{"id":87673,"plugin_id":"plugin_asdk_app_6ac768a541148191adc9588132f9f44c","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-09T18:08:40.523Z","digest":"eec6577b5b1729c6b0e0df0db1256263f31aff6ce04ad88ed9751f8d61b38c6c","against":null,"payload":{"description":"Find where visitors leave a Captino funnel - visits and conversion rate per step, the biggest step-to-step drop, and per-question response counts of form steps. Use when the user asks \"where do people drop off\", \"which step loses visitors\", \"conversion per step\", \"why doesn't my form convert\", \"which question do people abandon\", or wants to improve a funnel's opt-in rate.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":467}],"name":"funnel-dropoff","skill_md_contents":"---\nname: funnel-dropoff\ndescription: Find where visitors leave a Captino funnel - visits and conversion rate per step, the biggest step-to-step drop, and per-question response counts of form steps. Use when the user asks \"where do people drop off\", \"which step loses visitors\", \"conversion per step\", \"why doesn't my form convert\", \"which question do people abandon\", or wants to improve a funnel's opt-in rate.\n---\n\n# Funnel drop-off\n\nDiagnoses where a funnel loses visitors, step by step and, for form steps,\nquestion by question.\n\n## Prerequisites\n\nUses the Captino MCP tools `list_funnels`, `list_funnel_steps`,\n`get_funnel_conversion_stats` and `get_form_step_performance`. Needs\n**Funnel: read** and **Performance: read** access, granted when connecting\nCaptino. If they are missing or fail with 401/403, run the `captino-setup`\nskill. Never invent numbers.\n\n## Workflow\n\n### 1. Resolve the funnel and its steps\n\n1. Find the funnel id with `list_funnels` (`query` = the name). Ask if the name\n   is ambiguous.\n2. Call `list_funnel_steps` with the funnel id. Note each step's `order`, `name`,\n   `type` (`Site` or `Form`), `isPublic`, `hasDraft` and `hasFormDraft`.\n   - Unpublished steps (`isPublic: false`) get no traffic; they are not\n     drop-off, they are gaps in the funnel. Say so.\n   - `hasDraft` / `hasFormDraft` means the live page differs from what the user\n     sees in the editor. Mention it when the user is puzzled by the numbers.\n\n### 2. Get per-step conversion\n\nCall `get_funnel_conversion_stats` with the funnel id. Pass `from` and `to`\n(ISO 8601 with offset) when the user names a period; omit both for all-time.\n\nEach item has `order`, `funnelStepName`, `visits` and `conversionRate`.\n`conversionRate` is the percentage of the **first step's** visits that reached\nthis step, not the rate from the previous step. Compute:\n\n- step-to-step retention = `visits[n] / visits[n-1]`\n- step drop = `visits[n-1] - visits[n]` (absolute) and `1 - retention` (relative)\n\nThe biggest problem is usually the largest **relative** drop between two\nconsecutive published steps, but show the absolute numbers too because a 50%\ndrop on 10 visitors matters less than a 20% drop on 5,000.\n\n### 3. Drill into form steps\n\nFor every step with `type: Form` that sits at or right after the biggest drop,\ncall `get_form_step_performance` with the step id and a period (`from` and `to`\nare required here; use the same period as step 2, or the last 30 days for\nall-time questions).\n\nThe result lists `questions` in order, each with `responses`, plus total\n`submissions`. Responses fall as visitors abandon the form, so the largest drop\nbetween consecutive questions is the question that costs the most submissions.\n`submissions` versus the first question's responses is the overall form\ncompletion rate.\n\n### 4. Report\n\nLead with the single biggest leak in one sentence, then one table of the steps:\n\n> Most visitors leave between the landing page and the quiz: only 31% continue.\n>\n> | # | Step | Visits | From first step | From previous step |\n> | --: | :-- | --: | --: | --: |\n> | 1 | Landing | 1,240 | 100% | - |\n> | 2 | Quiz (form) | 385 | 31% | 31% |\n> | 3 | Thank you | 290 | 23% | 75% |\n\nIf you drilled into a form, add a second small table of questions and their\nresponse counts and name the question where abandonment peaks.\n\nClose with two or three concrete suggestions tied to the numbers: shorten the\nform before question N, move the value proposition above the fold on step 1,\npublish the missing step, and so on. Keep them specific to what the data shows.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}