1. Descriptive Analytics

KPIs and Denominators

Define metrics that preserve population, units, incentives and operational meaning

KPIs and Denominators

A metric is a contract

“Late-delivery rate” is undefined until all fields below are fixed:

FieldHarborMart contract
numeratoraccepted orders delivered after promised end
denominatoraccepted orders with a completed delivery outcome
unitorder, not parcel or delivery event
periodpromise date in the completed calendar week
timezoneEurope/London
exclusionsdocumented test orders; cancellations reported as a separate guardrail
revisionrestate for late-arriving completion events for 14 days
ownerfulfilment analytics lead

If 80 of 1,000 completed accepted orders are late, the rate is 8%. Dividing by 1,200 checkout attempts instead gives 6.7% and answers a different question.

Ratios do not average themselves

StoreLateCompletedRate
A91009%
B189002%

The unweighted average of store rates is 5.5%. The network rate is

9+18100+900=2.7%.\frac{9+18}{100+900}=2.7\%.

Use the first only if the estimand is “equally weighted typical store.” Use the second for “share of all completed orders.” Weighting is part of the question, not a formatting choice.

Py

Audit a pooled KPI

Idle

Driver trees expose arithmetic, not causality

Weekly contribution may be decomposed as

visits×conversion×average order value×contribution margin rate.\text{visits}\times\text{conversion}\times\text{average order value} \times\text{contribution margin rate}.

This identity helps locate a change. It does not prove that changing one component while holding the others fixed is feasible: a discount may raise conversion and lower margin simultaneously.

Outcome, driver and guardrail

RoleExampleWhy needed
outcomecontribution per eligible customerlinks to value
drivercheckout conversionreacts quickly
guardrailrefunds, lateness, complaintsdetects displaced harm
processpicker minutes per orderdiagnoses implementation

Optimising a driver without guardrails invites gaming. Faster picking can raise errors; fewer refunds can reflect a harder claims process rather than better service.

Metric-change protocol

When a definition changes:

  1. version the formula and effective date;
  2. backfill only when the required historical fields exist;
  3. show old and new definitions during transition;
  4. explain the break rather than drawing a continuous trend;
  5. record affected decisions and dashboards.

Quick check

The late rate falls after orders without a valid delivered timestamp are excluded. What must be checked?

Answer
Whether missing outcomes are selective. If failed deliveries are more likely to lack a timestamp, exclusion biases the denominator. Report missing-outcome rate and investigate its process.

Next: Segments, Cohorts and Funnels

Copyright © 2026