audit-conversion-tracking

# Audit Conversion Tracking This skill produces a structured conversion-tracking audit. Output is a prioritised fix list with severity, root cause, impact, fix, and verification per issue. The skill is opinionated about a few things: every brand should triangulate against source-of-truth (Shopify / payment processor / CRM); attribution model choice is a documented decision (not a default to ignore); Consent Mode v2 is non-negotiable in EU/UK; cluster-attribution capability is increasingly necessary for SEO ROI reporting; conversion tracking drifts and quarterly re-audit is the safety valve. ## When to use this skill - The user is preparing to report SEO revenue and wants confidence in the data. - The user has noticed GA4-vs-source-of-truth discrepancies. - The user is launching a major SEO investment and wants the conversion measurement clean upfront. - The user is running a quarterly analytics review. ## When NOT to use this skill - The user wants to set up conversion tracking from scratch (point to GA4 official docs; this is an audit). - The user wants to audit GA4 broadly — use `audit-ga4-setup` (overlaps; conversion-tracking is the deeper slice). - The user wants to analyse conversion data — use `analyse-traffic-patterns`. - The user wants the SEO KPI dashboard — use `define-seo-kpis`. ## Inputs required 1. **Business context and SEO foundation** — read `businesses/<slug>/business_context.md` and, for SEO-specific work, `businesses/<slug>/seo_foundation.md`. Resolve `<slug>` by listing `businesses/`: if exactly one folder exists, use it; if multiple, ask which business; if none, recommend running `business-context` first. If SEO foundation is missing, recommend running `seo-foundation` before producing output. Critical sections: products/services (3) for goal-event identification, goals (7). 2. **GA4 access** — required. Need to inspect events, conversions, attribution settings. 3. **Source-of-truth data source** — required for triangulation. Shopify / WooCommerce / payment processor / CRM. 30-day window of conversions and revenue. 4. **Tag manager / data layer access** — preferred. GTM container (Web + Server-side if applicable). 5. **Geographic operations** — required. EU / UK presence triggers Consent Mode v2 review. 6. **Reporting destinations** — list of dashboards / reports referencing conversion data (Looker Studio, internal reports). Audit needs to flag attribution-model mismatches. 7. **Cluster map** — read `businesses/<slug>/clusters/cluster-map.md` if available. Cluster-attribution capability is part of the audit. If GA4 access or source-of-truth data is missing, ask. Audit can't proceed without both. ## Process 1. **Goal-event identification.** Confirm which events should be conversions for this business model. - Ecommerce: `purchase`. - Subscription / SaaS: `sign_up`, `start_trial`, `subscribe`. - B2B / services: `generate_lead`, `book_meeting`, `submit_form`. - Content / media: depends on monetisation model. - Verify only goal events are marked as conversions (funnel events should NOT be conversions). 2. **Event implementation.** For each goal event: - Event firing correctly (verify in DebugView / GTM Preview). - All required parameters populated. - Event firing once per real conversion (no duplicate fires). - Cross-domain handled if applicable. 3. **Ecommerce parameter audit.** For ecommerce specifically: - `purchase` event has `transaction_id`, `value`, `currency`, `items` array. - `items` array populated (item_id, item_name, item_category, price, quantity). - All 5 funnel events firing in order (`view_item_list`, `view_item`, `add_to_cart`, `begin_checkout`, `purchase`). - Refunds tracked if applicable. 4. **Attribution model.** - Reporting attribution model identified (DDA, last-click, first-click, linear, position-based, time-decay). - Document the decision: which model, why, where reported. - Cross-check downstream reports (Looker Studio, internal): do they all use the documented model? Flag mismatches. - For SEO specifically: compute organic-search revenue under DDA and last-click; compare; document the gap. 5. **Consent Mode v2.** - EU/UK presence assessed (% of traffic). - Consent Mode v2 implementation status: yes / partial / no. - Test: verify GA4 receives `consent_default` and updates correctly. - If EU traffic >5% and Consent Mode v2 missing: flag as P0 (legal + measurement issue). 6. **Source-of-truth triangulation.** - 30-day window: conversions + revenue from GA4 vs Shopify / payment processor / CRM. - Acceptable: ≤5% discrepancy. - 5-10%: investigate but acceptable. - 10%: likely measurement issue; flag as P0 / P1 depending on cause. 7. **Cluster-attribution capability.** - Is there a way to attribute conversions to the cluster that brought the visitor in? - Options: custom dimension at session start; landing-page-to-cluster mapping in BigQuery; entry-page custom dimension. - If none: flag as P1 (limits SEO ROI reporting capability). 8. **Server-side tracking (large brands).** - GTM Server-side container in place? - Flag as P2 for small brands; P1 for ecom brands at scale (>$5m/yr). 9. **Cross-device / cross-session.** - User-ID setup if available (CRM-side ID linked to GA4). - Cookie-based stitching working. 10. **Conversion value population.** - `value` parameter populated on `purchase` and other goal events where applicable. - Currency consistent. 11. **Permissions hygiene.** - GA4 / GTM Editor access reviewed; remove stale. 12. **Write the audit.** ## Output format ```markdown # Conversion Tracking Audit — [Business name] **Audit date:** [date] **GA4 property:** [ID] **Source-of-truth:** [Shopify / Stripe / CRM / etc.] **Auditor:** [name] ## Executive summary - **P0 issues:** [n] - **P1 issues:** [n] - **P2 issues:** [n] - **Triangulation:** GA4 [$n] vs source-of-truth [$n] → [n%] discrepancy - **Estimated total fix effort:** [n] days - **Headline:** [1-2 sentence summary] ## Findings ### P0 — Critical (data is wrong; reporting unreliable) #### [Issue title] - **Symptom:** [user-visible] - **Root cause:** [technical reason] - **Impact:** [reports / decisions affected] - **Fix:** [steps] - **Effort:** [hours / days] - **Verification:** [how to confirm] [Repeat] ### P1 — Material [Same structure] ### P2 — Hygiene [Same structure] ## Configuration summary ### Goal events | Event | Firing? | Marked as conversion? | Parameters complete? | |-------|---------|------------------------|----------------------| | purchase | yes | yes | partial (`items` missing) | | add_to_cart | yes | **incorrectly marked as conversion** | yes | | ... | ... | ... | ... | ### Ecommerce parameter audit | Event | items | transaction_id | value | currency | |-------|-------|---------------|-------|----------| | purchase | partial (~85%) | yes | yes | yes | | view_item | yes | n/a | yes | yes | | add_to_cart | yes | n/a | yes | yes | | begin_checkout | yes | n/a | yes | yes | | view_item_list | **missing** | n/a | yes | yes | ### Attribution - **Reporting model in GA4:** DDA - **Reporting model in Looker Studio reports:** last-click - **30-day organic revenue (DDA):** [$n] - **30-day organic revenue (last-click):** [$n] - **Gap:** [n%] - **Flag:** **mismatch** — Looker Studio reports under-credit organic by [n%] ### Consent Mode v2 - **EU/UK traffic share:** [%] - **Implementation:** [yes / partial / no] - **Estimated under-counted EU conversions:** [%] ### Source-of-truth triangulation | Metric | GA4 (30d) | [Source-of-truth] | Discrepancy | Verdict | |--------|-----------|-------------------|-------------|---------| | Conversions | n | n | n% | Pass / Investigate | | Revenue | $n | $n | n% | Pass / Investigate | | AOV | $n | $n | n% | Pass / Investigate | ### Cluster attribution - **Capability:** [yes (mechanism) / no] - **Recommendation:** [implementation path if missing] ### Server-side tracking - **Status:** [yes / no / N/A] - **Estimated signal recovery if implemented:** [n%] ### Cross-device / user-ID - **User-ID:** [yes / no] - **Cross-device stitching:** [working / partial / unavailable] ### Permissions - **Stale users:** [n] ## Prioritised fix list (one-page version) | # | Issue | Severity | Effort | Owner | |---|-------|----------|--------|-------| | 1 | `items` missing on ~15% of purchases | P0 | 3-5 days | Dev | | 2 | `add_to_cart` marked as conversion | P0 | 30 min | Marketing | | 3 | Consent Mode v2 partial | P0 | 1-2 days | Dev | | 4 | Looker Studio reports use last-click vs GA4 DDA | P1 | 4 hrs | Marketing | | 5 | No cluster attribution capability | P1 | 2-3 days | Analytics | | ... | ... | ... | ... | ... | ## Recommendations beyond fixes - [ ] Re-audit in 90 days. - [ ] Standardise attribution model across all reports. - [ ] Document the choice in the per-business AGENTS.md. - [ ] Set up cluster-attribution before next strategic review. - [ ] Schedule quarterly conversion-tracking re-audit. ## Open questions for the user - [ ] Confirm preferred attribution model (DDA recommended for most). - [ ] Confirm budget for server-side tracking (if applicable). - [ ] Confirm cluster-attribution implementation owner. - [ ] Confirm Consent Mode v2 implementation owner. ``` Save the produced file to `businesses/<slug>/analytics/conversion-tracking-audit-[YYYY-MM-DD].md`. After writing, tell the user the file path so they can open it. ## Quality bar - Every finding has severity, root cause, impact, fix, effort, and verification. - P0 / P1 / P2 buckets used consistently (P0 = data wrong; P1 = data incomplete; P2 = hygiene). - Triangulation done against source-of-truth on conversions, revenue, AOV. - Attribution-model audit covers GA4 default + downstream reports; mismatch flagged. - Consent Mode v2 status checked if EU/UK traffic >5%; flagged P0 if missing. - Cluster-attribution capability assessed. - Server-side tracking flagged as recommendation for ecom brands at scale. - Output actionable for both technical and marketing owners. ## Common mistakes to avoid - Don't trust GA4 numbers without source-of-truth triangulation. The discrepancy is the signal. - Don't ignore last-click vs DDA mismatches. They materially change the SEO revenue story. - Don't propose server-side tracking for tiny brands. Cost > benefit below ~$1m/yr ecom. - Don't skip Consent Mode v2 in EU/UK. Legal + measurement. - Don't over-mark conversions. Funnel events should NOT be conversions. - Don't audit without checking attribution alignment in downstream reports. Single source of truth fails when reports disagree. - Don't neglect cluster-attribution. SEO ROI reporting at the cluster level is increasingly the standard. - Don't audit and walk away. Schedule re-audit in 90 days. ## Example **Input (abbreviated):** Field & Sun. GA4 + Shopify + Looker Studio. EU + US operations. **Output (abbreviated):** P0 (3): - `items` missing on ~15% of purchases (custom checkout flow); 3-5 days dev. - `add_to_cart` marked as conversion; 30 min unmark. - Consent Mode v2 partial; 1-2 days dev. P1 (3): - Looker Studio uses last-click while GA4 default is DDA; $124k vs $147k organic; 4 hrs to align. - No cluster attribution; 2-3 days analytics work to add via BigQuery mapping. - `view_item_list` event missing in Shopify-GA4 integration. P2 (2): - 3 ex-contractors retain GA4 Editor access. - Server-side tracking absent; recommended for current scale; 4-6 weeks dev. Triangulation: GA4 30-day revenue $124k vs Shopify $146k → 15% gap. Mostly explained by (a) ~3% under-count from missing `items` on ~15% of orders; (b) ~10% under-count from Consent Mode partial implementation; (c) ~2% other (cross-device). Attribution: GA4 30-day DDA-organic $147k vs last-click $124k. Looker Studio reports last-click; team had been reporting "$124k organic" — should be $147k under DDA. Cluster attribution: not currently possible. Recommend BigQuery-side landing-page → cluster mapping (2-3 days analytics). Server-side: Field & Sun at $5-10m/yr is at the threshold; recommend P2 / next-quarter scoping. Total fix effort: 6-7 days dev + 1 day marketing config + 2-3 days analytics. Re-audit scheduled: 2026-08-01. Saved to `businesses/field-and-sun/analytics/conversion-tracking-audit-2026-05-01.md`.