Choosing store metrics for smart decisions: Name the decision a metric supports to ensure relevance.; Use context-specific thresholds, not generic benchmarks.; Count by session, order, item or customer—match unit to purpose.
Image: Ecommerce Insight Desk

Measurement Planning

Part of Ecommerce measurement planning

Choosing store metrics from business decisions

Choose ecommerce metrics by linking each business decision to a counting unit, comparison, owner and action.

Pick a store metric by naming the decision it informs. Specify the outcome, a useful comparison and the action someone can take. If a measure cannot support that chain, it may belong in a diagnostic report, not the main review.

Write a measurement brief

For each decision, name its owner and review period. Describe what change would prompt investigation or action; set any numerical threshold from the store's own context rather than treating an example as a benchmark.

DecisionCandidate measureContext before acting
Check whether checkout needs attentionQualifying orders and a defined checkout-stage rateTraffic mix, checkout starts and incidents
Decide whether to keep a product offerProduct sales or contribution under a stated ruleUnits, discounts, reversals and stock
Decide whether acquisition spend should riseOutcomes for a defined customer groupSpend, first-order mix and later outcomes observed so far

A store may need paid orders for one decision and all orders placed for another. Give each measure a label that states which population it counts.

Choose the counting unit

Write the unit beside the metric: event, session, order, item or customer. Shopify's Conversion rate breakdown report calculates each step's rate as sessions for that step divided by total sessions. That is a session-based measure, not an order count. A different rate needs its own numerator and denominator, with a clear definition of the visitor or session population.

For a product decision, use item-level units and amounts. For a repeat-customer decision, count distinct customers under a documented identity rule. Do not combine a customer numerator with a session denominator. When comparing session rates over time, check whether platform measurement changes affected the periods.

Recommended Counting Units for Different Decisions

  1. Product decisionItem-level units and amounts
  2. Repeat-customer decisionDistinct customers under documented identity rule
  3. Session-based conversion rateSessions for step ÷ total sessions

Choose the comparison before reading the result

Select a reference period or group suited to the trading pattern. Note promotions, site changes and stock gaps, and show the underlying count beside each rate. A lower order count on a low-traffic day can call for a different investigation from a lower completion rate while traffic is steady.

Match the measure's timing to the decision. Basket additions may prompt a same-day check; sales after refunds may suit a later commercial review. Item views and basket additions are observed actions, not confirmed sales.

Keep a short metric contract

For each selected measure, record the decision, owner, formula, source, unit, time zone, update time, exclusions and action rule. For a sales amount, state whether tax, shipping, discounts and later reversals are included. Shopify's sales reports define terms such as gross sales, discounts and sales reversals; an internal measure called “net” may follow another rule.

After a review, ask whether the measure informed an action or whether its definition consumed the discussion. Keep useful diagnostic detail available, and remove a headline measure that repeatedly fails its assigned decision.

More from Measurement Planning