agentsclimarketplace

Ratings mechanics

Skill rshankras/claude-code-apple-skills/skills/app-store/ratings-mechanics

Claude Code skills for Apple platform development (iOS, macOS, iPadOS) — product validation, code generation, App Store optimization, and more

Install
npx -y skills add rshankras/claude-code-apple-skills --skill ratings-mechanics

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

How App Store ratings actually behave — per-storefront isolation (your US stars show nowhere else), the never-reset rule, phased release + manual release as rating protection, and where prompting/replying fit. Use when planning ratings strategy for new markets, considering a ratings reset, setting release options, or diagnosing "why is my rating missing in country X."

SKILL.md

5.3 KB, as published. Nobody here has run it

Ratings Mechanics

The rating is an asset with mechanics most developers learn the hard way. This skill covers the four rules that aren't obvious from the ASC UI. Prompting code lives in generators/review-prompt; reply writing lives in app-store/review-response-writer — this skill is the strategy layer that tells you when each matters.

When This Skill Activates

  • Localizing or expanding into new storefronts ("why does my app show no rating in Japan?")
  • Considering the "reset ratings summary" option on a version release
  • Choosing release options before submitting (phased vs immediate, auto vs manual)
  • Planning a per-market ratings strategy alongside product/localization-strategy
  • A bad build or review-bomb is threatening the rating

Rule 1: Ratings are per-storefront — they do not travel

Your 4.8★ from 2,000 US ratings renders as no rating at all on the Japanese storefront until Japanese users rate the app there. Every storefront starts from zero.

Consequences:

  • ✅ Entering a new market = re-running the early-days ratings playbook in that market: prompt eagerly (within guidelines), localize the prompt moment, reply to every early review.
  • ✅ Weight requestReview triggers by storefront maturity — a market with 12 ratings needs the prompt more than the home market with 5,000.
  • ❌ Assuming social proof transfers with the binary. A localized listing with zero local ratings converts like an unknown app, because there it is one.
  • The written-review pool is also per-storefront: expect empty review sections in fresh markets and seed them via TestFlight communities or launch outreach in that region.

Rule 2: Never reset the ratings summary

ASC offers a reset when you release a new version. It is almost always a mistake:

  • Reset discards the count as well as the average — 4.2★ from 3,000 ratings converts better than a naked 5.0★ from 6, and the count never comes back except one rating at a time.
  • The instinct to reset ("v2 is a big rewrite, old reviews don't apply") is better served by replying to outdated negative reviews (updated ratings replace the old score — see review-response-writer) and by the What's New copy.
  • ✅ Legitimate near-exception: a catastrophic launch (sub-3★, low count, fixed root cause) on an app with almost no ratings mass. Even then, run the math on count loss first.
  • ❌ Resetting an established app to chase a higher average. You'll rank and convert worse for months.

Rule 3: Phased release + manual release are rating armor

Two ASC toggles turn a bad build from a rating catastrophe into a contained incident:

  • Manual version release — approval ≠ release. Release when you're awake and watching crash dashboards, not whenever review finishes.
  • Phased release — 7-day staged rollout (1% → 2% → 5% → 10% → 20% → 50% → 100%) to users with automatic updates. A crashing build caught on day 1–2 has burned ~3% of your users; pause the rollout, fix, resubmit. Without it, 100% of users get the bad build and the 1-star flood arrives before the hotfix does.
  • ✅ Default for every release: phased ON + manual release + monitor day-1 crash rate.
  • ❌ Halting a phased release as a rollback. Pausing stops new deliveries only — users who got the build keep it, and manual App Store downloads always get the new version. The armor is damage limitation; the fix still has to ship.

Rule 4: The prompt-and-reply loop is the only rating input you control

  • Prompt at success moments with requestReview — never at launch, never mid-task. Aim the moment, cap the frequency, and localize what "success" means per market. Implementation: generators/review-prompt.
  • Reply to negative reviews — when a user updates their review after a reply, the new score replaces the old one in the average. Replies are the only mechanism that converts existing 1-stars into 4-stars. Templates: app-store/review-response-writer.
  • Per-storefront corollary: track "unanswered negative reviews" per market, not globally — a 5-review storefront is one bad review away from 60% negative.

Output Format

When auditing an app's ratings posture, report per storefront:

Storefront | Rating (count) | Unanswered 1–3★ (90d) | Prompt localized? | Phased+manual habit?

flagging: fresh storefronts with no prompting plan, any reset consideration (🔴 stop), and releases going out unphased.

References

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.