rshankras/claude-code-apple-skills/ad-attribution

Claude Code and agent skill by rshankras/claude-code-apple-skills

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Privacy-preserving ad measurement with AdAttributionKit (SKAdNetwork's successor) — install and re-engagement attribution, conversion-value strategy under crowd anonymity, and end-to-end postback testing. Use when running paid acquisition beyond Apple Ads, measuring re-engagement campaigns, designing conversion values, or migrating from SKAdNetwork.

Install

npx vskill@latest install rshankras/claude-code-apple-skills/ad-attribution

SKILL.md

Ad Attribution (AdAttributionKit)

You can't optimize paid traffic you can't measure. AdAttributionKit is Apple's privacy-preserving attribution framework — interoperable with SKAdNetwork, and the go-forward path (it also works in alternative marketplaces). This skill covers the advertised-app side: what an indie running paid campaigns must implement and how to design conversion signals that survive privacy thresholds.

Note the seam: Apple Ads attribution uses the AdServices framework, not AdAttributionKit — campaigns there are measured in the Apple Ads console (see app-store/apple-search-ads). This skill is for every other paid channel (ad networks, social, publishers).

When This Skill Activates

  • Running or planning paid user acquisition outside Apple Ads
  • "How do I know which campaign drove installs?" / measuring re-engagement ads
  • Designing or debugging conversion values / postbacks
  • Migrating from SKAdNetwork (fully interoperable — registered networks need no re-enrollment)

How it works (the 60-second model)

Ad network signs an ad (compact JWS impression) → publisher app shows it → user installs (or re-engages) → after a privacy-delayed measurement window, a postback with your conversion info goes to the ad network (and optionally to you). Attribution requires no user tracking; the price is coarser data, governed by crowd anonymity.

Conversion-value strategy (the part that's actually yours)

  • Two granularities: fine value 0–63 and coarse low/medium/high. Low crowd sizes

downgrade fine → coarse and can trim the 4-digit campaign source-identifier to as few as its first 2 digits — so encode what matters most (campaign family, geo tier) in the first two digits, detail in the rest.

  • Map values to revenue-predictive early behavior (completed onboarding, trial start, first

purchase), not vanity events. 64 slots go fast; reserve ranges (e.g. 0–15 activation, 16–47 monetization ladder, 48–63 reserved).

  • Update with updateConversionValue(_:) / PostbackUpdate; set lockPostback when the value

is final to schedule transmission (still privacy-delayed, never instant).

  • Re-engagement postbacks are separate from install postbacks: update them independently via

conversionTypes (.install / .reengagement); the first re-engagement update must land within 48 hours.

  • Multiple simultaneous re-engagement campaigns need conversion tags (iOS 18.4+): read the

tag from the re-engagement URL's query items, store it, and pass it in PostbackUpdate — without a tag, updates hit only the most recent conversion.

Publisher-side mechanics (if your app also shows ads)

  • Click-through: UIEventAttributionView over the tappable ad + appImpression.handleTap()

within 15 minutes of creating the impression.

  • View-through: beginView()/endView() on the same instance; ≥2 seconds on screen to

count; one open view-through per network per advertised app.

  • SKOverlay / SKStoreProductViewController: attach the impression — they count view on

present, click on tap.

Attribution windows & cooldowns (iOS 18.4+ configurability)

  • Defaults: 30 days click-through, 1 day view-through (install ads).
  • Override per ad network and interaction type in Info.plist

(AdAttributionKitConfigurations → AttributionWindows); you can ignore "view" or "click" (not both) per network.

  • Set cooldowns (install-cooldown-hours, reengagement-cooldown-hours) so a

re-engagement tap minutes after an install isn't double-attributed.

  • Country-level postback data (iOS 18.4): storefront country at install, gated behind an extra

crowd-anonymity tier — a bonus signal when volume allows, never guaranteed.

  • Overlapping re-engagement conversions require opting in:

EligibleForAdAttributionKitOverlappingConversions = YES in Info.plist.

Testing — never ship blind

  • Developer Mode → Settings → Developer → Ad Attribution Testing: removes time

randomization, shortens all conversion windows, sends postbacks on demand — full end-to-end rehearsal in minutes instead of days.

  • Point development postbacks at a separate dev endpoint, never production: they're signed

with a different key (kid), the network ID is always development.adattributionkit, and the advertised item ID may be 0 from Xcode.

  • ✅ Verify the postback pipeline before spending a dollar. ❌ Debugging attribution after

a live campaign burns budget on unmeasurable installs.

Checklist

  • [ ] Conversion-value map documented (0–63 + coarse fallback + first-2-digit priority)
  • [ ] lockPostback fired once value is final; re-engagement first update < 48h
  • [ ] Conversion tags wired if >1 re-engagement campaign runs at once
  • [ ] Cooldowns configured; windows tuned per network where defaults don't fit
  • [ ] Dev-endpoint postback rehearsal done in Developer Mode
  • [ ] Campaign readout joined with growth/analytics-interpretation (source-type funnel) —

attribution tells you which ad, App Analytics tells you what they did after

References

  • https://developer.apple.com/documentation/adattributionkit
  • https://developer.apple.com/videos/play/wwdc2024/10060/ (Meet AdAttributionKit)
  • https://developer.apple.com/videos/play/wwdc2025/221/ (What's new in AdAttributionKit)
  • Related skills: app-store/apple-search-ads (Apple Ads uses AdServices, not this), growth/analytics-interpretation, growth/store-growth-audit (P7.3 halo tracking)

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
770
Repository forks
74
7-day trend score
33

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/rshankras/claude-code-apple-skills/ad-attribution/badge)](https://verified-skill.com/skills/rshankras/claude-code-apple-skills/ad-attribution)

Something wrong with this skill or its page? Tell us.