analyse-traffic-patterns

# Analyse Traffic Patterns This skill produces a structured quarterly traffic analysis. Output is an insight-rich report — channel mix, page-type rollup, cluster aggregation, segmentation, anomaly investigation, prioritised actions. The skill is opinionated about a few things: QoQ + YoY is the default comparison (28-day windows are too noisy); cluster-level rollup is where strategic insight lives (page-level alone is too granular); every meaningful pattern shift gets a "what changed" investigation; analysis without prioritised actions isn't analysis, it's reporting. ## When to use this skill - The user wants a quarterly traffic analysis for stakeholders. - The user is investigating an apparent surge or decline. - The user is briefing strategy on which channels / pages / clusters are working. - The user is preparing a board / executive update on marketing performance. - The user wants to baseline traffic before launching a major initiative. ## When NOT to use this skill - The user wants a monthly metrics dashboard — use `define-seo-kpis` to build the dashboard; this skill is heavier and quarterly. - The user wants to audit GA4 / GSC config — use `audit-ga4-setup` / `audit-search-console-setup`. - The user wants to analyse search-specific (query / position) data — use `analyse-search-performance`. - The user wants to identify refresh candidates — use `plan-content-refresh`; this skill surfaces decay but the prioritised refresh list lives there. ## 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), goals (7). 2. **Cluster map** — read `businesses/<slug>/clusters/cluster-map.md`. Pages → clusters mapping required. 3. **GA4 access (or exports)** — required. Sessions, users, engaged sessions, conversions, revenue. Last 90 days + prior 90 days minimum. 4. **GSC access (or exports)** — strongly preferred. Cross-reference impressions / clicks with sessions. 5. **Source-of-truth revenue** — preferred. Shopify / CRM / payment processor for triangulation. 6. **Comparison window** — required. Default: last 90 days vs prior 90 days. Add YoY (last 90 days vs same 90 days last year) if retention covers it. 7. **Recent changes** — strongly preferred. Site changes, marketing campaigns, algorithm-update timing, external events in the analysis period. Used for "what changed" investigation. If GA4 access or cluster map is missing, ask. Both are foundational. ## Process 1. **Pull the data.** - GA4: sessions, users, engaged sessions, engagement rate, conversions, revenue. Dimensions: channel, page-type, page (URL), device, country, new/returning, audience. - GSC: clicks, impressions, position by query and page (for cross-reference layer). - Source-of-truth revenue: triangulate. 2. **Triangulation check.** - GA4 vs source-of-truth revenue: should match within 5% over the period. - If discrepancy >10%, flag as data-integrity issue and note caveat. 3. **Channel mix.** - Sessions / conversions / revenue by channel (current + prior). - Channel-level QoQ + YoY trends. - Channel mix shift: which channel grew share, which lost. 4. **Page-type rollup.** - Pages mapped to types: blog, PDP, collection, landing page, tool, programmatic, other. - Sessions / conversions / revenue per type. - Conversion-rate by type. 5. **Cluster-level rollup.** - Map every page in the analysis to a cluster from cluster-map.md. - For each cluster: sessions, conversions, revenue, conversion rate, organic-search share. - Identify clusters that gained / lost / are flat. 6. **Segmentation.** - Mobile vs desktop: sessions, conversions, conversion rate. - New vs returning: same metrics; identifies acquisition vs retention. - Geo (multi-market only): per-country breakdown. - Audience-based (if defined): per-audience. 7. **Funnel analysis.** - Session → engaged session → key event(s) → purchase. - Drop-off rates per stage. - Compare current vs prior to identify funnel-stage regressions. 8. **Anomaly detection.** - >20% MoM swings on segments worth >5% of total traffic. - Date-specific anomalies (sharp drops / rises on specific days). - Cross-reference with known external events (algorithm updates, campaigns, site changes, holidays). 9. **"What changed" investigation.** - For each meaningful pattern shift, evaluate against the change-set: - Site changes (releases, redesigns, migrations, content launches). - Tracking changes (GA4 config, GTM, consent updates). - Algorithm updates (Google official + community-reported). - Marketing campaigns (paid, email, PR, social). - Seasonality (annual / category cyclicality). - External events (competitor launches, news cycles, regulatory shifts). 10. **Insights & decisions.** - 5-10 insights, each with: pattern observed, hypothesis, recommended action, handoff skill. 11. **Prioritised action list.** - Each action: skill / handoff, severity (P0 / P1 / P2), effort estimate, owner. 12. **Open questions.** - 3-5 questions for the user / team that the analysis raised but couldn't fully answer. 13. **Write the report.** ## Output format ```markdown # Traffic Analysis — [Business name] **Analysis period:** [start date] → [end date] **Comparison period:** [start date] → [end date] **Analysis date:** [date] ## Executive summary - Total sessions: [n] ([+/- %] vs prior; [+/- %] YoY) - Total conversions: [n] ([+/- %] vs prior) - Total revenue (GA4-reported): [$n] ([+/- %] vs prior) - Triangulated revenue (source-of-truth): [$n] (discrepancy [%]) - **Headline:** [1-2 sentences capturing the dominant story] ## Channel mix | Channel | Sessions | Conversions | Revenue | Sessions Δ | Revenue Δ | |---------|----------|-------------|---------|------------|-----------| | Organic search | n | n | $n | +n% | +n% | | Direct | ... | ... | ... | ... | ... | | Paid search | ... | ... | ... | ... | ... | | ... | ... | ... | ... | ... | ... | **Channel insight:** [observation about channel mix; e.g. "organic +24% QoQ vs paid +6% — organic now 38% of revenue (was 31%)"] ## Page-type rollup | Page type | Sessions | Conversions | Conversion rate | Revenue | Sessions Δ | |-----------|----------|-------------|-----------------|---------|------------| | Blog | n | n | n% | $n | +n% | | PDP | ... | ... | ... | ... | ... | | Collection | ... | ... | ... | ... | ... | | Landing page | ... | ... | ... | ... | ... | | Tool | ... | ... | ... | ... | ... | | Other | ... | ... | ... | ... | ... | **Page-type insight:** [observation; e.g. "blog +40% but PDP organic flat — internal-link leak from blog → PDP"] ## Cluster rollup | Cluster | Sessions | Conversions | Conversion rate | Sessions Δ | Notes | |---------|----------|-------------|-----------------|------------|-------| | Sun-care | n | n | n% | +47% | Pillar launch | | Skincare-ingredients | ... | ... | ... | ... | ... | | Refills | ... | ... | ... | -22% | Decay; refresh recommended | | ... | ... | ... | ... | ... | ... | **Cluster insight:** [observation about cluster-level dynamics] ## Segmentation ### Mobile vs desktop | Segment | Sessions | Conversion rate | Sessions Δ | CR Δ | |---------|----------|-----------------|------------|------| | Mobile | n | n% | +n% | +/-pp | | Desktop | n | n% | +n% | +/-pp | **Segmentation insight:** [observation; e.g. "mobile sessions +28% but mobile CR dropped from 1.8% to 1.4% — checkout funnel investigation needed"] ### New vs returning [same shape] ### Geo (if multi-market) [same shape] ## Funnel analysis | Stage | Current | Prior | Δ | Drop-off rate | |-------|---------|-------|---|---------------| | Sessions | n | n | +n% | — | | Engaged sessions | n | n | +n% | n% drop | | Product views | n | n | +n% | n% drop | | Add-to-cart | n | n | +n% | n% drop | | Begin checkout | n | n | +n% | n% drop | | Purchase | n | n | +n% | n% drop | **Funnel insight:** [stage with the most concerning drop-off change] ## Anomalies & "what changed" ### Anomaly 1: [name / date] - **Pattern:** [what happened] - **Hypothesis:** [likely cause] - **Evidence:** [supporting data] - **Action:** [what to do about it] [Repeat for 2-4 anomalies if relevant] ## Insights & decisions 1. **[Insight title].** [Pattern observed; hypothesis; recommended action; handoff skill.] 2. **[Insight title].** [...] 3. **[Insight title].** [...] 4. **[Insight title].** [...] 5. **[Insight title].** [...] ## Prioritised action list | # | Action | Skill / handoff | Severity | Effort | Owner | |---|--------|-----------------|----------|--------|-------| | 1 | Audit blog → PDP linking | `audit-page-internal-links` | P0 | 1 day | SEO | | 2 | Investigate mobile checkout drop | UX / CRO handoff | P0 | 2-3 days | UX | | 3 | Refresh refills cluster | `plan-content-refresh` | P1 | 4-6 days | Content | | 4 | Standardise on DDA attribution | Reporting update | P1 | 4 hrs | Marketing | | 5 | Schedule next quarterly analysis | Calendar | P2 | — | — | ## Open questions for the team - [ ] Confirm the algorithm-update / external-event dates that may have driven the April anomaly. - [ ] Confirm whether the mobile-CR drop coincides with a checkout flow change. - [ ] Confirm the attribution-model decision (DDA vs last-click) for next quarter. - [ ] Confirm the cluster-priority list for the next 90 days (drives next-quarter reporting). ``` Save the produced file to `businesses/<slug>/analytics/traffic-analysis-[YYYY-MM-DD].md`. After writing, tell the user the file path so they can open it. ## Quality bar - Triangulation done with source-of-truth revenue; discrepancy quantified. - Comparison windows are QoQ + YoY (where retention allows), not 28-day. - Channel mix, page-type, and cluster rollups all present — none skipped. - Segmentation covers mobile vs desktop, new vs returning at minimum. - Funnel drop-offs identified and compared to prior period. - Anomalies investigated with "what changed" hypothesis, not just observed. - Insights have pattern + hypothesis + action + handoff skill — not just observations. - Action list is prioritised + effort-estimated; not a wishlist. - Open questions cover the 3-5 things the analysis surfaced but couldn't fully answer. ## Common mistakes to avoid - Don't compare 28-day windows. Too noisy; QoQ + YoY are the right defaults. - Don't skip triangulation. GA4-only revenue numbers in isolation aren't auditable.