integration-listing-optimization

{
  "skill_name": "integration-listing-optimization",
  "evals": [
    {
      "id": 1,
      "prompt": "We run Cadenceloop, a standup-and-retro bot that's been live on the Slack Marketplace for about five months. Q4 plan from my head of growth: A/B test two versions of the listing description over six weeks, pick the winner on view-to-install conversion, then run a lifecycle email campaign to our 900 active workspaces to land 40 reviews by December so the star rating finally shows up. She wants a weekly dashboard review with me. Turn that into an execution plan with owners and a measurement setup.",
      "expected_output": "A plan that classifies the Slack Marketplace as data-dark, refuses both the A/B test and the review campaign as unexecutable on that platform, and substitutes a write-once listing, external traffic generation, and a self-built install ledger based on the OAuth token store plus uninstall events.",
      "files": [],
      "expectations": [
        "States that the Slack Marketplace exposes no listing analytics to publishers: no views, no installs, no conversion, and no search-term data.",
        "Classifies the Slack Marketplace as a data-dark marketplace and says the measure-and-iterate loop does not apply there.",
        "Declines the six-week A/B test specifically because no funnel exists to read a result from, not merely because the sample is small.",
        "States that Slack exposes no review or rating system to submitters, so the 40-review goal cannot be executed on that marketplace at all.",
        "Recommends a self-built install ledger and names the vendor's own OAuth token store as the install record.",
        "Names subscribing to the app_uninstalled event, or an org-level uninstall event, as the churn signal to wire up.",
        "States that Slack itself monitors the app's usage and uninstalls and contacts the developer directly.",
        "Advises writing the best listing once and driving traffic from outside the marketplace instead of iterating on the listing.",
        "Notes that Slack discoverability runs on editorial curation and categories rather than a published or inferable ranking algorithm.",
        "Presents the absent analytics as deliberate platform design rather than a documentation gap or a request the vendor could file.",
        "Flags an unmaintained Slack listing as a removal-risk trigger and assigns a named owner to a periodic refresh."
      ]
    },
    {
      "id": 2,
      "prompt": "Trailhaus here, we make a bundle and volume-pricing app for Shopify merchants, live 8 months, sitting at about 140 installs. The app is currently called 'Trailhaus' with the subtitle 'Smarter pricing for growing stores'. Everything I've read says the name and subtitle are basically the whole ranking game, so I want to rename it to 'Bundle Discounts & Volume Pricing - Trailhaus' and set our five search terms to: 'bundle discounts for shopify stores', 'volume pricing app', 'increase average order value', 'quantity breaks', 'tiered pricing'. Sanity-check this and hand me the final strings.",
      "expected_output": "A corrected title that leads with the brand per Shopify's own naming instruction, rejected search terms that violate the one-concept rule, a provenance label on the 80% weighting claim, and a keyword set sourced from real buyer language.",
      "files": [],
      "expectations": [
        "Recommends a brand-first title with 'Trailhaus' leading, and rejects the proposed keyword-first name.",
        "States that Shopify's own guidance instructs brand-first naming, and that the mobile app-store keyword-first convention does not transfer to it.",
        "Rejects 'bundle discounts for shopify stores' and 'increase average order value' as search terms because each must express one concept.",
        "States the search-terms rule that each term may be up to five complete words.",
        "Labels the claim that name plus subtitle carry roughly 80% of keyword ranking weight as a practitioner estimate rather than a Shopify-official statement.",
        "Establishes which fields the marketplace actually indexes for search instead of assuming all listing text counts.",
        "Sources the keyword set from language buyers actually type, citing support tickets, sales-call notes, or the marketplace's own search-term report.",
        "Distinguishes keyword coverage (terms present in indexed fields) from keyword performance (where the listing ranks), and says to track them separately.",
        "Places the keyword rewrite as the highest-efficiency lever, ahead of media overhaul, review programs, and badge work.",
        "Does not recommend reusing the same title convention on the vendor's listings on other marketplaces."
      ]
    },
    {
      "id": 3,
      "prompt": "Parcelith, Shopify app for shipping-delay notifications. We have 9 reviews at 4.6 stars, our main competitor has 240. Marketing's proposal: (a) email our 1,200 merchants offering a $50 gift card to anyone who posts a review, (b) have the six of us on the team each write one so the count doesn't look dead, and (c) do a review swap with a partner app whose founder I know, they review us and we review them. It's a one-time push for our Black Friday launch window and then we stop. Write me the campaign brief.",
      "expected_output": "A refusal of all three tactics with the specific enforcement consequences named, replaced by a sanctioned-trigger review program built on the platform's own mechanics plus post-support asks.",
      "files": [],
      "expectations": [
        "Refuses the $50 gift-card incentive and states that all five benchmarked marketplaces ban incentivized reviews by agreement.",
        "States the incentive ban covers offers made through external channels such as email, not only in-product prompts.",
        "Names the graduated penalty ladder ending in termination of the Partner account and delisting of the app, not merely removal of reviews.",
        "Warns that genuine reviews tied to an untrusted reviewer can be unpublished collaterally.",
        "Refuses the six employee-written reviews.",
        "Refuses the review swap with the partner vendor.",
        "States that reviewer-attestation systems catch both employee reviews and cross-vendor swaps.",
        "Rejects the 'one-time push then we stop' framing rather than granting a time-boxed exception.",
        "Replaces the plan with triggers the platform itself operates: its auto-invite or copyable review link, and collection routed through its reviews API.",
        "Adds an ask after a positive support interaction, specifying it must not block the user's workflow and must carry an opt-out from future asks.",
        "States that review recency and velocity outweigh lifetime totals, so closing the 240-review gap directly is not the objective."
      ]
    },
    {
      "id": 4,
      "prompt": "Board meeting in 11 days. I'm presenting the AppExchange plan for Northvane (revenue-forecasting software, listing live two years, 3.9 stars, 47 reviews). My two draft slides read: 'Title and subtitle carry ~80% of ranking weight, so rewriting them is our highest-leverage move' and 'Moving 3.9 to 4.4 stars drives +89% listing conversion, worth roughly $400K in new ARR at current traffic'. Tighten the wording and make the case airtight.",
      "expected_output": "Both claims relabelled by provenance tier, the ARR extrapolation refused, and the board commitment rewritten to promise direction rather than magnitude.",
      "files": [],
      "expectations": [
        "Labels the roughly 80% name-and-subtitle weighting as practitioner inference, and notes it originates as an estimate about a different marketplace rather than an AppExchange fact.",
        "States that Salesforce publishes nothing about its AppExchange search algorithm, so ranking advice for it is inferred end to end.",
        "Labels the +89% conversion figure as consumer mobile app-store data whose direction transfers but whose magnitude does not.",
        "Applies an explicit provenance tier to each ranking claim, ordered vendor-official above corroborated practitioner inference above single-source inference above mobile-ASO analogy.",
        "Rewrites the board commitment to promise direction rather than a specific magnitude of lift.",
        "Refuses to derive the $400K ARR figure from the +89% number.",
        "Identifies the Trailblazer Score as a partner-level health metric tied to visibility, not a per-listing ranking factor.",
        "Names the only vendor-official ranking statements available anywhere in this discipline: Atlassian's on reviews, rating and rich media, and Shopify's Built for Shopify search boost.",
        "Treats the newer intent-based AppExchange search favouring outcome-oriented language as emerging and still settling rather than established fact.",
        "Does not quote a generic trust-badge conversion-lift figure as applicable to a B2B app listing."
      ]
    },
    {
      "id": 5,
      "prompt": "Okrahead is listed on both the Shopify App Store and AppExchange. I have exactly one engineering quarter to spend on trust and credibility work, and my CTO's ordering is: (1) the Salesforce Security Review badge first, because enterprise logos care about it most, (2) Atlassian Cloud Fortified next since we launch there in H2, (3) Built for Shopify last, it's just a sticker. Confirm the sequencing and give me the quarter plan.",
      "expected_output": "A corrected ranking that puts Built for Shopify first on confirmed ranking effect, reclassifies Security Review as a listing gate rather than a badge, and questions whether the badge quarter should be spent at all before cheaper levers ship.",
      "files": [],
      "expectations": [
        "Corrects that the Salesforce Security Review is a listing gate rather than a badge, and that failing it delists the app.",
        "Ranks Built for Shopify first among the badge candidates because it is the only one carrying a vendor-official ranking-boost statement.",
        "States that Atlassian Cloud Fortified and Runs on Atlassian buy enterprise-procurement trust with no published quantified ranking lift.",
        "Ranks badge candidates by confirmed ranking effect rather than perceived prestige.",
        "Places badge pursuit as the lowest-efficiency rung of the optimization ladder, costing a quarter or more of engineering and process work for movement other levers buy faster.",
        "Names the two conditions that promote badge pursuit anyway: an officially confirmed ranking effect, or enterprise procurement in the segment gating on the badge.",
        "Asks whether procurement in their segment actually demands a badge before the quarter is committed.",
        "Recommends working one marketplace at a time rather than running badge pursuit on both in parallel.",
        "Says the keyword rewrite and first-screen media work should consume hours before the badge quarter, because they buy more rank and conversion movement per hour.",
        "Does not present passing the Security Review as a ranking lever."
      ]
    },
    {
      "id": 6,
      "prompt": "Fernlight goes live on the Atlassian Marketplace and the Shopify App Store on the same day next month. Our designer produced one asset kit we plan to use on both: a 2 minute 40 second promo video with voiceover, customer logos, and a closing card reading 'from $19/mo', plus six screenshots exported at 1600x900 where image 1 is our marketing site homepage, image 2 is our logo on a gradient, and images 3 to 6 are product views with a 5-star badge overlaid in the corner. Give me the upload checklist for both marketplaces.",
      "expected_output": "A per-marketplace asset checklist that splits the video, resizes the screenshots, strips the banned pricing and rating claims, replaces the homepage and logo images, and reorders the first three gallery slots.",
      "files": [],
      "expectations": [
        "States that Atlassian caps listing video at 30 seconds and requires a function demonstration in an anti-marketing tone, so the 2:40 promo cannot be reused there.",
        "States that Shopify allows a two-to-three-minute promotional video but caps screencast footage at 25% of runtime.",
        "Gives different screenshot dimensions per marketplace, naming 1600x900 for Shopify and 1840x900 for Atlassian.",
        "Flags the 'from $19/mo' closing card as a pricing claim banned from listing imagery.",
        "Flags the 5-star badge overlay as a review claim banned from listing imagery.",
        "Rejects the logo-on-gradient image and notes that Shopify hard-rejects logo-only and duplicate images.",
        "Rejects the marketing-site homepage as the first image and requires screenshots to be unique, in-product, and captioned.",
        "Orders the first three gallery assets as hero, then key differentiator, then most-used or most-loved feature.",
        "Labels the first-three gallery ordering and the evidence that most visitors never scroll past the third image as consumer app-store analogy rather than B2B fact.",
        "States that demo video is platform-gated: worth building where the marketplace surfaces it prominently, skippable where it is buried.",
        "Notes that Atlassian caps the listing at two categories and downgrades category-stale apps to uncategorized."
      ]
    },
    {
      "id": 7,
      "prompt": "Glasswing, revenue-ops app on AppExchange, listing live 14 months and nobody has touched it. I want a full keyword and screenshot overhaul shipped next week. Two things to know: we have no listing numbers because Salesforce doesn't hand analytics to partners unless you're Summit tier, and roughly 31% of the orgs that install us uninstall within 45 days. My theory is that a stronger listing brings in better-fit customers and the churn sorts itself out. Plan the overhaul.",
      "expected_output": "Instrumentation turned on first using the free AppExchange analytics the user wrongly believes are gated, the 31% uninstall rate treated as a blocker ahead of the overhaul, and a standing refresh cycle with a named owner.",
      "files": [],
      "expectations": [
        "Puts instrumentation before any content change, on the grounds that an optimization that cannot be measured is a guess.",
        "Corrects the Summit-tier assumption: AppExchange marketplace analytics are free to any partner with an active listing.",
        "Names what that analytics surface exposes, including tile views, unique visitors, installs and lead events broken out by traffic source.",
        "Names the report of top AppExchange search terms driving visits and installs, and routes the keyword rewrite through it instead of through guesswork.",
        "Distinguishes the post-install telemetry API as churn and renewal tooling rather than listing-conversion data.",
        "Treats the 31% uninstall rate as a blocker to fix before the listing overhaul rather than a problem a better listing solves.",
        "Rejects the user's stated theory that better-fit traffic will make the churn sort itself out.",
        "States that uninstall rate feeds ranking negatively where corroborated and that platforms monitor it independently of the vendor.",
        "Warns that new installs won by a better listing become the uninstalls that sink the listing if the product gap stands.",
        "Establishes a refresh cycle of 30 to 60 days with a named owner, and asks who owns the listing day to day and what changed on it in the last 90 days.",
        "Lists view-to-install conversion, review velocity over the trailing 90 days, and uninstall rate among the operating KPIs."
      ]
    },
    {
      "id": 8,
      "prompt": "Perchcode, we're on the HubSpot App Marketplace. Nine days ago we swapped our tagline and installs are up 1.6% since. I want to (a) roll the new tagline out to our Shopify and Atlassian listings this week, and (b) commit $8K/month of standing budget to paid marketplace placement on whichever of our marketplaces sells it, because organic has been flat since spring. Also pull our real view-to-install conversion so I can report it, HubSpot must surface that somewhere. Write it up as a one-pager for my CEO.",
      "expected_output": "The 1.6% movement classified as noise, the HubSpot funnel correctly reported as nonexistent, the cross-marketplace rollout refused, and paid placement scoped to a time-boxed accelerant after the content levers ship.",
      "files": [],
      "expectations": [
        "Treats the 1.6% movement as noise and names the threshold below which a lift is not acted on as roughly 2%.",
        "Labels the significance thresholds as borrowed from consumer app-store A/B practice rather than measured on B2B marketplaces.",
        "Notes that B2B listing traffic is low, so reaching significance takes proportionally longer than the consumer benchmarks imply.",
        "States that HubSpot exposes install counts but no view-to-install funnel, so the requested conversion figure does not exist to report.",
        "Gives HubSpot's active-install definition: unique production accounts unaffiliated with the vendor's own organization showing successful app activity within the past 30 days.",
        "Refuses to roll the tagline out to the Shopify and Atlassian listings on the strength of this result.",
        "Rejects the standing $8K/month paid budget and confines paid marketplace placement to a time-boxed accelerant.",
        "States that paid installs mask whether the listing converts on its own.",
        "Requires the content levers, keyword rewrite and first-screen media, to ship before paid placement is switched on.",
        "States that HubSpot publishes nothing official about its ranking, so any HubSpot ranking advice in the one-pager is inference."
      ]
    }
  ],
  "trigger_queries": [
    {
      "query": "Our Salesforce AppExchange listing gets about 4,000 tile views a month and 30 installs. What do we fix first?",
      "should_trigger": true
    },
    {
      "query": "How do we rank higher in Shopify App Store search?",
      "should_trigger": true
    },
    {
      "query": "We submit to the Atlassian Marketplace in five weeks. Help me write the listing.",
      "should_trigger": true
    },
    {
      "query": "Need a plan to get our first 20 reviews on our marketplace app without breaking any rules",
      "should_trigger": true
    },
    {
      "query": "Is it against the rules to give merchants a discount code for leaving us a review?",
      "should_trigger": true
    },
    {
      "query": "our app shows up on page 4 when people search the one term we care about. how do we move up",
      "should_trigger": true
    },
    {
      "query": "What should the first three screenshots on our listing actually show?",
      "should_trigger": true
    },
    {
      "query": "we're on the hubspot app marketplace and have no idea if anyone even sees the page",
      "should_trigger": true
    },
    {
      "query": "Should we chase the Built for Shopify badge this quarter?",
      "should_trigger": true
    },
    {
      "query": "Installs have been flat for two quarters even though traffic to the page looks fine.",
      "should_trigger": true
    },
    {
      "query": "help me rewrite our app title and subtitle so buyers actually find us",
      "should_trigger": true
    },
    {
      "query": "Do marketplace search ads actually work or are we just buying installs?",
      "should_trigger": true
    },
    {
      "query": "Our competitor has 400 reviews and we have 12. How do we catch up?",
      "should_trigger": true
    },
    {
      "query": "What's a sane refresh cadence for a listing we published a year ago and never touched?",
      "should_trigger": true
    },
    {
      "query": "Our Slack app is live but we can't tell whether anyone installs it from the directory.",
      "should_trigger": true
    },
    {
      "query": "We got rejected for duplicate images. What now?",
      "should_trigger": true
    },
    {
      "query": "Which category should we pick for our app, and does the choice affect search?",
      "should_trigger": true
    },
    {
      "query": "Somebody left us a 2-star review saying onboarding is broken. Do we reply publicly?",
      "should_trigger": true
    },
    {
      "query": "Audit our app store page and tell me what's costing us installs",
      "should_trigger": true
    },
    {
      "query": "How long should the demo video on our listing be?",
      "should_trigger": true
    },
    {
      "query": "We want to go from 3.8 to 4.4 stars this quarter. What's a realistic plan?",
      "should_trigger": true
    },
    {
      "query": "Is it worth wiring GA4 and a pixel into our partner dashboard listing?",
      "should_trigger": true
    },
    {
      "query": "we launch in six weeks and the listing is still a placeholder. triage it for me",
      "should_trigger": true
    },
    {
      "query": "What words should go in the search terms field, and how many?",
      "should_trigger": true
    },
    {
      "query": "Our page gets plenty of views and almost nobody clicks install.",
      "should_trigger": true
    },
    {
      "query": "How do I tell whether the ranking advice I'm reading came from the platform or from someone's guess?",
      "should_trigger": true
    },
    {
      "query": "Our directory listing has been slowly sliding down the results all year.",
      "should_trigger": true
    },
    {
      "query": "Can we ask a customer for a review right after we close their support ticket?",
      "should_trigger": true
    },
    {
      "query": "Planning our AppExchange presence for next year. What does the standing program look like?",
      "should_trigger": true
    },
    {
      "query": "we're an unknown brand in this ecosystem. should our app name lead with the brand or the keyword?",
      "should_trigger": true
    },
    {
      "query": "What should I report monthly on our marketplace app?",
      "should_trigger": true
    },
    {
      "query": "Nearly everyone who installs us churns within a month, but we still want more installs.",
      "should_trigger": true
    },
    {
      "query": "Should we use the same copy and screenshots on all four marketplaces we're on?",
      "should_trigger": true
    },
    {
      "query": "Help me build a review request flow inside the product.",
      "should_trigger": true
    },
    {
      "query": "We got 60 installs after certification came through. Was it the certification?",
      "should_trigger": true
    },
    {
      "query": "I need to convince my VP that rewriting the tagline is worth two days of work.",
      "should_trigger": true
    },
    {
      "query": "we swapped our screenshots and conversion moved 1.8%. ship it everywhere?",
      "should_trigger": true
    },
    {
      "query": "What does the partner dashboard actually give us for our listing?",
      "should_trigger": true
    },
    {
      "query": "Our listing converts fine on one marketplace and terribly on another. Why?",
      "should_trigger": true
    },
    {
      "query": "Is it okay if a few of us on the team write honest reviews for our own app?",
      "should_trigger": true
    },
    {
      "query": "We're writing the listing rulebook every third-party app on our marketplace must follow.",
      "should_trigger": false
    },
    {
      "query": "Define objective reject criteria for screenshots submitted to our app directory.",
      "should_trigger": false
    },
    {
      "query": "Design the approval pipeline and permission audit for apps submitted to our platform.",
      "should_trigger": false
    },
    {
      "query": "How should we handle appeals when we reject a partner's app?",
      "should_trigger": false
    },
    {
      "query": "Pick the founding launch cohort for the app marketplace we're standing up.",
      "should_trigger": false
    },
    {
      "query": "Design our app-of-the-month spotlight governance and the MDF budget split.",
      "should_trigger": false
    },
    {
      "query": "What take-rate should our connector marketplace charge partners?",
      "should_trigger": false
    },
    {
      "query": "Should we build our own integration marketplace or just join other people's?",
      "should_trigger": false
    },
    {
      "query": "Merchant-of-record posture and payout hold windows for our app store.",
      "should_trigger": false
    },
    {
      "query": "How do we collect revenue share from paid apps listed on our marketplace?",
      "should_trigger": false
    },
    {
      "query": "Design the sandbox provisioning and dev-account flow for partner developers.",
      "should_trigger": false
    },
    {
      "query": "Which of these 30 integration partners deserve a deep co-build?",
      "should_trigger": false
    },
    {
      "query": "Set up sourced versus influenced pipeline attribution for our ISV alliances.",
      "should_trigger": false
    },
    {
      "query": "We want to be a source connector on the big ELT platforms. Which build path?",
      "should_trigger": false
    },
    {
      "query": "Does our API support log-based CDC or only pseudo-CDC for connector certification?",
      "should_trigger": false
    },
    {
      "query": "Design our public API's error-code taxonomy.",
      "should_trigger": false
    },
    {
      "query": "Write the rate-limit policy for our public API.",
      "should_trigger": false
    },
    {
      "query": "How should we version our public API without breaking integrators?",
      "should_trigger": false
    },
    {
      "query": "Plan the information architecture of our developer portal.",
      "should_trigger": false
    },
    {
      "query": "Should we ship an MCP server for our product?",
      "should_trigger": false
    },
    {
      "query": "What does a staff-level platform engineer interview loop look like?",
      "should_trigger": false
    },
    {
      "query": "Run an ASO audit on our iOS app and fix the keyword field.",
      "should_trigger": false
    },
    {
      "query": "Our Google Play install rate dropped after the last update. What happened?",
      "should_trigger": false
    },
    {
      "query": "Design App Store screenshots for our consumer fitness app.",
      "should_trigger": false
    },
    {
      "query": "Help us get featured on the App Store Today tab.",
      "should_trigger": false
    },
    {
      "query": "A/B test our Google Play store listing with Store Listing Experiments.",
      "should_trigger": false
    },
    {
      "query": "Reply to these one-star reviews on our iPhone game.",
      "should_trigger": false
    },
    {
      "query": "Configure EAS submit to push our builds to both mobile app stores.",
      "should_trigger": false
    },
    {
      "query": "What's the current review turnaround time for Apple's App Review team?",
      "should_trigger": false
    },
    {
      "query": "Localize our mobile app store listing into 12 languages.",
      "should_trigger": false
    },
    {
      "query": "Optimize our Amazon product listing: title, bullets, and A+ content.",
      "should_trigger": false
    },
    {
      "query": "We sell on a Shopify storefront. Improve our product page conversion rate.",
      "should_trigger": false
    },
    {
      "query": "Write the npm README for our SDK so more people adopt it.",
      "should_trigger": false
    },
    {
      "query": "Improve our SEO so we rank on Google for 'invoice automation software'.",
      "should_trigger": false
    },
    {
      "query": "Write a case study about a customer using our Salesforce integration.",
      "should_trigger": false
    },
    {
      "query": "How should we price the paid tiers of our developer tool?",
      "should_trigger": false
    },
    {
      "query": "Draft our answers to an enterprise buyer's security questionnaire.",
      "should_trigger": false
    },
    {
      "query": "Set up webhook delivery retries and signature verification.",
      "should_trigger": false
    },
    {
      "query": "Design the OAuth authorization server our partners authenticate against.",
      "should_trigger": false
    },
    {
      "query": "Build a bulk data export API with signed URLs.",
      "should_trigger": false
    }
  ]
}

# Mobile-ASO analogy toolkit

Audit rubrics and gallery-ordering frameworks from consumer mobile app stores (Apple App Store, Google Play) transfer structurally to B2B integration marketplaces, but the benchmark numbers do not. B2B listing traffic is orders of magnitude lower and the buyer is deliberate rather than impulse-browsing, so no figure below is verified to hold at the same magnitude on AppExchange, HubSpot, Atlassian, Shopify, or Slack. Use the frameworks; quote the numbers only as "in consumer app stores, X - direction likely, magnitude unverified here."

## Converged audit dimensions

Two unrelated authors independently arrived at near-identical weighted audit rubrics: one six-way, one ten-way splitting the same factors finer and adding Keyword Rankings as its own dimension.

| Dimension            | Six-way weight | Ten-way treatment  |
| -------------------- | -------------- | ------------------ |
| Title & Subtitle     | 20%            | split finer        |
| Description          | 15%            | split finer        |
| Visual Assets        | 25%            | split finer        |
| Ratings & Reviews    | 20%            | split finer        |
| Metadata & Freshness | 10%            | split finer        |
| Conversion Signals   | 10%            | split finer        |
| Keyword Rankings     | not separate   | own dimension, 10% |

The convergence, not either weighting, is the finding. Any listing audit should score:

- Title text.
- Description.
- Visual assets.
- Ratings and reviews.
- Keyword and category placement.
- Freshness.
- Conversion signals.

Track **keyword coverage** (terms present in indexed fields) separately from **keyword performance** (whether the listing actually ranks for them). Visual assets plus ratings carry the most combined weight in both rubrics; text metadata alone is a minority of either score.

Data-collection pattern worth copying:

1. Try live listing data first: fetch the public listing page, and capture images rather than relying on screenshot captions, which are usually not text-extractable.
2. Fall back to asking the user to paste their current listing fields.

For a marketplace-generic skill with no API, the fallback is the default path.

## Brand-maturity tiering

Classify the listing before scoring it: **Dominant / Established / Challenger**. A deviation from textbook optimization - brand-only title, no keyword-first strategy, no demo video - is a legitimate deliberate choice for a market leader whose buyers search its name, and a real gap for an unknown entrant. Rule of thumb: before docking points, ask whether this is a mistake or a deliberate choice by a team that has data you don't.

This transfers directly: an established SaaS's marketplace listing can lean on brand recognition in ways a new entrant's cannot, so the same checklist scores them differently. It also interacts with title convention, since at least one B2B marketplace instructs brand-first titles for everyone.

## Gallery ordering: the first-3 rule

Consumer data (two independent sources):

- Roughly 80-90% of store visitors never see past the third screenshot.
- Average scroll rate ~17%.
- The decision takes 6-10 seconds.

The ordering framework this justifies transfers to any marketplace listing with an ordered gallery:

| Position | Content                         | Purpose                                                                                                                                                |
| -------- | ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
| 1        | Hero - core value, best feature | Stop the scroll; say what the product does                                                                                                             |
| 2        | Key differentiator              | Why this one over competitors                                                                                                                          |
| 3        | Most-used / most-loved feature  | What users actually value                                                                                                                              |
| 4        | Social proof or outcome         | Ratings, results, testimonials - where the marketplace's rules allow them in images; several B2B marketplaces ban review claims baked into screenshots |
| 5+       | Supporting features             | Settings, integrations, edge cases                                                                                                                     |

Optimal consumer count: 4-6 images - more produced decision paralysis, not conversion.

## Conversion benchmarks (consumer-only numbers)

- **Ratings**:
  - 3.0 → 4.0 stars: +89% conversion.
  - 4.0 → 4.5 stars: +20-30% conversion.
  - A 0.4-star gap vs a competitor: ~25% lost installs from the same search.
  - 4.0 is a hard floor for featuring; below 3.5, visibility drops sharply.
  - 79% of users check ratings before installing.

  The strongest reusable _direction_: rating improvements compound into conversion, so review work is never cosmetic.

- **Screenshots**: well-designed sets lifted conversion 20-35%; A/B winners typically +10-25%.
- **Preview video**: +20-40% on the platform that autoplays it, while only ~6% of visitors tapped play on the platform that doesn't - the evidence behind "demo video is platform-gated, not universal." Check how prominently the target B2B marketplace surfaces video before investing in one.

## A/B significance thresholds

A generic before/after rule reusable on any marketplace with funnel data, at whatever traffic volume - but note B2B listing traffic is low, so reaching significance takes proportionally longer:

- Lift >10%: strong winner, apply immediately.
- 5-10%: meaningful.
- 2-5%: marginal.
- <2%: noise - do not act on it.

# Five-marketplace lever benchmarks

Submitter-side comparison of Salesforce AppExchange, HubSpot App Marketplace, Atlassian Marketplace, Shopify App Store, and Slack Marketplace. All five platforms actively change listing rules (image-uniqueness rules, recategorizations, search-engine migrations) - verify current specs against official docs before acting on any guidance below.

## Ranking factors: confirmed vs inferred

| Marketplace            | Officially confirmed                                                                                                                                                                                                                                                                                                  | Practitioner-inferred                                                                                                                                                                                                                                                                                                                                                                                           |
| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Atlassian              | Search runs on OpenSearch (replaced Algolia, 2026), relevance scored via NDCG; "more reviews and a higher rating improve your engagement score and search rank"; "rich media drives engagement and trial starts"; relevance prioritized over keyword matching alone. All verbatim from Atlassian's own rankings page. | Summary field (250-character cap) as the primary search-relevance string.                                                                                                                                                                                                                                                                                                                                       |
| Shopify                | "Built for Shopify apps appear higher in Shopify App Store search rankings" - the clearest official badge-ranking statement on any of the five.                                                                                                                                                                       | Name + subtitle ≈80% of keyword ranking weight (one tracker's estimate); install velocity over lifetime totals; listing view-to-install conversion; review recency weighted, especially the last 90 days ("50 reviews in 30 days outranks 500 static reviews"); uninstall rate feeding back negatively. Search drives ~60% of installs; one app reported +11.8% installs from changing only its featured image. |
| Salesforce AppExchange | Nothing on the search algorithm. The Trailblazer Score (a partner-health metric) is confirmed to influence "enhanced AppExchange visibility" - partner-level, not per-listing. Security Review is a listing gate, not a ranking factor: failing it delists.                                                           | Clean SEO title/description/categories/keywords drive indexing; media quality supports engagement and dwell; review volume/recency and install counts matter. The 2026 unified marketplace's intent-based search reportedly rewards outcome-oriented language - recent and still settling, treat as emerging.                                                                                                   |
| HubSpot                | Nothing.                                                                                                                                                                                                                                                                                                              | Install growth, ratings/reviews, certification status, content quality, internationalization - from third-party grader criteria and general practitioner guidance.                                                                                                                                                                                                                                              |
| Slack                  | Nothing published, and no analytics surface exists to reverse-engineer one from. Discoverability runs on editorial curation, categories, and the app's own external marketing.                                                                                                                                        | Nothing exists to infer from.                                                                                                                                                                                                                                                                                                                                                                                   |

## Native submitter analytics

| Marketplace            | What submitters get                                                                                                                                                                                                                                                                                                                      | How                                                                                   |
| ---------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- |
| Shopify                | Fullest stack: partner-dashboard installs/uninstalls/net installs, ratings, revenue, API error rates; officially supported GA4 Measurement ID and Meta Pixel wiring that passes server-side install events, giving a complete listing-view → install-click → install funnel per shop. Largest third-party tracker ecosystem of the five. | Partner dashboard; Distribution → Manage listing → Tracking (must be wired manually). |
| Atlassian              | Richest native funnel: sales data, license report, evaluations/leads, marketing-funnel and evaluation insights (paid apps; up to 24h lag). Do not confuse with "Atlassian Analytics", a separate end-user BI product.                                                                                                                    | Manage Vendor screen, mirrored by the sales-report REST API.                          |
| Salesforce AppExchange | Free to any partner with an active listing: tile views, hovers, unique visitors, lead events, installs, by traffic source - plus top AppExchange search terms driving visits and installs. A separate post-install telemetry API serves churn/renewal work, not listing conversion.                                                      | Partner Community, Publishing → Analytics tab.                                        |
| HubSpot                | Active-install counts ("unique production accounts, unaffiliated with your organization, showing successful app activity within the past 30 days") plus customer feedback. No view→install funnel - a confirmed gap, not an extraction miss.                                                                                             | Developer account, Development → App Listings.                                        |
| Slack                  | Nothing. No views, installs, conversion, or search-term data for publishers; Slack itself monitors usage/uninstalls and contacts the developer. Self-build instead: your OAuth token store is the install ledger; subscribe to the `app_uninstalled` (and org-level uninstall) events for churn.                                         |

## Review mechanics and enforcement

All five ban incentivized reviews by agreement. Enforcement maturity diverges - Shopify is the only one with a published graduated-penalty schedule and documented mass takedowns.

- **Shopify**: policy states incentivizing reviews risks "removal of a portion of reviews, demotion or delisting of your app, or termination of your Partner account"; the ban covers incentives "through the app or external channels", and genuine reviews tied to untrusted reviewers can be unpublished collaterally. Sanctioned mechanics are prescribed:
  - In-product asks at natural moments that never block workflow.
  - Asks after positive support interactions.
  - Merchant opt-out.
  - Collection routed through the Reviews API.

  Email campaigns to active users convert at 5-10% **[corroborated practitioner report]**.

- **Atlassian**: solicitation is platform-driven. Customers are auto-invited 30 days after purchase, and a listing needs 10 reviews before a star rating displays. Publishers get a copyable review link and one public reply per review, and a reviewer must have the app installed or have uninstalled within 14 days.
- **Salesforce AppExchange**: actively coaches solicitation:
  - Target specific customers.
  - Time the ask to a success milestone.
  - Solicit even from evaluators.
  - Engage negative reviews publicly to get them amended.
  - Keep reviews fresh.

  The Trusted Reviews program badges reviews from high-credential community members, so reviewer credibility is weighted visibly. The incentive ban is explicit for consulting listings; the app-listing wording is thinner but the principle is contractual.

- **HubSpot**: reviewer attestation - reviewers confirm they are actual users, one review per app, no reviewing your own, your employer's, or a competitor's app. A dedicated ticket type exists for reporting violations, but no active solicitation mechanic is documented.
- **Slack**: no review or rating system exposed to submitters at all - no tactic to execute and no policy to violate.

## Listing field constraints submitters optimize within

The operator-side rulebooks converge on one skeleton - short hook string, benefit-led body, fixed-geometry gallery, capped video, capped categories, off-listing trust surfaces gating approval - but the specifics diverge, so never carry one marketplace's numbers to another:

- **Title convention**: Shopify instructs brand-first naming ("QTeck Announcement Bar", not "Announcement Bar QTeck") - the opposite of mobile keyword-first advice. Atlassian's 250-character summary is the load-bearing search string. Establish which fields the target marketplace indexes before writing.
- **Search terms [official, Shopify]**: up to five complete words, one concept per term ("email marketing" valid; "email marketing for leads" not).
- **Screenshot geometry**: no universal ratio exists.

  | Marketplace | Ratio                                 |
  | ----------- | ------------------------------------- |
  | Shopify     | 1600×900                              |
  | Atlassian   | 1840×900                              |
  | Slack       | 1600×1000                             |
  | Salesforce  | 700×467 (roughly 3:2)                 |
  | HubSpot     | content-correctness rules, not pixels |

  Shopify's 2026 image-uniqueness rule hard-rejects duplicate and logo-only images. Common content bans: browser chrome, sensitive data, pricing/review claims baked into images.

- **Video ceilings**:

  | Marketplace | Ceiling                                                                    |
  | ----------- | -------------------------------------------------------------------------- |
  | Atlassian   | ≤30 seconds, function demo only, anti-marketing tone                       |
  | Slack       | 30-90 seconds                                                              |
  | Shopify     | 2-3 minutes, promotional tone allowed, screencast capped at 25% of runtime |
  | AppExchange | 1-3 minutes                                                                |

- **Taxonomy caps**:

  | Marketplace | Caps                                                              |
  | ----------- | ----------------------------------------------------------------- |
  | Salesforce  | one category plus two industries (tightest)                       |
  | Atlassian   | two of ten, with a staleness downgrade to "uncategorized"         |
  | Shopify     | broader categories plus up to 25 structured features per category |
  | HubSpot     | operator-curated persona collections layered over self-selection  |

- **Freshness signals**: Atlassian downgrades category-stale apps, Slack names an unmaintained listing a removal-risk trigger, and Shopify treats update recency as a freshness signal. This is why the defend stage's 30-60 day refresh exists.

Exact character-limit tables for AppExchange's and HubSpot's long-form fields are not fully published anywhere - they surface only inside each platform's own listing builder. Read them there at submission time.

## Badges with confirmed effects

- **Built for Shopify** - the only badge on the five with a vendor-official ranking-boost statement. Highest-priority pursuit where eligible.
- **Atlassian Cloud Fortified / Runs on Atlassian** - enterprise-procurement trust; no quantified ranking lift published. **[official programs, unquantified effect]**
- **HubSpot Certified App** - search-filter and curated-collection eligibility; losable and re-appliable on a cooldown.
- **Salesforce** - Security Review is a gate, not a badge (failure delists); the Trailblazer Score is the visibility-adjacent lever.
- Generic "trust badges lift conversion 15-30%" figures originate in consumer e-commerce storefronts, not B2B app listings - never quote them at face value here.

## Practitioner ecosystem

Depth is uneven - deepest where the platform exposes data worth analyzing:

- **Salesforce AppExchange**: Invisory, AppX, Salesforce Ben, Beyond The Cloud, Synebo, Peeklogic; Salesforce's own Summit-tier listing consultation.
- **Shopify**: AppJubilee, Prys, Fusionmetrics, AppstorePulse, Big Moves Marketing - all grounded in Shopify-native signals.
- **Atlassian**: thin - aety.io, SaaSJet's Marketplace Reporter; Atlassian's own docs are the primary framework.
- **HubSpot**: thin - third-party listing graders and HubSpot Academy's certification course.
- **Slack**: none, because there is no data or review surface to build a discipline on.
- **Mobile-ASO firms** (Gummicube, Phiture, SplitMetrics and peers): consumer app-store specialists whose frameworks are frequently analogized to B2B marketplaces - treat any advice taken from them as analogy, however confidently it is stated.

---
name: integration-listing-optimization
description: Optimize a B2B SaaS vendor's own listing on a third-party app marketplace - Salesforce AppExchange, HubSpot, Atlassian Marketplace, Shopify App Store, Slack, or comparable directories. Covers the view-to-install funnel, keyword placement in title and tagline, screenshot and demo-video ordering, category choice, compliant review-velocity programs (incentivized reviews are banned everywhere), badge pursuit ranked by confirmed ranking effect, and defending rank against listing decay. Use whenever the user mentions AppExchange, marketplace listing conversion, marketplace search rank, app installs, or app reviews - even if they never say "listing optimization". Do NOT use for the operator's own rulebook - use samber/developer-platform-skills@app-marketplace-listing-standards instead.
license: MIT
metadata:
  author: Samuel Berthe
  version: "1.0.2"
---

# Integration Listing Optimization

You are a marketplace-growth advisor to a B2B SaaS vendor whose product is listed - or about to be - on a third-party app marketplace. Optimize the listing the vendor owns on someone else's platform: get found in marketplace search, convert views to installs, build review velocity without breaking platform policy, and defend the position once won. The method is marketplace-generic; the five platforms benchmarked behind it are Salesforce AppExchange, HubSpot App Marketplace, Atlassian Marketplace, the Shopify App Store, and the Slack Marketplace.

This skill is the mirror image of operator-side listing standards (`samber/developer-platform-skills@app-marketplace-listing-standards`): the rulebook that skill writes for a marketplace operator is exactly the constraint set you optimize within here. So read the target marketplace's own listing rules before touching anything - every lever below plays inside those rules, never against them.

**Memory (advised):** When memory lives in a file, consider using `developer-platform-context.md`; if a different memory system is in use, rely on that instead. The file is an advisory reference, not a mandatory requirement. Separate task info in different sections. Remove finished tasks. Add a date to a task; no date for general project context. Some interview responses may differ between 2 tasks.

## Interview

Ask these before proposing anything. One question per message, multiple-choice where offered - each answer redirects a menu below.

1. Which marketplace(s), and is each listing live or pre-submission? Work one marketplace at a time - every lever below is weighted per platform.
2. What funnel data exists today - can you see listing views, installs, and the conversion between them? This decides menu 1, and whether this marketplace is a data-driven play at all.
3. How recognizable is your brand inside this ecosystem - an established name buyers search for, or an unknown entrant? This moves the title strategy (see the brand-maturity tiers in [references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md)).
4. What words do buyers actually type when they look for what you do - from support tickets, sales-call notes, or the marketplace's own search-term report where one exists?
5. Where does the review base stand - count, average rating, recency of the newest ten? And what customer moments could time an ask: support resolutions, success milestones, renewals?
6. Any badge or certification earned, in progress, or gating deals - does enterprise procurement in your segment ask for one?
7. Who owns this listing day-to-day, and what changed on it in the last 90 days?
8. By when must results land - a launch, a renewal season, a hard date, or open-ended?
9. Is this a one-off win (unblock a launch, beat one competitor to a keyword) or a compounding asset (a listing that keeps ranking as the catalog grows)?
10. What is the effort ceiling - owner hours per month, engineering time for instrumentation, and appetite for changing support or lifecycle workflows to carry review asks?

Questions 8-10 exist because the menus below diverge sharply on time-to-effect, durability, and effort:

- A hard deadline promotes text-only levers over standing programs.
- A compounding-asset mandate promotes the standing programs.
- A low effort ceiling deletes options rather than demoting them.

Every ranking below is a default, not a law. Re-rank each menu against the answers, and against any unfair advantage the vendor already owns - an in-house designer, a large active-user email base, a support team with high satisfaction scores. Say which answer moved which rung.

## Two splits that change the plan

All five benchmarked marketplaces sell software to businesses; the consumer app stores enter this discipline only as structural analogy - their audit rubrics transfer, their numbers do not ([references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md)). Two splits change the plan:

- **Data-rich vs data-dark marketplace.** What each platform exposes to submitters:
  - Shopify, Atlassian, and AppExchange: a real view-to-install funnel.
  - HubSpot: install counts, no funnel.
  - Slack: nothing - no listing analytics, no review system. This is deliberate platform design, not a documentation gap; the platform itself watches your usage and uninstalls, then contacts you.

  On a data-rich marketplace, run the full measure-and-iterate loop below. On a data-dark one the loop collapses: write the best listing you can once, drive traffic from outside the marketplace, self-track installs (menu 1's starved rung), and skip the review program entirely - there is nothing to iterate against.

- **Confirmed vs inferred ranking factors.** Only two platforms in this entire discipline say anything vendor-official about ranking:
  - Atlassian: marketplace search runs on OpenSearch scored by NDCG, and more reviews, higher ratings, and rich media improve rank.
  - Shopify: the Built for Shopify badge boosts search ranking.

  Everything else - including the widely quoted "app name plus subtitle carry ≈80% of Shopify keyword weight" - is practitioner reverse-engineering, and AppExchange and HubSpot ranking advice is inferred end to end. Carry the label on every claim you make: vendor-official > well-corroborated practitioner inference > single-source inference > mobile-ASO analogy. Never present a lower tier as a higher one ([references/marketplace-lever-benchmarks.md](references/marketplace-lever-benchmarks.md)).

## Workflow

Five stages - the first pass runs them in order; stages 3 and 5 then run standing. Treat the ordering as a default, adapted to the marketplace's own review and certification lead times.

1. **Instrument before optimizing** (menu 1): turn on every analytics surface the platform grants before touching content - an optimization you cannot measure is a guess, and question 2's answer may reclassify the whole engagement as a data-dark play.
2. **Win the keyword and first-screen battle** (menu 2's top rungs): put question 4's real buyer language into the fields that carry search weight, in the title convention this marketplace rewards, then fix the first three gallery assets.
3. **Build compliant review velocity** (menu 3): a standing program on sanctioned triggers only.
4. **Pursue the badge that provably moves ranking** (menu 2's starved rung, promoted only when its condition holds).
5. **Defend the position**: refresh on a 30-60 day cycle, watch uninstalls, monitor competitors' keyword and badge moves, and re-verify the platform's current rules. All five benchmarked platforms revise their listing rules on an ongoing basis, so if you can browse the web, check the official docs before each cycle.

Validate each stage's plan with the user before moving to the next.

## 1. Instrumentation

Ranked by funnel visibility bought per setup hour:

- effort, setup plus upkeep (descending): `self-built install ledger > third-party tracker > platform-supported external wiring > native dashboards`
- funnel visibility bought: `platform-supported external wiring > native dashboards > third-party tracker > self-built install ledger`
- efficiency: `native dashboards > platform-supported external wiring > third-party tracker > self-built install ledger`

- **Default rung: native dashboards, on day one, on every marketplace that has them.**
  - AppExchange: marketplace analytics free to any partner with an active listing, including the search terms that drive visits.
  - Atlassian: vendor console exposing a view-to-evaluation-to-purchase funnel.
  - Shopify: partner dashboard showing installs, uninstalls, and revenue.

  Costs minutes. Skipping it is the most common self-inflicted blindness in this discipline.

- **Promote to platform-supported external wiring** - connecting your own web-analytics property or ad pixel into the listing where the marketplace officially supports it, which yields the full listing-view → install-click → completed-install funnel - once weekly decisions depend on funnel data. It is the highest-visibility rung but requires deliberate setup the default rung does not.
- **Third-party trackers and listing graders** add what native dashboards lack: keyword rank tracking, competitor monitoring, benchmark comparisons. Worth adding once you are contesting specific keywords against named competitors; before that they measure a battle you have not entered.
- **Starved option: the self-built install ledger** - your own auth-token store doubling as the install record, plus a subscription to the platform's uninstall event. Highest effort, least data, so it loses every efficiency round; promote it to mandatory on a data-dark marketplace, where it is the only instrumentation that exists at all.

Per-marketplace tooling detail: [references/marketplace-lever-benchmarks.md](references/marketplace-lever-benchmarks.md).

## 2. Optimization-investment ladder

Which lever gets the next block of hours. Efficiency here is search-rank and conversion movement bought per hour of vendor effort:

- effort (descending): `badge pursuit > review-velocity program > paid marketplace ads > demo video > first-screen media overhaul > keyword rewrite`
- rank and conversion movement bought: `review-velocity program > keyword rewrite > badge pursuit > first-screen media overhaul > demo video > paid marketplace ads`
- efficiency: `keyword rewrite > first-screen media overhaul > review-velocity program > demo video > paid marketplace ads > badge pursuit`

- **Default rung: the keyword rewrite** - hours of work on the fields that carry search weight (name, tagline or subtitle, search terms), fed by question 4's real buyer language. Check the marketplace's own title convention first: the benchmarked commerce marketplace instructs brand-first titles, the exact opposite of mobile-ASO keyword-first advice, so the convention does not transfer between marketplaces. Establish which fields the marketplace actually indexes for search rather than assuming all listing text counts.
- **Then the first-screen media overhaul**: unique, in-UI, captioned screenshots ordered hero → differentiator → most-loved feature (an analogy-labelled ordering - [references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md)), cropped of sensitive data, with no pricing or review claims baked into the image. One benchmarked platform now hard-rejects duplicate and logo-only images, so on it this rung is compliance, not polish.
- **The review-velocity program** (menu 3) is the standing companion, not a one-off rung. It tops the value axis because review recency and velocity are the one input both vendor-official statements and the practitioner consensus agree moves rank, but it pays over quarters, not weeks.
- **Demo video is platform-gated, not universal.** Add one where the marketplace surfaces video prominently, at that marketplace's ceiling: 30 seconds of function proof at the strictest benchmarked platform, two to three promotional minutes at the loosest. Skip it where video is buried - the analogy data shows video only converts where the surface autoplays it.
- **Paid marketplace search ads**, where the platform offers them: a time-boxed accelerant when organic install velocity stalls after the content rungs have shipped, never a standing subsidy - paid installs mask whether the listing converts on its own.
- **Starved option: badge pursuit** - a quarter or more of engineering and process work against certification criteria, and the strongest trust asset on the menu. It loses every efficiency round; promote it when the marketplace officially confirms a ranking effect (exactly one benchmarked platform does, in writing) or when question 6 says enterprise procurement gates on the badge. Rank candidate badges by confirmed effect, never by perceived prestige - and treat a mandatory security review as a listing gate to pass, not a badge to pursue.
- **Deleted, not demoted: buying installs or reviews in any form.** It sits on no rung of this ladder - see menu 3's deletion for why, and for what happens to Partner accounts that try.

## 3. Review velocity

Review recency and velocity outweigh lifetime totals: one well-corroborated practitioner claim holds that an app gaining 50 reviews in 30 days outranks one sitting on 500 static reviews. Label it inference, but plan as if the direction is true, because the one vendor-official statement on reviews agrees with it. On a marketplace with no review system, skip this menu entirely.

Ranked by published reviews per hour of outreach effort:

- effort (descending): `milestone-timed personal ask > lifecycle email campaign > negative-review engagement > post-support prompt > platform-automated trigger`
- reviews bought: `platform-automated trigger > lifecycle email campaign > post-support prompt > milestone-timed personal ask > negative-review engagement`
- compliance cost, as the policy review it triggers and the reversibility it costs: `lifecycle email campaign > milestone-timed personal ask > post-support prompt > platform-automated trigger == negative-review engagement`
- efficiency: `platform-automated trigger > post-support prompt > lifecycle email campaign > milestone-timed personal ask > negative-review engagement`

The compliance `==` holds because both tactics run entirely inside surfaces the platform itself operates: an invite the platform sends, a public reply posted through the platform's own mechanism. Neither can drift into selective or incentivized asking - they are equally exposure-free, not merely both cheap.

- **Default rung: every trigger the platform itself operates, plus the post-support ask.**
  - Turn on the platform's auto-invite where one exists.
  - Use its copyable review link.
  - Route in-product prompts through its review API where offered.
  - Add an ask after a positive support interaction, the highest-conversion sanctioned moment: prompted at a natural pause, never blocking the user's workflow, with an opt-out from future asks.
- **The lifecycle email campaign** to active users converts at a practitioner-reported 5-10%. Keep the ask unconditional and sent to all qualifying users - its compliance cost above is precisely the drift risk into asking only the happy ones, which several platforms treat as manipulation.
- **Starved option: the milestone-timed personal ask** - a named human asking a specific customer at a success milestone. Fewest reviews per hour on the menu, highest review quality, and the only tactic that reliably produces the detailed, credible reviews that platforms with reviewer-credibility weighting surface first. Promote it when the marketplace weights reviewer credibility, or when the review base is small enough that each review still moves the average.
- **Negative-review engagement**: reply publicly, once per review, aiming to fix the issue and get the review amended. Low volume bought - but the reply is read by every future evaluator, and one platform's own partner guidance explicitly coaches it.
- **Deleted, not demoted: incentivized reviews** - discounts, gift cards, feature unlocks, or withheld functionality in exchange for a review, through the product or any external channel. All five benchmarked platforms ban it by agreement, and one publishes a graduated penalty ladder that ends at Partner-account termination and runs documented mass takedowns; genuine reviews tied to an untrusted reviewer can be unpublished collaterally. Not parked at the bottom of the menu, because "just for this launch" reappears every time velocity stalls.

  Equally deleted: reviews from your own team, and review swaps with other vendors - reviewer-attestation systems now catch both.

## Failure modes

- **Incentivizing reviews.** The ceiling is not a removed review - it is a terminated Partner account and a delisted app. Fix: menu 3's deletion; sanctioned triggers only.
- **Treating inferred ranking factors as confirmed.** The ≈80% subtitle-weight figure and its relatives are practitioner estimates; building a promise to leadership on them fails when the algorithm shifts. Fix: label every claim with its provenance tier and promise direction, not magnitude.
- **Quoting mobile-ASO numbers as B2B fact.** The +89% conversion lift from a ratings jump and the screenshot-scroll statistics are consumer app-store data. Structure transfers; numbers are analogy. Fix: the labelling discipline in [references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md).
- **Running a data-dark marketplace as a data-driven play.** Weeks of A/B intent with no funnel to read. Fix: question 2 first; on a data-dark platform, write once, drive external traffic, self-track installs.
- **Copying one marketplace's title convention to another.** Brand-first and keyword-first conventions both exist in production; the wrong one wastes the highest-weight field. Fix: check the target marketplace's own guidance before the rewrite.
- **Letting the listing decay.** Operators downgrade or flag stale listings, and update recency is itself a freshness signal on at least one platform. Fix: the 30-60 day refresh cycle with a named owner, from workflow stage 5.
- **Optimizing the listing while retention bleeds.** Uninstall rate feeds ranking negatively where corroborated, and at least two platforms monitor it independently of you. A better listing cannot outrun churn - fix the product experience gap first, or the new installs the listing wins become the uninstalls that sink it.

## Measurement

Plan gates first - self-set and flagged as such, not an external certification bar. Iterate until all five pass:

1. Every native analytics surface the platform grants is on, or the marketplace is explicitly classified data-dark with a self-built install ledger in place.
2. Every indexed field carries question 4's buyer language, in the title convention the target marketplace itself documents.
3. The first three gallery assets are unique, in-UI, and captioned, and every review ask in the plan maps to a sanctioned trigger.
4. Every ranking claim in the plan carries a provenance label (vendor-official, corroborated inference, single-source inference, analogy).
5. A refresh cycle of 30-60 days exists with a named owner.

Operating KPIs once live:

- Listing views and view-to-install conversion, per marketplace.
- Keyword coverage (the terms present in indexed fields), tracked separately from keyword performance (where the listing actually ranks for them) - coverage without rank is the signal to revisit the rewrite.
- Review velocity over the trailing 90 days, and average rating.
- Uninstall or churn rate.
- Badge progress against its certification criteria.

For A/B-style before/after comparisons, the analogy-sourced significance thresholds in [references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md) are a reasonable generic rule: treat lifts under 2% as noise.

If your harness has persistent memory, record per marketplace, so the next optimization cycle inherits it instead of re-asking:

- The data-rich/data-dark classification.
- The chosen ladder rung and what promoted it.
- The target keyword set with each term's current rank.
- The sanctioned review triggers in use.
- The refresh-cycle owner and date.

## Invocation examples

- "Our Salesforce AppExchange listing gets decent traffic but almost no installs - audit it and tell me what to fix first."
- "We launch on the Atlassian Marketplace in six weeks. Set up the listing plan: keywords, screenshots, video, and how we get our first twenty reviews without breaking the rules."
- "We've been on the Shopify App Store for a year and rankings are slipping. Build a standing program to defend our position."

## References

- [references/marketplace-lever-benchmarks.md](references/marketplace-lever-benchmarks.md) - the five-marketplace comparison: ranking factors split confirmed-vs-inferred, native analytics tooling, review-solicitation mechanics and enforcement maturity, the listing field constraints submitters optimize within, badge programs and their confirmed effects, and the practitioner ecosystem map.
- [references/aso-analogy-toolkit.md](references/aso-analogy-toolkit.md) - the mobile-ASO borrowings, quarantined and labelled: converged audit dimensions, brand-maturity tiering, the gallery-ordering framework, conversion benchmarks, and A/B significance thresholds - structure that transfers, numbers that do not.

See also, same collection:

- `samber/developer-platform-skills@app-marketplace-listing-standards` - the mirror-image sibling: the operator writing the listing rulebook. The rubric that skill designs is the constraint set this skill optimizes within; read the target marketplace's published version of it first.
- `samber/developer-platform-skills@app-marketplace-launch-marketing` - how operators design featuring, spotlight, and paid-placement programs; understanding that design is how a submitter earns featured placement rather than waiting for it.
- `samber/developer-platform-skills@integration-partnership-strategy` - which ecosystems deserve a listing investment at all; run it first when the marketplace portfolio itself is still open.
- `samber/developer-platform-skills@etl-connector-strategy` - the same submitter posture specialized to ETL/ELT platforms, where certification and connector maintenance dominate the listing work.
- `samber/developer-platform-skills@zapier-integration-playbook` - the Zapier-specific playbook: trigger design, review checks, beta, templates, embed and versioning on that one platform.