills (same collection):
- `mbfinotti/partnerships-skills@partner-ecosystem` - ranks ecosystem segments and categories; hands its named candidates to this skill.
- `mbfinotti/partnerships-skills@joint-gtm-planning` - planning the joint motion after a candidate earns "go".
- `mbfinotti/partnerships-skills@partner-tiering` - tier criteria inside an existing program; not a candidate-ranking tool.
- `mbfinotti/partnerships-skills@partner-marketplace-strategy` - third-party marketplace listing strategy.