KPI Design for Processes

KPIs translate a process into facts: speed, quality, cost, and control. Good KPIs use operational definitions, trustworthy data, and a review cadence that drives action. Track the critical few; make them testable.

Overview

KPIs are the small set of measures that show if a process delivers its outcome. They must be defined, reproducible, and tied to the step where value or risk occurs. Use one source of truth for daily and month-end views. Avoid parallel spreadsheets.

Principles

Outcome first

State the customer, the promise, and the test for success. Pick 3–5 KPIs per flow.

Defined & testable

Write the operational definition. Name the event timestamps, the unit of analysis, and the sampling rule.

Actionable

Assign owners and thresholds. When a KPI breaks, action follows within the review window.

Types of KPIs

Flow (speed)

  • Lead time, cycle time, throughput, WIP
  • On-time %; queue time; touch time

Quality

  • First-pass yield, defect rate, rework rate
  • Right-first-time; escape rate

Cost & service

  • Cost per unit/case; unit margin
  • SLA attainment; backlog size & aging

Also track control health (KCIs): late approvals, failed reconciliations, access review exceptions, control test pass rate.

Operational definitions & formulas

Core formulas

  • Lead time = end_ts − start_ts (request → delivery)
  • Cycle time = active work time per unit
  • Throughput = completed units / period
  • WIP = units in process now
  • First-pass yield = good units / total units
  • Defect rate = defects / opportunities (or per unit)
  • Cost per unit = total process cost / output count
  • SLA attainment = met commitments / total commitments

Little’s Law (queueing)

WIP = Throughput × Lead time. If WIP grows and throughput is flat, lead time rises. Use this to detect hidden queues.

Reference: MIT/Stanford explanations; Lean VSM practice. MIT OCW · Lean Institute

Measurement system design

Data model

  • Define the unit (order, ticket, batch) and the lifecycle events (created, started, paused, completed).
  • Specify sources and system clocks; store UTC and local when needed.
  • Keep IDs stable; track rework links and parent/child items.

Operational definition

  • Exact filter (inclusions/exclusions); numerator/denominator
  • Time base (calendar vs. business hours; holiday rules)
  • Sampling: full population preferred; else risk-based sample

Data quality

  • Completeness, accuracy, validity, timeliness (CAVT)
  • Detect duplicates; record late/edited timestamps
  • Document lineage; avoid manual re-key

References: ISO 9001 documented information; NIST e-Handbook (measurement & statistics). ISO · NIST

Baselines, targets & segmentation

Baselines

  • Use last 90–180 days for a stable baseline where seasonality is low.
  • Show distribution (median, p90/p95) not only the mean.

Targets

  • Set targets above common cause variation (use SPC limits).
  • Define acceptable error bands; avoid whiplash from outliers.

Segmentation & normalization

  • Segment by product, channel, region, customer tier.
  • Normalize by size/complexity (per item, per 100 orders, per $1k).
  • Watch for Simpson’s paradox when mixing segments.

Visuals & review cadence

Daily

Flow board: WIP, starts, completes, blockers. Show yesterday and 7-day trend.

Weekly

Trend board: lead time (median/p90), FPY, backlog aging, SLA hits/misses, actions closed.

Monthly

SPC/control view, conformance checks, cost per unit, and benefits tracking vs. plan.

Statistical control (SPC)

Why SPC

Distinguish common-cause noise from special-cause signals. Act only when there is a signal.

Charts

  • p-chart for proportions (defect rate, FPY)
  • X-bar/R (or XmR) for continuous measures (cycle time)
  • u-chart for defects per unit

References

NIST e-Handbook of Statistical Methods; ASQ SPC primers. NIST · ASQ

Pitfalls & anti-patterns

Averages that lie

Averaging ratios or mixing segments hides reality. Show distribution and per-segment views.

Vanity metrics

Counts without denominators; dashboards with color but no decisions. Tie each KPI to an owner and a threshold.

Utilization obsession

Maxed utilization increases WIP and lead time (Little’s Law). Protect flow first.

90-day starter

Days 0–30

  • Pick one flow. Define 3–5 KPIs with operational definitions.
  • Baseline last 90 days; add median and p90.

Days 31–60

  • Install daily/weekly boards and owners.
  • Add SPC to one KPI; stop reacting to noise.

Days 61–90

  • Segment and normalize; resolve one chronic bottleneck.
  • Publish the scale plan and retire vanity metrics.

References

  • ISO 9001 documented information / process approach — iso.org
  • NIST e-Handbook of Statistical Methods — nist.gov
  • MIT OCW: Little’s Law — ocw.mit.edu
  • Lean Enterprise Institute: Value-stream mapping — lean.org
  • ASQ: Control charts (SPC) — asq.org

Design the critical few KPIs. Make them testable. Review them on a clock.

If you want a KPI dictionary template and SPC starter, ask for a copy.

Contact us