Skip to main content

Feature Adoption over the Product Lifecycle

One adoption rate cannot answer every lifecycle question. Discovery, first successful value, repetition, habitual use, and retirement have different eligible populations, observation windows, and interventions. A feature can gain first uses while repeated use and support burden worsen.

Prerequisites​

Complete metric alerts and anomaly triage. Use validated, deduplicated, restated events before assigning lifecycle stages.

Learning goals​

  • Define mutually interpretable lifecycle stages from observable events.
  • Match the metric to a launch, growth, maturity, or retirement decision.
  • Keep exposure, eligibility, first value, and repetition distinct.
  • Avoid declaring maturity from a short before/after window.

Decision contract​

For the staged report-export rollout, the owner asks whether discovery reaches eligible actors, whether a use completes, and whether use repeats. The current fixture observes all four actors visiting the report page and eventually using export. Only two actors have at least two valid uses across the two-day history. The short history cannot establish habit, durable value, or market maturity.

Launch decisions emphasize valid exposure and first-value failures. Growth decisions add time-to-value, repeat-use cohorts, support, and subgroup access. Mature-feature decisions watch durable penetration, reliability, cost to serve, and substitution. Retirement decisions require migration coverage and harm monitoring, not a falling use count alone.

Worked example​

import json
from collections import Counter
from pathlib import Path

rows = [json.loads(line) for line in Path("tests/fixtures/product-analytics/review-events-v1.jsonl").read_text().splitlines()]
deduped = {row["event_id"]: row for row in reversed(rows)}
baseline_active = {row["actor_id"] for row in deduped.values() if row["event_name"] == "page_view" and row["occurred_at"].startswith("2026-10-01")}
baseline_users = {row["actor_id"] for row in deduped.values() if row["event_name"] == "feature_used" and row["occurred_at"].startswith("2026-10-01")}
current_active = {row["actor_id"] for row in deduped.values() if row["event_name"] == "page_view" and row["occurred_at"].startswith("2026-10-02")}
current_users = {row["actor_id"] for row in deduped.values() if row["event_name"] == "feature_used" and row["occurred_at"].startswith("2026-10-02")}
first_close_users = {actor for actor in current_users if actor != "d"}
uses = Counter(row["actor_id"] for row in deduped.values() if row["event_name"] == "feature_used")
stages = {"discovered": len(current_active), "first_value": len(current_users), "repeated": sum(value >= 2 for value in uses.values())}
print(f"baseline={len(baseline_users)}/{len(baseline_active)} first_close={len(first_close_users)}/{len(current_active)} restated={len(current_users)}/{len(current_active)}")
print(f"stages={stages} maturity_claim=False")

Expected output:

baseline=2/4 first_close=3/4 restated=4/4
stages={'discovered': 4, 'first_value': 4, 'repeated': 2} maturity_claim=False

Limitations and common mistakes​

  • Visiting a page is not exposure if the feature was below the fold, disabled, or unavailable.
  • A successful event may miss failed attempts and accessibility barriers.
  • Repeat use can indicate repeated need or repeated failure; pair counts with outcome evidence.
  • Cohorts need sufficient follow-up and a declared censoring policy.
  • Lifecycle labels summarize evidence; they do not cause behavior or dictate investment.

Practice​

For one feature, predeclare observable discovery, first-value, repeated-value, and retirement events. Name the decision and guardrail at each stage and state the minimum follow-up needed.

Next steps​

Monetization and unit economics adds explicit financial scope without reducing the product decision to money alone.

Primary sources​

Checked 2026-10-19: Google, HEART framework paper; Google Analytics, recommended events; NIST, reliability data analysis.