# Fullstory Conversion Investigation > Investigate why conversion rate is down — or answer any question about checkout/booking funnel performance — using Fullstory data. Loads org context, identifies the right funnel, finds what's worsening among dropouts, pulls session evidence, and delivers a plain-English narrative a trading manager or COO can act on. Use this skill whenever someone asks "why is conversion down?", "why are users abandoning checkout?", "what's causing our funnel drop-off?", "our bookings are down this week", "conversion has dropped", or any question about checkout or purchase funnel performance. Also trigger when someone asks what's changed in the user journey or why revenue metrics have shifted. - Skill: `aaronbali/fullstory-conversion-investigation` (Agent Skill) - Install (CLI): `npx skillmds@latest add aaronbali/fullstory-conversion-investigation` - Raw SKILL.md: https://api.skillmd.com/api/skills/aaronbali/fullstory-conversion-investigation/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Finance & Business - Author: aaronbali (https://skillmd.com/u/aaronbali) - Updated: 2026-09-22 - Page: https://skillmd.com/skills/aaronbali/fullstory-conversion-investigation --- # Fullstory Conversion Investigation The goal is a credible, evidence-backed answer to "why is conversion down?" — specific enough to act on, written in plain English that doesn't require Fullstory expertise to understand. Speed matters: get to insight fast, don't over-engineer. ## Step 1 — Load org context Look for the org context file at `~/claude/accounts/{slug}-fullstory-context.md`. - If the user specified an org name or slug, load that file directly. - If not, check `~/claude/accounts/` for available context files and use the most relevant one. If multiple exist, ask which org. - If no context file exists, tell the user: "I don't have org context for this yet — run `/fullstory-org-discovery` first to set it up, then I can do the analysis." Read the context file fully before proceeding. You need: funnel IDs, the active Revenue Event ID, the confirmed active checkout page IDs, and the Notes section flags. ## Step 2 — Identify the conversion funnel From the Funnels section of the context file, identify the primary end-to-end conversion funnel. Prefer: 1. The funnel explicitly described as end-to-end (e.g. "Car Booking Full Funnel", "Purchase Funnel", "Checkout From Search") 2. Any funnel that ends at the success/confirmation page Note its `funnel_id` — you'll use this in the next step. If multiple plausible funnels exist, pick the broadest one unless the user's question implies a specific step (e.g. "why is checkout abandonment up?" → use the checkout-specific funnel). ## Step 3 — Find what's worsening Run these two calls in parallel: 1. `discover_groups(funnel_id=, compare_to_previous=true, relative_time_range="7d", limit=5)` — Top frustration/error groups among funnel dropouts in the last 7 days, ranked by regression (biggest new or worsening issues first) 2. `get_opportunities(funnel_id=, time_range="30d")` — StoryAI's pre-ranked issues specifically impacting this funnel The combination gives you both fresh regressions (discover_groups) and the broader ranked picture (StoryAI). ## Step 4 — Pull session evidence From the top 2–3 issues identified in Step 3, get sessions: `get_sessions_for_opportunity(metric_id=, group_id=, relative_time_range="7d", limit=3)` Then for 1–2 sessions per issue, call `get_session_events(device_id=..., session_id=...)` to read what actually happened. Look for: the moment the issue occurs, what the user did immediately before and after, whether they abandoned or found a workaround. ## Step 5 — Synthesise the answer Write a plain-English response structured as: --- **Conversion is [up/down/flat] this week.** [One sentence on the headline — percentage or directional if available from opportunity stats.] **The main drivers:** **1. [Issue title]** — [N] users affected [2–3 sentences: what's happening, where in the funnel, what the session evidence shows. Avoid jargon. Say "users hitting an error when trying to pay" not "UncaughtException on the payment step".] **2. [Issue title]** — [N] users affected [Same structure.] **3. [Issue title if relevant]** **What we'd recommend looking at next:** [1–2 actionable bullets. Keep them specific — "Investigate the /api/collect 503 error on the cart page" not "investigate errors".] Session replays: - [session_url] — [one line description of what this session shows] - [session_url] — [one line description] --- ## What to watch out for - **Don't use the duplicate Revenue Event** — the context file will flag which one is active. Using the wrong one gives false revenue numbers. - **Check the Notes section** for implementation flags before interpreting data. If the context file flags a URL migration or noisy events (e.g. Usetiful errors), exclude those from the narrative. - **StoryAI opportunities show user counts, not conversion rate** — frame them as "X users experienced this issue" not "conversion dropped by X%". Only make percentage claims if `get_opportunity_stats` gives you rate data. - **If discover_groups returns no groups**, the funnel may have low traffic this week or the funnel IDs may reference old page paths. Check the Notes section for URL migration flags and try `get_opportunities` scoped to the funnel as a fallback.