Activation and Feature Adoption
Activation is a product hypothesis expressed as observable behavior. It is not a universal event named “activation.” Feature adoption likewise requires an eligible population, a successful-use event, a time window, and a threshold connected to a decision. A cutoff chosen after inspecting outcomes invites circular reasoning.
Prerequisites
Complete retention and churn. Retain the fixture's event validation, consent, identity, and restatement rules.
Learning goals
- Translate a value hypothesis into an auditable behavioral milestone.
- Separate eligibility from success and exposure from use.
- Report threshold sensitivity rather than canonizing one cutoff.
- Avoid treating adoption association as evidence of product impact.
From hypothesis to threshold
Suppose the team asks whether active actors successfully use report export. The denominator is actors with valid product activity, not all event rows. The numerator is actors with at least a declared number of valid feature_used(feature="report_export") events.
One use answers discovery or first successful use. Two uses asks for repetition. The fixture yields different answers. Neither threshold is “the truth”: select it before the analysis based on the decision, validate that the event means successful completion, and retain the sensitivity table.
Metric contract
- Population: actors with a valid, consented page view or feature use on 2026-10-01 UTC.
- Unit: resolved actor × UTC day.
- Observation window:
[2026-10-01T00:00:00Z, 2026-10-02T00:00:00Z)at the restated close. - Denominator: distinct eligible active actors, including eligible non-adopters.
- Missingness: malformed feature properties and denied measurement are excluded and counted; lack of a use event may mean no use or missing instrumentation.
- Identity: 2026-10-02 snapshot; no household inference and no probabilistic merging.
Worked example
import json
from collections import Counter
from datetime import datetime
from pathlib import Path
root = Path("tests/fixtures/product-analytics")
rows = [json.loads(line) for line in (root / "events-v1.jsonl").read_text().splitlines()]
plan = json.loads((root / "tracking-plan-v1.json").read_text())
parse = lambda value: datetime.fromisoformat(value.replace("Z", "+00:00"))
deduped = {}
for row in sorted(rows, key=lambda item: parse(item["received_at"])):
deduped.setdefault(row["event_id"], row)
def valid(row):
required = plan["events"][row["event_name"]]["required_properties"]
kinds = {"string": str, "boolean": bool}
return all(key in row["properties"] and isinstance(row["properties"][key], kinds[kind]) for key, kind in required.items())
def actor(row):
return row["user_id"] or plan["identity_snapshot"]["links"].get(row["anonymous_id"], row["anonymous_id"])
eligible = [row for row in deduped.values() if row["analytics_consent"] and row["purpose"] == "product_measurement" and valid(row)]
active = {actor(row) for row in eligible if row["event_name"] in {"page_view", "feature_used"}}
uses = Counter(actor(row) for row in eligible if row["event_name"] == "feature_used" and row["properties"]["feature"] == "report_export")
at_one = sum(uses[key] >= 1 for key in active)
at_two = sum(uses[key] >= 2 for key in active)
print(f"eligible_active={len(active)} adopters_at_1_use={at_one} rate={at_one/len(active):.3f}")
print(f"adopters_at_2_uses={at_two} rate={at_two/len(active):.3f} threshold_is_decision_specific=True")
Expected output:
eligible_active=4 adopters_at_1_use=2 rate=0.500
adopters_at_2_uses=0 rate=0.000 threshold_is_decision_specific=True
Limitations and common mistakes
- Calling any easy-to-track action activation without evidence that it represents value.
- Denominating adoption by all users when only exposed or eligible users could use the feature.
- Choosing a threshold after looking at retention and then claiming it predicts success out of sample.
- Ignoring failed attempts because only successful events are instrumented.
- Claiming the feature caused retention because adopters retain more; motivation and opportunity confound the comparison.
Practice
Write decision rationales for one-use and two-use thresholds. Add a hypothetical feature_exposed event and explain how exposure-based eligibility changes the denominator. Predeclare what result would cause the team to revise instrumentation rather than the feature.
Next steps
Engagement frequency and intensity compares counts, rates, and distributions without hiding threshold sensitivity.
Primary sources
Checked 2026-10-19: Google's HEART framework paper, Google Analytics event collection model, and W3C privacy principles.