Conversions Dropped to Zero

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.

Conversions for [client: if not listed, call roasr_list_clients and let me pick] dropped almost to zero. Find out when, whether it is a tracking break or a real drop, who is affected, and what I should check by hand first.

1. **Load the RoASr method.** Call roasr_load_skill('roasr_frame') and roasr_load_skill('root_cause') and follow them 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. Look at [window: last 14 days (default) | last 7 days | last 30 days | custom], plus the same-length period before it for comparison. If a window starts before dataSince, say so and narrow it. An account with sync failed or never may simply be stale: name it and make no conclusion about it, since a stale feed looks exactly like zero conversions.
3. **What dropped.** Focus on [what dropped: all conversions (default) | purchases | leads | custom]. roasr_metrics_daily per account (accountId): conversions, clicks, spend and impressions.
4. **Find the date.** The first day conversions fell below about a fifth of the previous 7-day median and stayed there. Say how sure you are of the date, and that a one-day gap alone is not a break.
5. **Break or real drop?** Compare clicks, spend and impressions around that date. Clicks and spend held while conversions went to zero: probable tracking or site break. Clicks, spend and conversions fell together: delivery problem (budget exhausted, billing, rejected ads, a paused campaign), and RoASr cannot see those causes directly, so name them as things to verify. Clicks held, conversions fell partially: possible landing or checkout issue.
6. **Which actions.** roasr_metrics_actions before and after the date. Every bucket vanishing together points at the site, tag or consent layer. One bucket vanishing points at that event's setup. Buckets are not additive: compare each on its own.
7. **Google specifics.** For Google accounts, roasr_signals with kind google_conversion_actions: did a conversion action change status, lose its category, stop being primary, or show no recent conversions?
8. **Scope.** Is it one account, one platform or all? One account suggests its own setup; one platform suggests that platform's tag or server-side events; everything at once suggests the site, checkout, payment or a consent banner.
9. **Hand-check list.** Rank where to look first, with the date from step 4: did the site or checkout deploy that day, does a test purchase or form fire the tag or pixel, did the consent banner change, is the server-side event flow up, was a conversion action or attribution setting edited, does the payment or form backend work.

**Output:** open with one sentence: the date, the most likely class of cause (tracking, site, delivery or real drop) and your confidence (high / medium / low). Then a table: account · platform · date of drop · clicks before and after · conversions before and after · verdict. Then 3–6 hand checks grouped High / Medium / Nice-to-have, each with what a positive finding would mean.

**Output rules:**
- Recommendations only — never apply changes; RoASr tools are read-only.
- Do not tune bids, budgets or creatives until tracking is ruled out.
- A cause fits only if its timing matches the date; timing alone is correlation, so label unconfirmed causes as unconfirmed.
- With no margin set for the client, talk ROAS and CPA, not profit.
- Answer in the language of this prompt.