influencer-discovery-brief

{
  "skill_name": "influencer-discovery-brief",
  "evals": [
    {
      "id": 1,
      "prompt": "We're launching a barrier-repair serum at Silt Botanicals in March and my intern pulled six creators off Instagram: @dewpointdaily (412K), @barrierbabe (38K), @skinsciencesam (91K), @glowupgrace (1.3M), @quietroutine (12K), @acidsandanswers (156K). Rank them best to worst for us and tell me which three to sign.",
      "expected_output": "The run refuses to rank yet. It runs the interview, states that weights must be approved before any candidate is scored, presents the weighting approaches as a ranked menu with the two-stage audience-fit gate recommended, and asks for evidence (exports or profile review) rather than scoring these handles from general knowledge.",
      "files": [],
      "expectations": [
        "Does not produce a ranked order of the six creators in this reply",
        "Asks interview questions before any scoring, covering at least the campaign objective, the target customer, the converting markets, and the brand-safety redlines",
        "Asks interview questions one at a time rather than delivering the whole list as a single questionnaire block",
        "States explicitly that criteria weights must be approved by the manager before the first candidate is scored",
        "Presents the weighting approaches as a ranked menu and leads with a recommendation rather than listing them flat",
        "Names all three weighting approaches: two-stage audience-fit gate, objective-led fixed weights, and floor-gate plus equal weights",
        "Recommends the two-stage audience-fit gate as the default approach",
        "Declines to score audience fit from general knowledge of these creators, and asks for an audience-insights or discovery-platform export or a direct profile review instead",
        "States that knockout screening runs on every candidate before any scoring begins",
        "States that follower count alone is not evidence of audience fit and refuses to rank on it",
        "Names the brief's sections: header, criteria, knockouts, shortlist, and rejected list",
        "Requires a cited evidence source attached to every scored cell"
      ]
    },
    {
      "id": 2,
      "prompt": "Two finalists for our cold-brew campaign at Harrow & Pine. @mugshotmornings on Instagram has 84,000 followers and her last ten posts averaged 2,100 likes and 160 comments. @beanbuzz on TikTok has 210,000 followers and his last ten averaged 9,800 likes, 420 comments, 1,900 shares and 3,100 saves. Which one has the better engagement, and is either of them above average?",
      "expected_output": "Each rate is computed with its own platform's formula - Instagram roughly 2.7%, TikTok roughly 7.2% including shares and saves - the two are declared non-comparable across platforms, the cited platform-wide averages carry their reliability flags, and the real comparison is against tier-and-niche peers in the same pool.",
      "files": [],
      "expectations": [
        "Computes the Instagram engagement rate as (likes + comments) / followers x 100",
        "Computes the TikTok engagement rate including shares and saves: (likes + comments + shares + saves) / followers x 100",
        "Reports the Instagram engagement rate as approximately 2.7%",
        "Reports the TikTok engagement rate as approximately 7.2%",
        "Does not apply the Instagram formula to the TikTok candidate, and does not report a TikTok rate near 4.9%",
        "States that engagement rates from two different platforms cannot be compared directly against each other",
        "Cites the platform-wide average engagement rates of approximately 0.48% for Instagram and approximately 3.73% for TikTok",
        "Flags those platform-wide averages as vendor self-reported and not broken out by follower tier",
        "Recommends comparing each candidate against tier-and-niche peers inside the same candidate pool rather than against a platform-wide average",
        "Declines to name a winner on engagement rate alone and requires comment-quality review plus audience-fit evidence",
        "Recommends comparing engagement on disclosed sponsored posts against each creator's organic baseline"
      ]
    },
    {
      "id": 3,
      "prompt": "Hollowgate Compliance sells SOC 2 automation to heads of security at 100-1000 person companies in Germany and the Netherlands. I've been told to stand up a thought-leader program and my VP wants the shortlist sorted by follower count with a hard 50K minimum on LinkedIn. Set up the criteria for me.",
      "expected_output": "The 50K follower minimum is rejected for a B2B program. Reach is given near-zero weight; comment-section ICP match, professional credibility and newsletter subscriber quality carry the weight, with the practitioner-reasoning caveat stated, geography still verified, and brand safety kept out of the weighted set as a knockout.",
      "files": [],
      "expectations": [
        "Pushes back on the 50K follower minimum as a selection criterion for a B2B program",
        "States that follower reach carries near-zero weight in B2B",
        "Includes comment-section ICP match as a heavily weighted criterion, measured by sampling the job roles of commenters",
        "Includes professional credibility - job title, employer, speaking and publication history - as a heavily weighted criterion",
        "Includes newsletter subscriber quality, judged on subscriber roles and reply culture rather than list size, where the creator runs a newsletter",
        "Gives aesthetic or content fit negligible weight or omits it entirely",
        "Labels the B2B weighting as practitioner reasoning rather than a measured or published standard, and says to carry that caveat into the brief",
        "Still requires audience geography verification against Germany and the Netherlands",
        "Includes niche consistency measured by categorizing the creator's last 20-30 posts",
        "Keeps brand safety out of the weighted criteria and handles it as a knockout instead",
        "Does not present the criteria set as final before the manager explicitly approves the weights",
        "Notes that B2B redlines add employer conflicts, such as a creator employed by a competitor or by a key account"
      ]
    },
    {
      "id": 4,
      "prompt": "My agency's deck says to vet creators with the 3 Rs - Reach, Relevance, Resonance - and a colleague swears there's a 4 Rs version now. Build our vetting scorecard on whichever one is the real standard so I can defend it to my CMO.",
      "expected_output": "Neither is treated as the real standard. The 3 Rs is described as a floating industry mnemonic with no traceable originator, the 4 Rs as one vendor's framework, and the scorecard is built from explicit criteria the manager approves, with weights described as working conventions.",
      "files": [],
      "expectations": [
        "States that the 3 Rs is a recurring industry mnemonic with no traceable canonical originator",
        "Declines to present the 3 Rs as a citable or authoritative method",
        "Identifies the 4 Rs variant, which adds Reference, as one influencer-analytics vendor's framework rather than an industry standard",
        "Does not pick one of the two as the real or correct standard",
        "Builds the scorecard from explicit weighted criteria the manager approves instead of from a named framework",
        "States that a framework name may appear in the brief only as shorthand for criteria the manager already approved",
        "Names concrete criteria from the criteria menu, such as audience fit, engagement quality, niche consistency, geography match, platform fit, and sponsored-content track record",
        "States that the weights are working conventions to negotiate with the manager, not published or measured standards",
        "Keeps brand safety out of the weighted scorecard and treats it as a knockout",
        "Asks for the campaign objective before assigning any weights"
      ]
    },
    {
      "id": 5,
      "prompt": "There's no budget for a discovery tool here, so all I have is what I can see on public profiles. Ambervale Kitchens needs a shortlist of ten creators by Friday and it's Monday morning. How should I score them?",
      "expected_output": "The floor-gate plus equal weights approach is recommended and justified by the public-profile-only data and the sub-week deadline. The two-stage gate is dropped from the menu rather than offered, the floor-gate's ranking is flagged as a pass list rather than an order, and sourcing leans on the roster pull and customer mentions.",
      "files": [],
      "expectations": [
        "Recommends floor-gate plus equal weights as the weighting approach",
        "Justifies that choice by the public-profile-only data and the deadline falling inside a week",
        "Does not offer the two-stage audience-fit gate as a selectable option for this manager",
        "States that the floor-gate produces a pass list rather than an ordering, and says to note that in the brief",
        "Warns that ties near the top are common under equal weights",
        "States that the floor-gate hands the KPI loop nothing to tune",
        "Does not make an audience-insights demographic export a requirement for the scoring pass",
        "Puts the existing roster and CRM pull and the customer-and-community-mention pass first as the sourcing route under this deadline",
        "Does not recommend a discovery-platform export as the sourcing route here",
        "Still requires knockout screening on every candidate before scoring",
        "Uses fallback geography signals such as comment language mix, posting-time patterns or tagged locations in place of a geo export"
      ]
    },
    {
      "id": 6,
      "prompt": "I'm the only person on partnerships at Kilnwright - we sell furnace-monitoring hardware to plant managers at metal foundries. Nobody indexes creators in this space as far as I can tell. I need about 25 names in the pool by the end of the month and I have maybe six hours a week. Where do I actually find them?",
      "expected_output": "The roster pull and customer-and-community-mention pass run together as the core. The discovery-platform export is deleted for lack of coverage and licence, the scraper variant of hashtag search is deleted on platform-terms grounds, social listening is deleted against the solo effort ceiling and named as what the ordering starves, and a privacy review is flagged on the mention pass.",
      "files": [],
      "expectations": [
        "Runs the existing roster and CRM pull together with the customer-and-community-mention pass as the always-run pair",
        "States that the roster and CRM pull returns performance history with this brand but no net-new names",
        "States that the customer-and-community-mention pass is the only pass reaching podcasts, newsletters and closed communities that no platform indexes",
        "Deletes the discovery-platform export pass outright given no licence and no platform coverage of this niche, rather than demoting it",
        "Deletes the scraper variant of hashtag search and names the platform-terms breach and account-ban risk as the reason",
        "Deletes social listening given the solo effort ceiling and the absence of an owner for a standing monitoring job",
        "Treats the sourcing passes as a budget to stop spending once the pool is deep enough, not a checklist to complete",
        "Flags a privacy review before repurposing support, community or survey data into a campaign list",
        "Names competitor collaborations as the pass to add when the pool is still thin after the first passes",
        "States that manual niche keyword and hashtag search earns its place only where no platform covers the niche",
        "Names social listening as what the efficiency ordering starves, and states the condition that would promote it - a repeating program with someone owning the standing job",
        "Does not recommend buying, scraping or otherwise bulk-harvesting a creator database"
      ]
    },
    {
      "id": 7,
      "prompt": "Tallowmoor Distillery, single malt, we only sell in the US and Canada. Scoring is finished. @peatandsmoke scored 4.6 out of 5, our highest, but he posted a rant last month that got him piled on and he has an open deal with another distillery. @caskchaser scored 4.4 but the export says 68% of his audience is in India. @drammatic scored 3.9 and is clean. Who do we sign?",
      "expected_output": "The top scorer is knocked out on brand-safety and competitor-conflict grounds, the second is knocked out on geography rather than scored down, the ordering error (scoring before knockouts) is named, alcohol triggers regulated-category handling and legal review, and every rejection is recorded with its specific failed criterion.",
      "files": [],
      "expectations": [
        "Rejects @peatandsmoke on knockout grounds despite the 4.6 score being the highest",
        "States that a weighted total cannot buy back a brand-safety or exclusivity failure",
        "Treats the open deal with another distillery as a competitor conflict, and asks whether the manager requires category exclusivity rather than ruling on the ambiguous case alone",
        "Rejects @caskchaser as a geography knockout rather than as a low score, noting only about 32% of the audience sits in the converting markets",
        "Names the process error that knockouts were applied after scoring instead of before it",
        "Identifies alcohol as a regulated category and applies age-gating norms, jurisdiction rules and claim restrictions",
        "Routes the final shortlist through legal review before handoff because the category is regulated",
        "Records each rejection as the candidate plus the specific criterion failed, not a bare rejected line",
        "Does not propose lowering a brand-safety weight or averaging the controversy into the weighted score",
        "Recommends re-screening the surviving candidate immediately before handoff if more than a few weeks have passed since screening",
        "Checks audience-facing controversies such as pinned comments, community posts and podcast appearances, not only the creator's own posts",
        "Does not recommend signing on the strength of the highest weighted total"
      ]
    },
    {
      "id": 8,
      "prompt": "This is the third time we've run creator discovery at Norrhalt and I still can't tell my boss whether the process is getting better or worse. What should I be tracking, and what numbers should I be hitting?",
      "expected_output": "Four discovery-process KPIs are named - screen-to-shortlist, shortlist-to-contracted, criteria-weight stability, post-campaign fit rate - campaign-performance metrics are excluded, no benchmark targets are invented, and fit-rate collection is started despite its lag.",
      "files": [],
      "expectations": [
        "Distinguishes discovery-process KPIs from campaign-performance KPIs and keeps campaign metrics out of the answer",
        "Does not propose EMV, impressions, reach, CPM, ROAS or campaign engagement rate as the measure of the discovery process",
        "Names the screen-to-shortlist ratio",
        "Names the shortlist-to-contracted ratio",
        "Names criteria-weight stability between cycles",
        "Names the post-campaign fit rate",
        "States that no reliable published industry benchmarks exist for these ratios and declines to supply target numbers",
        "Tells the user to establish their own baseline and track the trend instead",
        "Explains that a falling screen-to-shortlist ratio means the sourcing passes are getting better targeted",
        "Explains that a persistently low shortlist-to-contracted ratio means the criteria do not match what the manager actually wants, and says to revisit the weights",
        "Recommends starting post-campaign fit-rate collection in this cycle despite its lag, because it is the only signal that corrects a wrong criteria set"
      ]
    },
    {
      "id": 9,
      "prompt": "Shortlist is done for Bricklane Cold Press - five micro creators on Instagram, all in our US and UK markets, all scored against the weights we agreed. Budget covers two. Two of them, @juiceruin and @pressedmornings, run in the same London wellness crowd and tag each other constantly. Pick the two and write me the DMs so I can send them tonight.",
      "expected_output": "No outreach copy is written. An audience-overlap check runs across the shortlist and against the roster, the two overlapping creators are demoted or staggered rather than both contracted, the reason is stated as paying twice for the same audience, and the brief is validated section by section before the pick is final.",
      "files": [],
      "expectations": [
        "Does not draft the DMs or any outreach message copy",
        "States that the outreach message belongs to a separate step or skill and names that handoff",
        "Runs an audience-overlap check across the shortlist and against the existing roster before finalizing the ranking",
        "Demotes or staggers @juiceruin and @pressedmornings rather than contracting both",
        "States the reason for the demotion as paying twice for the same audience",
        "Describes how to sample overlap without tooling - shared frequent commenters, mutual collaboration history, membership in the same micro-community",
        "Does not produce a creative brief, deliverable specs or key messages for the chosen creators",
        "Confirms that the weights were approved and that every scored cell cites an evidence source before finalizing the pick",
        "Presents the brief section by section for the manager to validate rather than delivering it as one finished block",
        "Notes that a nano/micro portfolio carries a disclosure obligation per creator and that smaller creators are likelier to need coaching on it",
        "Suggests memorizing the approved weights, redlines and target-customer definition for later cycles if persistent memory is available"
      ]
    }
  ],
  "trigger_queries": [
    { "query": "Build the discovery brief for our spring skincare launch", "should_trigger": true },
    { "query": "How do I decide which influencers to approach?", "should_trigger": true },
    { "query": "We need vetting criteria for creators before we contact anyone", "should_trigger": true },
    { "query": "Help me shortlist creators for a B2C campaign", "should_trigger": true },
    { "query": "What weights should I use to score candidate influencers?", "should_trigger": true },
    { "query": "I have 40 handles in a spreadsheet and no idea which ones are worth contacting", "should_trigger": true },
    { "query": "Set up the knockout screens for our creator sourcing", "should_trigger": true },
    { "query": "How do I check whether an influencer's audience is actually our customer?", "should_trigger": true },
    { "query": "My boss wants the biggest accounts and I think that's wrong - help me build the case", "should_trigger": true },
    { "query": "Define sourcing criteria for a B2B thought-leader program", "should_trigger": true },
    { "query": "Where do I even find creators in a niche nobody covers?", "should_trigger": true },
    { "query": "Score these candidates against our brand-safety and engagement criteria with evidence per score", "should_trigger": true },
    { "query": "Is a 40K follower account better for us than a 900K one?", "should_trigger": true },
    { "query": "We keep signing people whose followers aren't in our market - how do we stop that?", "should_trigger": true },
    { "query": "What should be in the brief my team uses to pick creators?", "should_trigger": true },
    { "query": "how do i vet an influencer before reaching out", "should_trigger": true },
    { "query": "Need a ranked shortlist of creators by Friday", "should_trigger": true },
    { "query": "Which of these YouTubers actually reaches finance buyers?", "should_trigger": true },
    { "query": "Set up a repeatable process for choosing creator partners", "should_trigger": true },
    { "query": "Our last three creator picks flopped - I think our selection is broken", "should_trigger": true },
    { "query": "How do I tell a real engaged audience from a bought one?", "should_trigger": true },
    { "query": "What criteria separate a good creator partner from a bad one for us?", "should_trigger": true },
    { "query": "Draft the criteria table and pass floors for our campaign", "should_trigger": true },
    { "query": "We want five nano creators instead of one big one - how do I choose them?", "should_trigger": true },
    { "query": "Help me pick who to send free product to", "should_trigger": true },
    { "query": "I need to justify to my CMO why we skipped the 2M follower account", "should_trigger": true },
    { "query": "Audit this creator list and tell me who fails our redlines", "should_trigger": true },
    { "query": "How many candidates should I screen to end up with ten good ones?", "should_trigger": true },
    { "query": "What data do I need from a creator before I can score them?", "should_trigger": true },
    { "query": "Our program repeats every quarter - how should we set up creator selection so it improves?", "should_trigger": true },
    { "query": "Compare these two candidates on audience fit", "should_trigger": true },
    { "query": "Who on this list is genuinely on-niche and who just went viral once?", "should_trigger": true },
    { "query": "Should newsletter writers count as creators for our program?", "should_trigger": true },
    { "query": "Build the scorecard we'll use to choose ambassadors", "should_trigger": true },
    { "query": "How do I avoid paying two creators for the same followers?", "should_trigger": true },
    { "query": "the agency sent us 20 names, now what", "should_trigger": true },
    { "query": "We only sell in Canada - how do I filter creators by where their audience actually is?", "should_trigger": true },
    { "query": "What's the process before we start pitching creators?", "should_trigger": true },
    { "query": "Set the minimum bar a creator has to clear to make our list", "should_trigger": true },
    { "query": "I want a defensible ranking, not a gut pick", "should_trigger": true },
    { "query": "Which podcasters should we partner with for our developer tool?", "should_trigger": true },
    { "query": "Help me plan how to source candidate partners for an ambassador program", "should_trigger": true },
    { "query": "Our exclusivity rules keep biting us after we've already reached out", "should_trigger": true },
    { "query": "Tell me what to check on a creator's profile before we commit", "should_trigger": true },
    { "query": "Rank these ten TikTokers for our launch and show your working", "should_trigger": true },
    { "query": "How should follower tiers factor into who we pick?", "should_trigger": true },
    { "query": "We got burned by a creator's old posts - how do we screen for that next time?", "should_trigger": true },
    { "query": "Write the DM I'll send to the creator we picked", "should_trigger": false },
    { "query": "Draft a first-touch pitch to an influencer we've already vetted", "should_trigger": false },
    { "query": "What should the creative brief say about deliverable specs and hashtags?", "should_trigger": false },
    { "query": "The creator signed - now write her brief with key messages and brand guardrails", "should_trigger": false },
    { "query": "She's asking for $8K plus perpetual usage rights - what's my counter?", "should_trigger": false },
    { "query": "Build my concession ladder for creator rate negotiation", "should_trigger": false },
    { "query": "What's a fair kill fee in a creator contract?", "should_trigger": false },
    { "query": "Set up KPIs and attribution for the campaign once it launches", "should_trigger": false },
    { "query": "How do I attribute conversions to each creator's posts?", "should_trigger": false },
    { "query": "Design the reporting cadence for our influencer campaign results", "should_trigger": false },
    { "query": "Recruit 200 affiliates into our commission program", "should_trigger": false },
    { "query": "Write the recruitment email sequence for coupon and cashback partners", "should_trigger": false },
    { "query": "What disclosure language do our creators need in their captions?", "should_trigger": false },
    { "query": "Are our affiliates compliant with disclosure rules?", "should_trigger": false },
    { "query": "Set our affiliate commission rates by product line", "should_trigger": false },
    { "query": "Should we pay recurring or one-time affiliate commissions?", "should_trigger": false },
    { "query": "Our affiliate payouts don't reconcile with the platform report", "should_trigger": false },
    { "query": "Detect click fraud in our affiliate program", "should_trigger": false },
    { "query": "Build the dashboard for ongoing affiliate performance", "should_trigger": false },
    { "query": "Draft the terms and conditions for our affiliate program", "should_trigger": false },
    { "query": "Design the onboarding emails new affiliates get in week one", "should_trigger": false },
    { "query": "Which of our 60 technology partners deserve investment next year?", "should_trigger": false },
    { "query": "Build our partner tier structure with the benefits at each tier", "should_trigger": false },
    { "query": "Plan a co-selling motion with our biggest ISV partner", "should_trigger": false },
    { "query": "Our reseller and our direct team are fighting over the same account", "should_trigger": false },
    { "query": "What should we put in our partner enablement portal?", "should_trigger": false },
    { "query": "Model the economics of our channel program", "should_trigger": false },
    { "query": "Design the referral incentive for existing customers", "should_trigger": false },
    { "query": "Stop people from gaming our referral program", "should_trigger": false },
    { "query": "Should we list our app on a major cloud marketplace?", "should_trigger": false },
    { "query": "Write a joint go-to-market plan with our new partner", "should_trigger": false },
    { "query": "What does a partnerships manager need on their resume?", "should_trigger": false },
    { "query": "Design the interview loop for a partner manager hire", "should_trigger": false },
    { "query": "Map our partner ecosystem and show me where the gaps are", "should_trigger": false },
    { "query": "How is our partner performance trending this quarter?", "should_trigger": false },
    { "query": "Write an SEO content brief for our pillar page", "should_trigger": false },
    { "query": "Prepare my discovery call questions for tomorrow's prospect", "should_trigger": false },
    { "query": "Build a design brief for our rebrand", "should_trigger": false },
    { "query": "Do a competitive brief on our top three rivals", "should_trigger": false },
    { "query": "Set up continuous product discovery interviews with our users", "should_trigger": false },
    { "query": "Find me the keywords our competitors rank for", "should_trigger": false },
    { "query": "Source candidates for our open backend engineer role", "should_trigger": false },
    { "query": "Which conferences should we sponsor next year?", "should_trigger": false },
    { "query": "Pick the best subreddit to post our launch in", "should_trigger": false },
    { "query": "Write the captions for our own brand Instagram account", "should_trigger": false },
    { "query": "Which journalists should I pitch our funding announcement to?", "should_trigger": false },
    { "query": "Help me choose which customers to feature in a case study", "should_trigger": false }
  ]
}

# Audience Fit and Engagement Signals

How to actually score audience fit, engagement quality, geography, and overlap - with the reliability flag on every number cited.

## Audience fit over creator fit

- Judge the audience, not the persona. "Being popular isn't the same as being influential." - Scott Guthrie (sabguthrie.info; the post dates from 2018-19 but remains hosted on his still-active site).
- Sample the comment section: who comments, and do they resemble the target customer (B2C: demographics, interests; B2B: job roles, employers)?
- Treat media-kit demographics as the creator's claim, not evidence - verify against an audience-insights export or comment sampling before scoring on it.
- If you can browse the web, read 10-20 recent posts and their comments yourself; otherwise ask the manager to paste an audience-insights or discovery-platform export. Never score fit from general knowledge of the creator.

## Engagement-rate formulas - one per platform

Formulas differ per platform. Applying one platform's formula to another silently misranks candidates.

The formulas below follow one vendor's published methodology (Socialinsider, a self-reported analysis of 70M posts, Jan 2024-Dec 2025):

- Instagram ER = (likes + comments) / followers x 100
- TikTok ER = (likes + comments + shares + saves) / followers x 100
- Facebook and X use their own reaction/repost variants - look up the platform's convention before computing.

Platform-wide averages from that same vendor study:

- TikTok: 3.73%
- Instagram: 0.48%
- Facebook: 0.15%
- X: 0.12%

Reliability flags:

- Vendor self-reported.
- Platform-wide, not broken out by follower tier.
- The vendor itself notes the figures are 2025 values presented as 2026.

Use them as sanity anchors only. The honest comparison is a candidate against tier-and-niche peers inside the same candidate pool.

Tier-level engagement claims circulate (e.g. a 2025 Qoruz study seen only as a search snippet, methodology unconfirmed) - treat as low-confidence. Do not cite them as fact in a brief.

## Engagement quality beyond the rate

- Read the comments, not just the count. "Reach and impression numbers are vanity metrics." - Scott Guthrie (same source and date flag as above).
- Substantive comments reference the content; a wall of generic praise or emoji strings signals a hollow or gamed audience.
- Check whether the creator replies - a reply culture indicates a real relationship with the audience.
- Compare engagement on disclosed sponsored posts against the creator's organic baseline: a collapse on #ad posts means the audience tunes out promotion.

## Shallow authenticity red flags - screening level only

These are the checks a shortlister runs; a full fake-follower audit is a separate discipline and out of this skill's scope.

- Generic or emoji-only comments that could attach to any post.
- Engagement wildly out of line with follower count, in either direction.
- Sudden follower spikes without a matching viral post or press moment.
- Commenter profiles that don't match the claimed audience demographics.
- Reciprocal engagement rings: the same accounts commenting on each other's posts within minutes (pod behavior).

When red flags stack, park the candidate as "authenticity unverified" in the rejected list rather than scoring around the doubt.

## Geography verification

Not ranked, deliberately: the export is strictly better on every axis and the fallback exists only when it is unavailable, so an efficiency ordering here would be false precision dressed as a choice.

- Primary: audience-insights geo split from the creator or a discovery platform.
- Fallback signals when no export exists: comment language mix, posting-time patterns, tagged locations, where engaged commenters say they are.
- Score the share of audience inside converting markets; a huge audience concentrated outside them is unusable reach, whatever the total.

## Audience overlap across the roster

- Check overlap between each shortlist candidate, the other shortlisted creators, and the existing roster.
- Not ranked, for the same reason as geography above: tool availability decides this, not a value-per-effort trade-off.
- With tooling: use follower-overlap exports where the manager's platform provides them.
- Without tooling: sample shared frequent commenters, mutual collaboration history, and membership in the same micro-community - heavy overlap shows up fast.
- Demote or stagger heavily overlapping candidates: contracting both pays twice for the same eyeballs.

# Brand Safety Screening

The knockout checklist a shortlister applies to every candidate before any scoring. A knockout failure sends the candidate straight to the rejected list with its specific reason - no weighted total can buy it back.

## Knockout checklist

- Scan recent content - working convention: the last ~3 months, longer for a high-stakes campaign - for controversy, hate or harassment, and anything crossing the manager's stated redlines.
- Reject on an active competitor partnership when the manager requires category exclusivity; log ambiguous conflicts (expired deals, adjacent categories) for the manager to rule on rather than deciding alone.
- Reject a pattern of undisclosed sponsorships - a creator who hides material connections today creates brand liability tomorrow, and platforms and regulators hold the brand responsible too.
- Reject a history of fabricated claims: fake results, implied unpaid-customer framing, unverifiable testimonials.
- Apply regulated-category constraints when the interview flagged one (alcohol, health, finance, gambling, kids' products): age-gating norms, jurisdiction rules, claim restrictions. Route the final shortlist through legal review before handoff in any regulated category.
- Treat a below-floor audience-geography result as a knockout, not a low score - an audience outside converting markets fails the campaign regardless of every other strength.
- Check the creator's audience-facing controversies, not just their own posts: pinned comments, community posts, and podcast appearances carry brand risk equally.

One vendor (HypeAuditor, in a 2024 article updated 2026) describes automated screening of creator content across markers such as toxic, political, or alcohol-related content - that is a vendor tool description, not an industry standard. Manual review of recent content is the baseline that needs no tool; a screening platform accelerates it but does not replace reading the actual posts.

## Screening workflow

1. Screen after sourcing and before scoring - screening after outreach wastes the contact, and public association with a later-rejected creator is itself a brand-safety cost.
2. Record every knockout failure as candidate + the specific criterion failed; a bare "rejected" line is unauditable.
3. Re-screen shortlisted candidates immediately before handoff when more than a few weeks have passed - recent content changes fast.
4. If you can browse the web, review the candidate's recent public content directly; otherwise ask the manager for screenshots or a screening-tool export, and mark any candidate you could not screen as "screening incomplete" rather than passing them silently.

## B2C vs B2B

The screening process is identical for both segments; only the redline content differs.

- B2C redlines skew to conduct, values conflicts, and category taboos.
- B2B redlines add employer conflicts (a creator employed by a competitor or a key account) and professional-claims accuracy.

Say which redline set applies in the brief's knockout section.

# Creator Tiers and Niche Mapping

Follower-tier definitions, what tiers mean per segment, and how to check niche consistency and platform fit.

## The tier taxonomy

Influencer-tooling vendors publishing tier definitions largely agree on this taxonomy:

| Tier  | Followers |
| ----- | --------- |
| Nano  | ~1K-10K   |
| Micro | 10K-100K  |
| Macro | 100K-1M   |
| Mega  | >1M       |

A separate "mid-tier" bracket appears inconsistently across the industry, with varying boundaries. State in the brief which taxonomy its tiers use, so scores stay comparable across cycles.

## What tiers buy (B2C)

Two ways to spend the same budget, ranked:

- **Nano/micro portfolio** - highest engagement and trust per follower; suits conversion and trust objectives, gifting-friendly budgets, and portfolio builds. It costs a contract, a briefing, and a disclosure check per creator, so the coordination hours scale with the count.
- **One macro/mega placement** - reach in a single simultaneous hit; suits awareness objectives and launch moments. One negotiation, one brief, one relationship, but expect lower engagement rates and per-follower conversion, and the whole spend rides on one creator's week.

- effort: nano/micro portfolio (contracting, briefing, and chasing five to ten creators) > one macro/mega placement (one negotiation, though a harder one)
- compliance cost: nano/micro portfolio (a disclosure obligation per creator, and smaller creators are likelier to need coaching on it - one missed disclosure is the brand's problem) > one macro/mega placement (one creator, usually agency-represented and disclosure-literate)
- value: nano/micro portfolio (more converted buyers and more usable content at equal spend - practitioner reasoning widely repeated across sources, with no single measured study confirming it) > one macro/mega placement (simultaneous reach no portfolio can replicate)
- efficiency: nano/micro portfolio > one macro/mega placement, except where the objective needs everyone to see it in the same week

Match the tier mix to the objective and the contracted-creator count from the interview:

- Five contracts skew micro/nano portfolio.
- One contract with an awareness objective skews macro.

Where the brand cannot police a disclosure obligation per creator, cut the portfolio size rather than accepting the exposure.

## B2B: follower tiers are near-worthless

This section is practitioner reasoning. Carry that caveat into any B2B brief.

- Judge who follows and engages, not how many. A small-audience niche operator whose commenters hold buying roles beats a large generalist - reasoning, not a measured claim.
- Credibility signals to score instead of tiers:
  - Job title and employer credibility - does the creator's own role make their opinion load-bearing for the ICP?
  - Comment-section ICP match - are commenters practitioners and buyers, or other creators?
  - Speaking and publication history - conference talks, industry bylines, podcast appearances.
  - Newsletter subscriber quality - subscriber roles and reply culture over raw list size.
- Platform mix in B2B skews to professional networks, newsletters, and industry podcasts; short-form video tiers rarely transfer.

## Niche consistency

- Categorize the creator's last 20-30 posts; compute the share inside the campaign's category.
- Enforce an on-niche floor (working convention: a clear majority of recent content on-niche - set the exact floor with the manager).
- Niche dilution - posting across unrelated categories - dilutes audience intent: followers who came for one topic ignore the others. No verified post-mortem study backs a specific threshold; the floor is a discipline, not a measured cutoff.
- Distinguish niche audiences from virality audiences: a creator whose numbers come from one off-niche viral hit carries an audience that never chose the niche.

## Platform fit

- Verify presence per campaign platform; a creator strong on one platform routinely has a thin, repost-only presence on another.
- Check that the creator natively produces the formats the objective needs (short video, long video, carousel, newsletter, live).
- Score each platform separately; never average a strong platform with a weak one into a single misleading number.

# Examples

One full positive worked example (B2C), the B2B criteria-weighting variant, and one annotated negative example. Every brand, creator, and number below is invented for illustration - never present these figures as real industry data. In a real engagement, every score comes from cited evidence gathered in the workflow.

## Positive: B2C discovery brief with scored shortlist

Scenario (fictional): Loam & Leaf, a DTC houseplant-subscription brand.

- Objective: conversion.
- Target customer: urban renters 25-40 who own fewer than five plants.
- Converting markets: US, UK.
- Platforms: Instagram, TikTok.
- Budget: mid, cash + product.
- Creators to contract: three.
- Redlines: no pesticide-misuse content, no competitor subscription boxes.

```
DISCOVERY BRIEF - Loam & Leaf, 2026-08-25
Header    : Loam & Leaf | conversion | urban renters 25-40, <5 plants | US+UK | IG, TikTok
Criteria  : approved via brainstorming gate (objective-led fixed weights)
```

| Criterion              | Weight | How measured                                     | Pass floor                     |
| ---------------------- | ------ | ------------------------------------------------ | ------------------------------ |
| Audience fit           | 35%    | Audience-insights demo export + comment sampling | >=50% in 25-40 urban segment   |
| Engagement quality     | 25%    | Platform-native ER + manual comment scan         | Above pool median for tier     |
| Niche consistency      | 15%    | Last 30 posts categorized                        | >=60% plant/home content       |
| Geography match        | 15%    | Geo split from export                            | >=55% US+UK audience           |
| Sponsored track record | 10%    | #ad posts vs. organic baseline                   | No collapse on sponsored posts |

Knockouts (applied before scoring):

- Redline scan of last 3 months.
- Active competitor-box partnership.
- Undisclosed-sponsorship pattern.
- Authenticity red flags stacking.
- Geography below floor.

Shortlist (scores 1-5 per criterion; every cell carries an evidence link in the real artifact):

| Candidate                    | Aud. fit | Engage. | Niche | Geo | Sponsored | Weighted | Rank |
| ---------------------------- | -------- | ------- | ----- | --- | --------- | -------- | ---- |
| @rootedinbrooklyn (IG, 24K)  | 5        | 4       | 5     | 5   | 4         | 4.65     | 1    |
| @flatplantclub (TikTok, 61K) | 4        | 5       | 4     | 4   | 4         | 4.25     | 2    |
| @thegreenlandlady (IG, 112K) | 4        | 3       | 5     | 4   | 3         | 3.80     | 3    |

Overlap check: @rootedinbrooklyn and @flatplantclub share <10% sampled frequent commenters; no overlap with the existing roster. No demotion.

Rejected with reason:

- @plantsandpups (89K): niche consistency 40% - below the 60% floor; audience follows for the dogs.
- @urbanjungleguy (240K): knockout - active partnership with a competitor subscription box.
- @leafdropdaily (33K): parked "authenticity unverified" - generic comment wall, engagement out of line with reach.
- @conservatorycait (150K): geography knockout - 71% of audience outside US/UK per export.

Why this brief passes the quality gate:

- Weights were approved before scoring.
- Knockouts ran first.
- Every score cites an export or profile link.
- Nobody was ranked on follower count.
- Geography and overlap are verified.
- Every rejection names its reason.

## B2B variant: criteria weighting for a thought-leader program

Scenario (fictional): Brightledger, an accounting-automation SaaS.

- Objective: trust-building.
- Target customer: heads of finance at 50-500-employee EU companies.
- Platforms: professional network + newsletters.

| Criterion                       | Weight | How measured                                       | Pass floor                            |
| ------------------------------- | ------ | -------------------------------------------------- | ------------------------------------- |
| Comment-section ICP match       | 30%    | Sample commenter roles on last 15 posts            | Majority practitioner/buyer roles     |
| Professional credibility        | 25%    | Title, employer, talks, bylines                    | Verifiable finance-domain credibility |
| Niche consistency               | 20%    | Last 30 posts categorized                          | >=70% finance-ops content             |
| Newsletter subscriber quality   | 15%    | Subscriber roles, reply culture (creator-provided) | Evidence of ICP subscribers           |
| Audience fit (geo/firmographic) | 10%    | Available insights + stated audience               | EU-weighted audience                  |

Follower reach is deliberately absent - near-zero weight in B2B. Note in a real B2B brief: this weighting is practitioner reasoning.

## Negative: the follower-count shortlist

A shortlist built the way this skill exists to prevent, annotated line by line.

> **Shortlist: top 5 biggest plant accounts we could find**

Sourcing by size, not criteria. No approved weights, no knockouts run - scoring started before the manager agreed what "good" means.

> **1. @plantfluencer_official - 2.1M followers. Huge reach, obvious pick.**

Follower-count worship. No audience-fit evidence, no engagement check; a mega account's audience is broad by construction, and the campaign converts on a narrow segment.

> **2. @greenhome_daily - 890K. Beautiful feed, very on-brand aesthetic.**

Over-indexing on aesthetics: production quality scored as if it were audience fit. Nobody sampled who follows or comments.

> **3. @viralgardenhacks - 1.4M. Went viral twice last month.**

Virality audience, not a niche audience - followers came for two spectacle posts, not for plant care. Niche consistency never checked.

> **4. @plantsofworld - 640K, mostly SEA-based audience.**

Geography mismatch, known and ignored. The converting markets are US/UK; most of this reach can never buy.

> **5. @leafqueen_x - 720K. Partnered with Fernbox last quarter.**

Brand-safety screening skipped: a live exclusivity conflict surfaces only after outreach, wasting the contact and risking public association.

> **All five confirmed interested - let's pick the two biggest.**

Final selection by size again:

- No overlap check (accounts 1-3 share large follower bases - paying twice for the same eyeballs).
- No evidence trail.
- No rejected-with-reason list.

Nothing in this shortlist is auditable. Every line is fixed by the positive example's structure: weights first, knockouts second, evidence-backed scores third, rank last.

# Sourcing Criteria and Weighting

The criteria menu, the objective-to-weight translation, the ranked weighting approaches to present at the brainstorming gate, and pass floors. Every weight and floor below is a working convention to negotiate with the manager, not a published or measured standard for shortlist criteria weights.

Do not invoke a named framework as the authority behind weights. The "3 Rs" (Reach, Relevance, Resonance) and the "4 Rs" variant adding "Reference" are floating mnemonics in industry use, not citable methods. Use a framework in a brief only as shorthand for what the manager already approved.

## Criteria menu

| Criterion                      | What it measures                                                      | How measured                                                                                                                                          |
| ------------------------------ | --------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- |
| Audience fit                   | Share of the creator's audience matching the target customer          | Audience-insights demographic export; comment-section sampling; media-kit data (treat self-reported media-kit numbers as claims to verify, not facts) |
| Engagement quality             | Real interaction rate plus comment substance                          | Platform-native engagement formula + manual scan of comment quality                                                                                   |
| Niche consistency              | Share of recent content in the campaign's category                    | Categorize the last 20-30 posts; compute the on-niche share                                                                                           |
| Geography match                | Share of audience in converting markets                               | Audience-insights geo split; fallback signals if unavailable                                                                                          |
| Platform fit                   | Active, native presence on the campaign's platforms                   | Per-platform profile check - never extrapolate strength from one platform to another                                                                  |
| Sponsored-content track record | Whether sponsored posts hold engagement vs. organic posts             | Compare engagement on recent disclosed posts vs. the creator's organic baseline                                                                       |
| Aesthetic/content fit (B2C)    | Visual and tonal match with the brand                                 | Recent-content review against brand guidelines                                                                                                        |
| Professional credibility (B2B) | Job title, employer, speaking/publication history, newsletter quality | Profile, bylines, event listings, newsletter samples                                                                                                  |
| Audience overlap (penalty)     | Duplication with the roster and other shortlisted creators            | Overlap tooling export if available; shared-commenter sampling otherwise                                                                              |

Brand safety is deliberately absent from this table: it is a knockout, never a weighted criterion. A high weighted total must not be able to buy back a safety failure.

## Objective-to-weight translation

The interview's objective answer sets the starting emphasis before the manager tunes it:

| Objective        | Weight toward                                                         | Weight away from                                            |
| ---------------- | --------------------------------------------------------------------- | ----------------------------------------------------------- |
| Awareness        | Reach within fit, geography match, platform fit                       | Deep engagement floors, sponsored track record              |
| Conversion       | Audience fit, engagement quality, sponsored-content track record      | Raw reach, aesthetic polish                                 |
| Content-for-paid | Content/aesthetic fit, production consistency, sponsored track record | Audience size (the content, not the audience, is the asset) |
| Trust-building   | Niche consistency, credibility (B2B), comment-section quality         | Reach, posting volume                                       |

## Weighting approaches for the brainstorming gate - ranked

Present them in this order, lead with the recommendation, and let the manager explicitly pick before any scoring.

1. **Two-stage audience-fit gate.** Gate first on audience fit alone (a minimum fit share), then weight-score only the gate survivors on the remaining criteria.
   - Needs one expensive figure per candidate rather than every figure for every candidate, so the full scoring pass runs on a much smaller set.
   - Enforces fit-first hardest, which is where shortlists actually fail.
   - Cannot compare candidates cut at the gate against each other, so a manager who wants to argue over a near-miss has nothing to argue from.
2. **Objective-led fixed weights.** Assign percentage weights per criterion from the table above, then score every survivor on every criterion.
   - The only approach defending each rank criterion by criterion.
   - The only approach producing weights the post-campaign fit rate can tune, so the only one that compounds across cycles.
   - Demands data for every cell of every survivor, which is the slowest thing this skill asks for.
3. **Floor-gate plus equal weights.** Set a pass floor per criterion, then equal-weight everything that passes. Floors read off public profiles, so it survives on the thinnest data and lands under a one-week deadline.
   - Separates pass from fail, but not the passes from each other; ties are common near the top.
   - Hands the KPI loop nothing to tune.

- effort, counting both scoring hours and the data each cell needs: objective-led fixed weights > two-stage audience-fit gate > floor-gate plus equal weights
- value, as shortlist quality the manager can defend: objective-led fixed weights > two-stage audience-fit gate > floor-gate plus equal weights
- efficiency: two-stage audience-fit gate > objective-led fixed weights > floor-gate plus equal weights, wherever one audience-fit figure per candidate is obtainable; floor-gate leads and the other two are unaffordable where the only data is public profiles

- Default to the two-stage gate.
- Drop to the floor-gate when audience exports are unavailable or the deadline is inside a week. Note in the brief that its ranking is a pass list, not an order.
- Pay the full objective-led pass at least once when the program repeats, since its weights are the asset the next cycle starts from.
- Delete the two-stage gate from the menu, rather than offering it as an option the manager cannot execute, when audience exports do not exist at all.

These orderings are defaults, not laws. Re-rank them against the manager:

- An audience-insights export already in hand moves the two-stage gate further ahead.
- An established weight set from a prior cycle makes the objective-led pass the cheap option rather than the expensive one.
- A first-ever cycle with no exports leaves only the floor-gate.

## Pass floors

- Set a floor per weighted criterion with the manager (e.g. minimum on-niche share, minimum engagement rate for the tier, minimum audience share in converting markets).
- A candidate below any floor is rejected with that floor as the stated reason, whatever the weighted total.
- Floors are working conventions for this brand and cycle - revisit them each cycle against the post-campaign fit rate.

## B2C vs B2B weighting divergence

| Criterion                       | B2C guidance                                           | B2B guidance                                                                                 |
| ------------------------------- | ------------------------------------------------------ | -------------------------------------------------------------------------------------------- |
| Follower reach                  | Meaningful, tier-dependent - keep it a minority weight | Near-zero weight; judge who follows, not how many                                            |
| Audience demographics/geography | Heavy                                                  | Moderate (ICP roles matter more than geo in many B2B motions - still verify geo)             |
| Aesthetic/content fit           | Real weighted criterion                                | Usually negligible                                                                           |
| Professional credibility        | Usually negligible                                     | Heavy: title, employer, speaking/publication history                                         |
| Comment-section ICP match       | Sample for demographic fit                             | Heavy: do commenters hold buying roles?                                                      |
| Newsletter subscriber quality   | Rare                                                   | Heavy when the creator runs a newsletter - subscriber roles and reply culture over list size |

The B2B column is practitioner reasoning. State that caveat in the brief when a B2B manager asks where the weights come from.

Identical for both: the audience-fit-over-creator-fit principle, geography verification, knockout screening, overlap checks, and weights agreed before candidates are viewed.

---
name: influencer-discovery-brief
description: Define creator sourcing criteria and produce the discovery brief used to shortlist candidate influencers and creators - weighted criteria for audience fit, engagement quality, niche consistency and brand safety, knockout screens, and a ranked shortlist scoring template with evidence per score. Covers B2C brand campaigns and B2B creator and thought-leader programs. Use whenever the user mentions finding creators, vetting influencers, audience fit, creator shortlists, or deciding which creators to approach, even if they never say discovery. Runs before any contact. Do NOT use for writing the pitch message itself - use mbfinotti/partnerships-skills@influencer-outreach instead.
license: MIT
metadata:
  author: Maya-Beth Finotti
  version: "1.1.1"
---

# Influencer Discovery Brief

Define the sourcing criteria and produce the discovery brief a partnership manager uses to shortlist candidate creators - who qualifies, by what weighted criteria, and with what evidence.

The "3 Rs" (Reach, Relevance, Resonance) is a genuinely recurring industry convention for judging influence, but research could not trace it to a canonical originator - treat it as a floating mnemonic, not a citable method. A "4 Rs" variant adding "Reference" is one influencer-analytics vendor's framework, not an industry standard. This skill therefore works from explicit weighted criteria the manager approves, never from a named model presented as authoritative.

## Interview

Ask before sourcing anything. One question per message; offer multiple-choice answers when possible. Skip any question the user already answered.

- Is this a B2C brand campaign or a B2B creator/thought-leader program? (Or both - then weight criteria separately per track.)
- What is the brand/product, and what is the value prop a creator's _audience_ would actually care about?
- What is the campaign objective: awareness, conversion, content-for-paid, or trust-building? The objective changes the criteria weights, so answer before any scoring.
- Who is the target _customer_ - demographics, role, buying context? (Not the target creator; the creator is a channel to this person.)
- Which platform(s) does the campaign run on?
- Which geographies/markets actually convert for this product? (Audience anywhere else is reach you pay for and can't use.)
- What is the budget tier, and is compensation cash, product, or a mix?
- How many creators will actually be contracted from the shortlist?
- What are the brand-safety redlines, and is the product in a regulated category (alcohol, health, finance, gambling, kids)?
- Any competitor or category-exclusivity conflicts to enforce? Who is already on the roster? (Needed for audience-overlap checks.)
- What audience data can you actually obtain: creator discovery platform exports, audience-insights exports, media kits, or public profiles only? (A licensed discovery platform moves its export pass from the slowest sourcing option to the fastest. Public profiles only deletes the two-stage weighting gate from the menu.)
- By what date must the shortlist land, and what is waiting on it? (A date inside a week promotes the roster pull, the customer-mention pass, and the floor-gate weighting. A quarter makes social listening and full objective-led scoring affordable.)
- One-off campaign or a creator program you will run again? (A one-off stops at a shortlist. A repeating program justifies the full objective-led scoring pass, whose approved weights are the only output the post-campaign fit-rate KPI can tune.)
- What is the effort ceiling - your own hours, whose data you can request, and whether you can commit to a standing monitoring job? (A solo ceiling deletes social listening. No budget owner for a platform licence deletes discovery-platform exports.)

Carry these last three answers into every ranking below. They are the only inputs that reorder the defaults for this manager.

Every ordering in this skill is a default, not a law. It ranks by qualified candidates returned per unit of effort, which is never the same ordering as cheapest-first, and it shifts with context and with who runs the pass.

Re-rank against what you already know about this manager before recommending anything. Say which answer moved which pass, so the reordering is auditable rather than a preference you asserted.

- A discovery platform already licensed moves its export pass to the front.
- An existing creator roster with performance history makes the roster pull the strongest pass rather than merely the cheapest.
- A niche too small for platform databases deletes the export pass outright and promotes customer mentions and social listening in its place.

## Workflow

1. Run the Interview; collect every answer before looking at a single candidate.
2. Translate the objective into a draft weighted-criteria set - awareness, conversion, content-for-paid, and trust-building each weight differently. See [references/sourcing-criteria-and-weighting.md](references/sourcing-criteria-and-weighting.md).
3. Enter brainstorming mode: present the weighting approaches in ranked order, lead with the recommendation, and give each one's trade-off in the manager's own vocabulary (criteria, floors, knockouts, shortlist). Full trade-offs in [references/sourcing-criteria-and-weighting.md](references/sourcing-criteria-and-weighting.md).

   - efficiency: two-stage audience-fit gate > objective-led fixed weights > floor-gate plus equal weights, wherever one audience-fit figure can be obtained per candidate; floor-gate leads and the other two are unaffordable where the only data is public profiles.

   Default to the two-stage gate. Drop to the floor-gate when the data answer was public profiles only, or the date answer leaves under a week. Pay the full objective-led scoring pass at least once when the answer was a repeating program - it is the only approach that produces tunable weights.

4. Ask brainstorming questions one at a time; score nothing until the manager explicitly picks a weighting approach.
5. Build the raw candidate pool through sourcing passes. Six passes exist; treat them as a budget, not a checklist. Run them in this order and stop once the pool is deep enough to survive the knockouts.

   - **Existing roster and CRM** - an hour, no tool access. The only pass returning performance history with _your_ brand: who already posted, what it did, who the manager would sign again. It returns no net-new names, and step 9's overlap check needs it regardless.
   - **Customer and community mentions** - a day pulling support and community threads, review text, and a "which creators do you follow" question into a survey already going out. Names arrive with fit evidence attached, because your buyers demonstrably watch them, and this is the only pass reaching podcasts, newsletters, and closed communities no platform indexes.
   - **Discovery-platform exports** - near-zero with a licence in hand: one filtered query returns breadth plus the audience-demographic export step 8 scores against. Without one, a quarter and a procurement conversation. Covers only what the vendor indexed - thin in narrow B2B niches and on non-social formats.
   - **Competitor collaborations** - an hour browsing competitors' tagged and disclosed posts. Pre-qualified: they take sponsorships, they are in category, and their sponsored-versus-organic engagement is visible before you make contact. Every competitor works the same list, and the exclusivity knockout removes part of it.
   - **Niche keyword and hashtag search** - an hour that keeps being an hour, no tool access. It mostly re-derives by hand what the export returns, at a low qualified-per-hour rate; it earns its place only where no platform covers the niche.
   - **Social listening** - a week to design queries and a standing job to maintain. It surfaces creators already talking about the category unprompted, the strongest unique-name yield of any pass, but it is a monitoring system whose discovery output is a by-product.

   - effort: social listening > discovery-platform exports without a licence > customer and community mentions > competitor collaborations == hashtag search (tied: both an hour of manual browsing, no tool access, no one else's permission) > roster and CRM == discovery-platform exports with a licence (tied: both an hour spent querying data you already hold)
   - compliance cost: customer and community mentions (a privacy review before repurposing support, community, or survey data into a list you build a campaign from) > hashtag search run by scraper (a platform-ToS breach that gets the account banned - near-zero read by hand) > discovery-platform exports (the vendor licence governs what you may re-export and retain) > roster and CRM (past contracts can carry exclusivity or re-approach clauses) == competitor collaborations == social listening (no exposure)
   - value, as qualified creators the other passes would have missed: customer and community mentions > social listening > discovery-platform exports > competitor collaborations > roster and CRM > hashtag search
   - efficiency: roster and CRM > customer and community mentions > discovery-platform exports > competitor collaborations > social listening > hashtag search

   **Always run:**

   - The roster pull and the customer-mention pass together - they cost a day and return the two kinds of name nothing else does.

   **Add next, in this order:**

   - The discovery-platform exports third; move it to first whenever the licence already exists - the same query then costs an hour and also supplies the scoring data.
   - Competitor collaborations, when the pool is still thin after three passes.
   - Social listening, the moment the answers said a repeating program and someone will own the standing job.

   **Delete rather than demote:**

   - The scraper variant of hashtag search: it breaks platform terms and the banned account takes the manual pass down with it.
   - The export pass entirely, when the niche has no platform coverage and no licence exists.

   Social listening is what this order starves - second on value, first on effort, so it loses every efficiency round and a manager following the ratio alone never builds the one pass that surfaces creators nobody else has found.

6. If you can browse the web, verify each candidate's public profile yourself; otherwise ask the manager to paste discovery-platform or audience-insights exports. Never score from memory or assumption.
7. Screen every candidate against the knockout checklist _before_ any scoring; a knockout failure goes straight to the rejected list with its reason. See [references/brand-safety-screening.md](references/brand-safety-screening.md).
8. Score survivors per criterion with the approved weights, attaching one evidence source per score. Signals in [references/audience-fit-and-engagement-signals.md](references/audience-fit-and-engagement-signals.md); tier and niche mapping in [references/creator-tiers-and-niche-mapping.md](references/creator-tiers-and-niche-mapping.md).
9. Rank by weighted total; check audience overlap across the shortlist and against the existing roster; demote candidates duplicating an audience already paid for.
10. Self-check the brief against the Quality gate below; iterate until every item passes.
11. Present the brief section by section - header, criteria, knockouts, shortlist, rejected list - validating each with the manager before finalizing.
12. If your harness has persistent memory, memorize the approved criteria weights, redlines, and target-audience definition - later runs skip straight past the interview.

Stop at the ranked shortlist. Outreach messages and the creative brief handed to a chosen creator belong to sibling skills, not this one.

## B2C vs B2B - where criteria diverge

- B2C: weight audience demographics, geography, and aesthetic/content fit; follower tiers carry real meaning (nano/micro for engagement and conversion, macro/mega for reach).
- B2B: weight the creator's job title and employer credibility, who is in the comments (do commenters match the ICP?), speaking and publication history, and newsletter subscriber quality. Treat raw follower counts as near-worthless. The guidance above is practitioner reasoning: state that caveat in the brief when a B2B manager asks where the weights come from.
- Identical for both: audience-fit-over-creator-fit, geography verification, brand-safety screening, audience-overlap checks across the roster, and the discipline of agreeing weights _before_ looking at candidates.

## Expected output

Deliver one structured artifact (full worked example in [references/examples.md](references/examples.md)):

```
DISCOVERY BRIEF - <brand>, <date>
Header    : brand, objective, target customer, converting markets, platforms
Criteria  : table - criterion / weight / how measured / pass floor
Knockouts : screening checklist applied before scoring, with redlines
Shortlist : table - candidate / score per criterion / weighted total / rank
            / evidence link per score
Rejected  : candidate + one-line reason (knockout failed or floor missed)
```

Every score cites its evidence; every rejection states its reason. A brief the manager can't audit line by line is not done.

## Quality gate

Check the finished brief against every item below, verified for every shortlisted candidate. Pass threshold: 7/7. Iterate - rescore, re-source, re-screen - until nothing fails.

These thresholds are working conventions for auditable shortlists, not measured industry standards.

1. Weights were explicitly approved by the manager before the first candidate was scored.
2. Every knockout criterion was applied to every candidate before scoring began.
3. Every scored cell carries a cited evidence source (profile link, export, or pasted data) - no memory-scored cells.
4. No candidate was scored on follower count alone; every shortlisted candidate has audience-fit and engagement evidence.
5. Audience geography is verified for every shortlisted candidate against the converting markets.
6. Audience overlap is checked across the shortlist and against the existing roster.
7. Every rejected candidate has a stated, specific reason.

## KPIs - shortlist quality, not campaign performance

Measure the discovery process itself. Campaign KPIs belong to the measurement skill.

No reliable published industry benchmarks exist for these ratios. Don't invent targets: establish your own baseline and track the trend.

- Screen-to-shortlist ratio: candidates screened per candidate shortlisted. Falling over time means sourcing passes are getting better targeted.
- Shortlist-to-contracted ratio: shortlisted candidates the manager actually signs. Persistently low means criteria don't match what the manager really wants - revisit weights.
- Criteria-weight stability: how much approved weights change between discovery cycles. Stabilizing weights mean the criteria are maturing; churning weights mean the objective is unclear.
- Post-campaign fit rate: share of shortlisted-then-contracted creators later judged a good fit. Feed it back into the weights.

- effort: post-campaign fit rate (a quarter of waiting plus a judgment collected per contracted creator) > screen-to-shortlist == shortlist-to-contracted == criteria-weight stability (near-zero - all three are counts the brief already produced)
- value: post-campaign fit rate (the only one saying whether the criteria themselves were right) > shortlist-to-contracted > screen-to-shortlist == criteria-weight stability (tied because neither judges the shortlist, only the pass that produced it, and neither reads at all until a second cycle exists)
- efficiency: shortlist-to-contracted > screen-to-shortlist == criteria-weight stability > post-campaign fit rate

Track the three free ratios from cycle one. Start collecting the fit rate in cycle one anyway - its lag is precisely why it never gets started, and it is the only signal that corrects a wrong criteria set.

## Common failure modes

| Failure                                                  | Fix                                                                                                       |
| -------------------------------------------------------- | --------------------------------------------------------------------------------------------------------- |
| Follower-count worship                                   | Weight audience fit and engagement first; cap or zero reach weight per the approved criteria              |
| Matching the creator's persona instead of their audience | Score who follows and comments, not who posts; sample the comment section against the target customer     |
| Audience-geography mismatch                              | Verify geo split for every shortlisted candidate; audience outside converting markets is unusable reach   |
| Engagement-rate gaming and pods                          | Apply the shallow red-flag checks; park unverifiable candidates as "authenticity unverified"              |
| Niche dilution (unrelated content categories)            | Categorize recent posts; enforce the on-niche share floor                                                 |
| Over-indexing on aesthetics/production quality           | Keep aesthetic fit a minor weighted criterion (B2C only), never a proxy for audience fit                  |
| Ignoring audience overlap across the roster              | Run the overlap check before ranking; demote duplicated audiences - never pay twice for the same eyeballs |
| Brand-safety screening after outreach                    | Knockouts run before scoring, which runs before any contact                                               |
| Scoring before weights are agreed                        | The brainstorming gate blocks scoring until the manager explicitly picks the weighting approach           |
| One platform's engagement formula applied to another     | Use each platform's own formula; never compare cross-platform rates directly                              |

## Invocation Examples

- "Build the discovery brief for a B2C skincare launch - we need 8 micro-creators, budget $40K, conversion objective."
- "We need B2B thought-leaders for a developer-tooling campaign. Define the vetting criteria and the knockout screens."
- "Score these 14 candidate creators against our brand-safety and engagement-quality criteria, with evidence per score."

## Reference

- See [references/sourcing-criteria-and-weighting.md](references/sourcing-criteria-and-weighting.md) when translating the objective into criteria - the criteria menu, objective-to-weight mapping, the ranked weighting approaches for the brainstorming gate, and pass floors.
- See [references/audience-fit-and-engagement-signals.md](references/audience-fit-and-engagement-signals.md) when scoring candidates - audience-fit tests, per-platform engagement formulas with benchmark reliability flags, shallow authenticity red flags, geography and overlap checks.
- See [references/creator-tiers-and-niche-mapping.md](references/creator-tiers-and-niche-mapping.md) when mapping tiers and niches - the verified tier taxonomy with its provenance, B2C tier use, B2B credibility signals, niche-consistency checks.
- See [references/brand-safety-screening.md](references/brand-safety-screening.md) when screening - the knockout checklist, regulated-category constraints, competitor-conflict handling.
- See [references/examples.md](references/examples.md) before writing the brief - a full worked B2C brief with scored shortlist, a B2B criteria-weighting variant, and an annotated negative example.
- See `mbfinotti/partnerships-skills@influencer-campaign-brief` for the creative brief handed to a creator after selection.
- See `mbfinotti/partnerships-skills@influencer-measurement-framework` for campaign KPIs; this skill measures only shortlist quality.
- See `mbfinotti/partnerships-skills@affiliate-disclosure-compliance` for the disclosure rules any contracted creator must follow.