Analytics Says Traffic Dropped, but Sales Are Flat
An e-commerce site on AEM Sites uses Adobe Analytics via a tag management setup. After a recent release that changed the page template and added a consent banner, the Analytics 'visits' metric dropped ~40% week over week, while orders (from the order system) are flat. The team also enabled a new CDN with bot filtering and changed the page load order so the Analytics library loads after the consent banner resolves. Some pages use client-side rendering for product grids. The business is asking whether the site is broken and whether to roll back.
1. Is the site broken? Diagnose the traffic drop and advise on rollback.
Read the worked answer
Sales flat + visits down 40% strongly suggests a measurement problem, not a traffic problem. Likely causes:
- Consent gating. If the Analytics library now loads only after consent resolves, visitors who never interact with the banner (or decline) are not tracked. That alone can cut measured visits substantially while real traffic is unchanged.
- CDN bot filtering. Filtering bots removes traffic that was previously counted (and inflated). A 40% drop is plausible if bot traffic was high.
- Load-order/timing. Analytics firing later can miss quick bounces or SPA route changes if the tracking call is not re-fired on client-side navigation.
- Template change. If the new template dropped or renamed a data element (page name, prop), reports can misattribute or lose hits.
Advice: do not roll back yet. First reconcile: compare order counts (flat) with Analytics visits; check consent-gated tracking rate, bot-filter config, and whether SPA route changes fire page views. Fix tracking (fire on consent where lawful, re-fire on route change, validate data elements) and re-baseline. Rollback would not restore real traffic and could undo legitimate bot filtering.
Judgement: separate 'traffic changed' from 'measurement changed' using an independent source (orders). Rollback is the wrong lever for a tracking regression.
2. The business wants a reliable weekly report going forward. What should the practitioner recommend?
Read the worked answer
Build a report that is resilient to tracking changes:
- Primary KPI from the order system (revenue/orders) as the source of truth; Analytics is the behavioral layer, not the revenue truth.
- Track a stable denominator (e.g., sessions with consent, or server-side page views) so consent/bot changes are visible as a step change, not a mystery.
- Annotate the report with release dates, consent changes, and CDN/bot-filter changes so week-over-week dips are explainable.
- Validate data elements after every template release (page name, product ID, event names) with a small QA checklist.
- SPA tracking - ensure virtual page views fire on route change.
- Alerting on sudden metric step-changes so the team investigates before the business panics.
Judgement: the report should make measurement changes explicit and tie behavioral metrics to a business outcome that is measured independently, so a tracking regression cannot masquerade as a demand collapse.