Analyze Attribution with an AI coding tool
The adapty-attribution skill lets an AI coding tool answer questions about your ad campaigns from your Adapty Attribution data. Ask in plain language — “Which channel paid back best last month?”, “Should we kill this campaign?” — and it runs the adapty attribution commands, reads the results, and answers with the numbers behind the verdict.
A skill is a set of instructions your AI tool loads on demand. This one contains the parts of the Attribution reports that are easy to misread: which revenue basis a ROAS is on, which values can’t be computed yet, and where Apple Search Ads spend comes from.
The skill only reads. It changes no campaign, tracking link, or integration, and the budget moves it suggests are yours to make.
What it answers
The skill works from playbooks — a set of commands, a way to read them, and the shape of the answer — for the questions teams ask most:
- Payback and budget: Ranks channels and campaigns by ROAS and says where extra budget should go, and in what steps.
- Young campaigns: Judges a new campaign at a day its cohorts have reached, instead of on a D30 that has only a week of revenue in it.
- Trial quality: Ranks campaigns by cost per paying trial rather than cost per install, so campaigns whose trials don’t convert show up.
- Weekly digest: Compares the last week with the ones before it and flags what moved.
- Countries and stores: Shows where to expand and where to cap spend.
- Winners inside a campaign: Ranks ad sets and ads within one campaign.
- Low-quality traffic: Flags campaigns with more refunds and chargebacks than the app’s average.
- Predicted payback: Reads predicted ROAS at longer horizons and turns it into a break-even CPI.
- Numbers that disagree: Explains why Meta, TikTok, or Google report a different ROAS than Adapty.
- Installs dropped: Reads the pattern of a drop to tell an app-side cause from a single network’s.
Setup questions — connecting Meta, TikTok, or Google Ads, tracking links, the web pixel, Stripe, or Paddle — go to the docs pages that own them.
Apple Search Ads
Attribution doesn’t collect Apple Search Ads spend, so its ROAS is empty there. The skill gets Apple Search Ads spend and ROAS from Ads Manager and reports them as their own line, in your campaign group’s currency, instead of merging them into the Attribution numbers. If you give it a spend amount yourself, it answers with a threshold — “Apple Search Ads beats Google if its spend was under $X” — rather than a ROAS built from your estimate.
Install the skill
The skill ships in the adapty-skills repository alongside adapty-integration, ads-manager, flow-generator, and others. Each of these commands installs every skill in the repository.
The skill needs the Adapty CLI, logged in, and Attribution access for your company. Then ask for it — /adapty-attribution in tools that map slash commands to skills, or “Use the adapty-attribution skill” in the ones that don’t.
Example prompts
Name the app, the period, and the decision you’re making. The skill works out the commands.
Our app is app_finance. Which paid channels paid back best in August, and where
should next month's extra $20k of budget go? Judge on revenue after the store's
cut.Our TikTok "Fall Promo" campaign started on Sep 15. Leadership's rule is to kill
anything under 80% D30 ROAS. How is it doing against "Spring Sleep", and should
we kill it?Meta Ads Manager says our "Broad US" campaign had a 2.4 purchase ROAS in August.
Adapty shows something much lower. Which number is right, and what do I tell my
boss?What’s next
- Query Attribution from the CLI — the same reports by hand, without an AI tool.
- Attribution analytics — the dashboard view of the same data.