--- name: run-a-monthly-performance-review description: >- The monthly deep pass that weekly reviews cannot do: month-over-month and year-over-year KPI assessment with full-year trajectory, tracking-integrity reconciliation against backend/GA4 numbers with explicit discrepancy bands, a competitive check, a settings-drift and auto-applied- changes audit through change history and recommendations, baseline/threshold recalibration, and a prioritized month plan, exportable to Sheets. VigilDog automates every pull and comparison; the human owns target judgment and strategy calls. Run in the first week of each month, or when onboarding an account to establish baselines. Not the weekly tactical loop (run-a-weekly-performance-review), not the strategic goal/structure reset (run-a-quarterly-business-review), and the competitive section delegates to analyze-auction-insights when it deserves depth. --- # Monthly Performance Review ## Purpose Bridge tactical weeks and strategic quarters. The monthly review catches what moves too slowly for weekly visibility, tracking drift, settings drift, competitive creep, stale thresholds, and converts a month of data into a recalibrated baseline and a concrete plan. It is decision-oriented and time-boxed (~90 minutes), not an open-ended data safari. ## When to run - First week of every month, covering the prior full month. - Onboarding a new account (initial baseline construction). - Month three of a quarter: run it as the input feed for run-a-quarterly-business-review. ## When NOT to run - Mid-month tactical questions, run-a-weekly-performance-review. - Goals/structure/channel-mix are in question, run-a-quarterly-business-review. - A single subsystem needs help (tracking broken, bids erratic), go straight to that skill; the monthly review is a sweep, not an ambulance. ## Prerequisites - The month's weekly review outputs (flags, actions, carried items). - Backend truth for reconciliation: CRM/ecommerce export or GA4 access via the VigilDog GA4 tools. - Documented targets and current status-board thresholds. - Prior month's review document for continuity. ## Procedure 1. Performance trends (~15 min). `gads_get_account_performance` for the full month vs prior month; `gads_run_gaql_query` for the same month last year (seasonality control). Table: conversions, spend, conversion value, CPA/ROAS, each vs target and vs both comparison periods. Project full-year pace from YTD via `gads_run_gaql_query` monthly buckets. Rank campaigns with `gads_get_campaign`/GAQL: top 5 by spend with target status, bottom 3 by efficiency as reallocation candidates. 2. Tracking integrity (~10 min). Reconcile the month's conversions and revenue against backend: `ga4_run_report` / `ga4_batch_run_reports` for GA4-side purchases/leads, plus the user's CRM/platform export. Compute discrepancy percentage and grade it (see decision rules). Then audit conversion plumbing with `gads_list_conversion_actions`: every intended action active and recording, correct primary/secondary split, windows and counting unchanged, no zombie duplicate actions. Any red here taints every other number in the review, flag prominently. 3. Competitive landscape (~10 min). Quick pass via `gads_get_auction_insights`: own IS trend, entrants, exits, and whether CPC movement correlates with competitor pressure. If flags fire (large IS swings, strong new entrant), spawn the full analyze-auction-insights procedure rather than deep-diving here. 4. Settings and structure audit (~15 min). - Drift check: `gads_get_campaign` across active campaigns, geo targeting, network settings, language, ad rotation, campaign conversion goals, versus the intended configuration on record. - Unattended automation: `gads_get_change_history` filtered to system/auto-applied sources for the month; list every change nobody made deliberately. Reversals of strategy-conflicting changes are executed via the relevant write skill with preview → user approval → commit (consult gads_policy_guardrail before the session's first write). - Recommendations queue: `gads_list_recommendations`; classify each as aligned (adopt via its proper skill), conflicting (dismiss, HUMAN STEP in the UI), or ignorable. - Structure: campaigns under ~30 conversions/month via GAQL = consolidation candidates for the structure backlog. 5. Baseline recalibration (~10 min). Compare each status-board threshold (CPA, ROAS, CVR, CTR bands) against the month's actuals. If a metric has sat consistently above or below its band for 2+ months, propose moving the band, up when green has become easy, down when the market has genuinely shifted. HUMAN STEP: the user ratifies threshold and target changes; document rationale. 6. Action plan (~10 min). Consolidate findings into P1/P2/P3 with owners and dates, each traced to the phase that produced it. Reassess the account's single active focus constraint: moved, resolved, or replaced. Write the review document; export tables via `sheets_export_report` for the stakeholder-facing copy. ## Decision rules - Tracking discrepancy bands: under 10% = normal attribution slack, no action; 10-20% = monitor, investigate if the trend is worsening; over 20% = tracking is broken, route to the measurement skills and treat this month's efficiency readings as suspect. - YoY beats MoM whenever seasonality exists; a MoM drop that matches last year's curve is not a finding. - Full-year projection off target by more than ~15% → escalate to goal conversation, not more tactical tweaks. - Threshold recalibration requires 2+ consecutive months of evidence, never off one month. - Auto-applied changes are guilty until proven aligned: default is revert unless they match documented strategy. - Sub-30-conversions/month campaigns get consolidation review, below that, Smart Bidding runs data-starved. - Time-box at 90 minutes; a review that needs 3 hours is doing the weekly reviews' missed work, fix the weekly cadence instead. - Reconciliation source hierarchy: platform/CRM backend beats GA4 beats Google Ads for "what actually happened"; GA4 is a cross-check with its own attribution lens, not ground truth. - If cart data is live (set-up-cart-data-and-profit-tracking), report gross profit alongside revenue in phase 1, a revenue-flat, profit-down month is a real finding revenue tables hide. ## Common failure modes - Re-running a big weekly review and skipping the monthly-only dimensions (tracking, drift, baselines), the checklist above exists to prevent exactly this. - Skipping reconciliation because "tracking was fine last month"; drift is gradual and only this check catches it early. - Never touching thresholds, so the status board goes stale and either everything is false-positive red or comfortably meaningless green. - Letting auto-applied recommendations accumulate unreviewed for months. - No written output, findings evaporate and next month rediscovers them. ## Related skills - Fed by: run-a-weekly-performance-review (four of them), and daily health checks where they exist. - Nested: analyze-auction-insights (competitive phase, when flags warrant depth). - Feeds: run-a-quarterly-business-review. - Routes findings to: set-up-google-ads-conversion-tracking / configure-conversion-adjustments (tracking reds), allocate-budget-across-campaigns (reallocation candidates), monitor-and-maintain-bid-strategy-health (target/limit issues), review-and-optimize-campaign-structure (consolidation backlog). - Ecommerce accounts additionally check: set-up-cart-data-and-profit-tracking coverage and merchant feed health during the tracking phase. - Lead-gen accounts additionally check: set-up-offline-conversion-tracking import reliability (upload success rate and CRM-vs-Ads counts) during the tracking phase.