I85PREP
⌘ K
Adobe Analytics Architect Master study materials
Adobe Analytics Architect Master study resources

Calculated Metric Drift After a Taxonomy Change

Adobe Analytics Architect Master · Put your knowledge to work in context.

All case studies

Calculated Metric Drift After a Taxonomy Change

A subscription media company tracks trial starts with a trialStart event and a planType eVar. A calculated metric 'trial conversion rate' = trialStart / visits is used on executive dashboards. After a taxonomy change, planType values were re-mapped (e.g., 'premium' split into 'premium_monthly' and 'premium_annual'), and a new trialStartV2 event was introduced for a new checkout flow while the old trialStart still fires on the legacy flow. Executives report the conversion metric 'dropped' and blame the new checkout. The team wants to know whether the metric is broken or the business changed.

1. Diagnose whether the conversion-rate drop is a metric-definition problem, a taxonomy problem, or a real business change.

Read the worked answer

Three candidate causes; test each:

  1. Metric-definition problem: the calculated metric uses only trialStart. If the new checkout flow fires trialStartV2 and not trialStart, the numerator is undercounted - the metric is broken, not the business. Check: trend trialStart vs. trialStartV2; if trialStartV2 grows while trialStart falls, the metric misses the new flow.
  2. Taxonomy problem: planType re-mapping can break any segment or breakdown used alongside the metric. If dashboards filter on planType == 'premium', the split into monthly/annual may exclude or mis-bucket traffic. Check: 'unspecified'/'none' rates for planType before/after the change.
  3. Real business change: if both events are counted and taxonomy is clean, the drop may be genuine. Check: reconcile trial counts against the billing system.

Most likely: a combination of 1 and 2 - the metric omits trialStartV2 and dashboard segments reference old planType values. The new checkout is a scapegoat.

2. Recommend a corrected metric definition and a governance process to prevent recurrence.

Read the worked answer

Corrected metric:

  • Redefine the numerator as trialStart + trialStartV2 (or, better, consolidate to a single event with a flow dimension so the metric has one source).
  • Keep the denominator as visits only if that is the intended business definition; otherwise use a more appropriate denominator (e.g., checkout starts) and document it.
  • Update dashboard segments to use the new planType values, and add a rollup (e.g., a planTier classification or derived dimension) so 'premium' still aggregates monthly+annual.

Governance process:

  • Metric registry: document every calculated metric, its numerator/denominator, owner, and dependencies on events/variables.
  • Change control: taxonomy or event changes require a review of dependent metrics and segments before release.
  • Reconciliation: monthly check of trial counts vs. billing system.
  • Monitoring: alert on 'none'/'unspecified' spikes and on divergence between paired events (e.g., trialStart vs. trialStartV2).
  • Deprecation plan: retire the legacy event on a schedule rather than running two indefinitely.