How to Measure Behavior Change

Definition. Measuring behavior change means measuring a specific target behavior against a baseline with explicit denominators and time windows. In Behavioral Strategy, you treat behavior as the KPI (not clicks or attitudes).

From Behavioral Strategy, developed by Jason Hreha.

Axiom. If two people can’t observe it and agree whether it happened, it’s not measurable.

How this differs from common analytics

  • Not engagement-first: you measure the target behavior, not clicks, opens, or time in app.
  • Not attitude-first: self-report and NPS are supportive signals; they do not define behavior change.
  • Not windowless: every metric is reported with an explicit denominator and time window.

The measurement setup (5 steps)

  1. Define the target behavior operationally.
    • Template: population does action in context with frequency/window.
    • If two people can’t observe it and agree whether it happened, it’s not measurable.
  2. Choose the denominator (who counts).
    • Examples:
      • “All exposed users”
      • “All eligible users”
      • “All new accounts in cohort week X”
    • Avoid denominator drift (changing who counts midstream).
  3. Choose the time window (when it must happen).
    • Examples:
      • “Within first session”
      • “Within 24 hours”
      • “Within 7 days of signup”
    • You can’t interpret behavior without a window.
  4. Instrument behavior events (or observation rules).
    • Minimum event set (example naming aligned to a typical product flow):
      • exposure_shown
      • behavior_attempt_started
      • behavior_completed
      • behavior_abandoned
    • See Measurement Standards for reporting rules and event fields.
  5. Pre-register what “success” means.
    • Define thresholds and what decisions they trigger (proceed, redesign, reselect).
    • If you don’t pre-register, you’ll “discover” wins in noise.

Core metrics (what to report)

  • Δ‑B: change in target behavior completion rate vs baseline (percentage points), with denominator and window. See Δ‑B.
  • TTFB: time from first exposure/start to first valid completion of the behavior. Shorter TTFB is often a leading indicator of higher retention in products; validate per cohort. See TTFB.
  • Behavior retention: D30/D180 retention of the behavior (not app opens), with cohort definition.
  • bPMF: sustained behavior at scale in market conditions (behavioral product-market fit). See bPMF.

Common traps (what breaks measurement)

  • Proxy metrics: “engagement” or clicks that don’t represent the target behavior.
  • Undefined windows: measuring “eventually” guarantees false positives.
  • Survivorship bias: looking only at retained users instead of the eligible/exposed denominator.
  • Mixed behaviors: bundling multiple behaviors into one KPI hides failure points in the chain.
  • Changing instrumentation mid-test: makes baselines incomparable (log changes and reset baselines if needed).

Outputs (what you should have when you’re done)

  • A target behavior definition (operational, observable)
  • A denominator and time window (documented and stable)
  • An instrumentation spec (events/fields + success rules)
  • A baseline vs post report that includes Δ‑B, TTFB, and retention (with denominators/windows)

Templates (copy/paste)

Example measurement spec 

target_behavior:
  population: "New teams (B2B)"
  action: "Send 3 messages in #general"
  context: "Within 24 hours of workspace creation"
denominator: "All new workspaces created"
window: "First 24 hours"
success_criteria:
  completion_rate_pp: ">= 35"
  ttfb_median: "<= 15 minutes"
instrumentation:
  exposure_shown: { surface: "onboarding", variant: "v2" }
  behavior_completed: { behavior_id: "send_3_msgs_24h", channel: "#general" }
reporting:
  delta_b: { baseline_pp: 22, post_pp: 37, delta_pp: 15 }
  retention_d30_pp: { baseline_pp: 8, post_pp: 14 }

Frequently asked questions

What is Delta-B (Δ‑B)?

Δ‑B is the change in target behavior completion rate versus a baseline, reported in percentage points with an explicit denominator and time window. For example, if 22% of new workspaces send 3 messages in 24 hours at baseline and 37% do after intervention, Δ‑B is +15 pp. Always report the denominator and window alongside the number.

What is Time to First Behavior (TTFB)?

TTFB measures the elapsed time from first exposure (or account creation) to the first valid completion of the target behavior. Shorter TTFB often predicts higher retention. Report median and P90 values with a defined start event. For example, if median TTFB drops from 4 days to 18 hours, the behavior chain has less friction.

What denominator should you use?

Use a stable denominator tied to who could realistically perform the behavior, such as all exposed users, all eligible users, or a defined cohort. Avoid denominator drift (changing who counts midstream). The choice of denominator shapes every metric you report, so document it before collecting data and keep it consistent across comparison periods.

What is behavior retention?

Behavior retention tracks whether a defined cohort repeats the target behavior over time, such as D30 or D180 repeat completion rate. It distinguishes one-time activation from sustained behavior. Report with an explicit cohort definition, denominator, and window. High initial completion with low D30 retention signals the behavior is feasible but not self-sustaining.

What is bPMF?

Behavioral Product-Market Fit (bPMF) means the target behavior persists at scale in market conditions with viable economics, not one-off activation wins. It is the final validation gate. You have bPMF when the behavior sustains across cohorts without intervention dependency and the unit economics work. It is the behavioral equivalent of traditional Product-Market Fit.

See also

Jason Hreha· Updated February 2, 2026
On this page