Conversion Mix & Funnel

Before you start, ask me about each [input] below — one question at a time, with ready-made answers to pick from; skip whatever you can already tell.

Break down what [client: if not listed, call roasr_list_clients and let me pick] is really counting as conversions, how the funnel looks, and whether the conversion numbers can be trusted.

1. **Load the RoASr method.** Call roasr_load_skill('roasr_frame') and follow it exactly.
2. **Frame the data.** Call roasr_limits, roasr_list_clients and roasr_list_accounts: plan, niche, margin, KPI, account currencies, sync status and dataSince. Window: [window: last 30 days (default) | last 14 days | last 90 days | custom]. Business type: [business type: detect from the data (default) | ecommerce | lead generation | custom]. If the window starts before dataSince, say so and narrow it. If an account has sync.state failed or never, name it and draw no conclusions from it. If currency.mixed is true, never sum money across accounts: do the work per account with accountId.
3. **Pull the numbers.** roasr_metrics_actions for the window: both raw (every action type with its value) and buckets (purchase, lead, add_to_cart, complete_registration, landing_page_view, engagement and others). Then roasr_metrics_summary for the same window: total conversions, clicks, spend, revenue.
4. **Respect what buckets are.** Buckets are not additive: engagement is an aggregate and reactions, comments and shares are its components, so summing them double-counts. Never add a funnel step to another or total the buckets. Treat engagement as context, not as a funnel step.
5. **Fork by business type.** Decide from the buckets, not from the client's description. Purchases, add_to_cart and revenue present: ecommerce, with the funnel landing_page_view, add_to_cart, purchase. Leads or registrations with no purchases: lead generation, with the funnel landing_page_view, lead or complete_registration. If both show up, name the main goal and say what the other bucket is doing there. Without a clear fork, say so and ask me.
6. **Measure the funnel.** Compute the step-to-step rates that make sense: clicks to landing_page_view (a rate under about 60-70% points to a slow page or a tracking gap), landing_page_view to the next step, the next step to the final one. Give each rate with its numerator and denominator. Do not benchmark against invented industry norms; say that no norm is available here.
7. **Hunt for broken or inflated goals.** Look in raw for (a) two action types with almost the same value, likely one event counted under two names, never to be summed; (b) cheap micro-actions, such as page or video views, mixed into total conversions: if summary conversions are much larger than the main outcome bucket, the account is probably optimizing toward a proxy goal; (c) purchases with zero revenue in the summary: the order value is not being passed, so ROAS is meaningless; (d) a main bucket that is missing or flat at zero while clicks are healthy: a tracking break comes before any performance story.
8. **Conclude.** State which bucket is the real main conversion, which counted goals are noise, and what to verify in the tracking setup.

**Output:** open with one sentence: what the account's real main conversion is and whether the totals can be trusted. Then a table: bucket · count · share of its funnel step · note (real goal / proxy / possible duplicate / context only). Then the funnel line with rates. Then 3-8 actions grouped High / Medium / Nice-to-have, each with the expected effect. Fixing tracking comes first.

**Output rules:**
- Recommendations only — never apply changes; RoASr tools are read-only.
- Buckets are not additive: never sum them and never call the sum "total conversions".
- Every rate shows its numerator and denominator; "possible duplicate" is a suspicion until the setup is checked.
- If the client has no margin set, talk ROAS and CPA, not profit.
- Answer in the language of this prompt.