"Is 38% D7 retention good?" is the question every Chrome extension founder asks once, gets no useful answer, and moves on. The Web Store doesn't publish benchmarks. Generic SaaS benchmarks don't fit because extensions live in a different shape — install is essentially free, use is silent, and the surface is fragmented. This guide is the actual numbers, split by category, with what to read from each: activation rate, D7 and D30 retention, time-to-uninstall distribution, churn rate, and review rate. All ranges derived from observed patterns across customer extensions and public Web Store data — your mileage will vary, but you'll know which direction to read it.
The five category buckets
All benchmarks below split extensions into five buckets, because intra-category variance is much smaller than inter-category variance — comparing a translator's retention to a wallpaper extension's is meaningless; comparing it to another translator's is informative:
- Daily-driver tools — ad/tracker blockers, dark-mode, translators that auto-trigger, accessibility rewriters, password managers. Run continuously. The user rarely thinks about them.
- Productivity utilities — note clippers, screenshot tools, formatters, color pickers, currency converters. User-invoked. Useful several times per week.
- Developer tools — DevTools panels, API inspectors, JSON formatters, regex testers. Specialist audience, deep but narrow use.
- Niche utilities — single-site enhancers (one platform, one workflow), counters, dashboard widgets. Small TAM, dedicated users.
- Entertainment / themes — wallpapers, new-tab replacers, sound effects, fun toys. High install volume, low engagement.
Activation rate by category
Activation = % of installers who fire at least one feature-use event within 24 hours of install. The definition matches the DAU/MAU guide. Range is the typical band; the top of the range is the moat.
Daily-driver tools activate on install because the tool runs automatically; entertainment installs are mostly impulse and rarely re-opened.
Three reads:
- Daily-driver tools win activation by default — their feature-use event fires the moment the user navigates to any matching page. The user doesn't have to do anything. Other categories require a deliberate interaction.
- Niche & entertainment under 45% are structurally limited, not broken. If you're in those buckets, optimize for the activated user's value, not for the activation gate.
- Productivity and developer tools at 55–60% have the biggest activation upside — these are the ones where onboarding work and permission migration pay back fastest.
D7 retention by category
D7 = % of installers who fire at least one feature-use event on day 7 (the seventh calendar day after install). Standard retention definition.
The week-after drop-off is roughly proportional to how often the category's use case appears in a normal week.
Daily-driver tools have the cleanest D7 because their use-case (browsing the web) repeats every day. Productivity tools depend on whether the user's typical week contains the task they automate — clippers retain readers/researchers more than office workers; formatters retain devs but not designers.
D30 retention by category
D30 = % of installers active on day 30. This is the cohort number that most predicts long-term value.
D30 separates daily-utility habit (40%+) from week-to-week use (25–30%) from try-and-discard installs (under 20%).
Three diagnostic patterns:
- D30 / D7 ratio above 0.75 = habit formed. Users who survived the first week are sticky. Whatever you built in the first 7 days is doing its job.
- D30 / D7 ratio under 0.5 = strong week-2 churn. Activated users used the tool a few times and stopped. Often a single-session pattern from the uninstall diagnostic — the use case didn't repeat enough to form muscle memory.
- D30 above 35% for non-daily-driver categories is excellent — top decile. Don't chase higher; spend the effort on growth instead.
The retention curve (D1 → D90)
The shape matters as much as the value. A "smile curve" (steep drop then flat) means you have a power-user base under the noise; a straight decay means you don't. Typical curves per category:
Daily-driver tools form a near-flat tail after day 14 — the surviving users are habitual. Entertainment extensions don't have a tail; they have a slope.
A useful heuristic: compare D30 to D60. If the drop between them is <25% of D30, your power-user base is established and you can plan for predictable long-term revenue. If the drop is >40%, you're still leaking retained users who never quite committed.
Time-to-uninstall distribution
When uninstalls happen across the lifecycle, % of total uninstalls falling into each bucket. The pattern is remarkably consistent across categories — >60% of all uninstalls happen in the first week.
The 1–24h bucket is dominant. Most uninstalls are first-day decisions — either never-activated or first-impression failures.
Three things this distribution tells you:
- 43% of uninstalls happen on day 1. That's almost half your churn in the activation window. Per the onboarding guide, fixing the first 60 seconds disproportionately reduces lifetime uninstalls.
- The 7–30d bucket is single-session churn. Activated users who didn't form a habit. Diagnostic SQL in why users uninstall §5.
- >90 days is small (7%). Long-tail churn isn't where to invest — once users are past 90 days, they mostly stay until you ship a regression or a competitor launches.
Uninstall rate (30-day churn)
30-day uninstall rate = uninstalls divided by installs in the same rolling 30-day window. Same metric as the "Uninstall rate" KPI in the Crxlytics dashboard.
Lower is better. Daily-driver tools churn least because their use case re-establishes itself every browsing day.
Calibration note: an extension with an uninstall rate better than the band above means one of:
- You're measuring it wrong — the most common reason. Per CWS data accuracy, the SDK can silently drop
ext.uninstalledevents if the post-uninstall page orsendBeaconisn't set up. - Your install base is small enough that you're still riding a launch cohort — bring back the 90+ day cohort before drawing conclusions.
- You've genuinely done unusually good activation + retention work. Verify against D30 retention; they should move together.
Review rate among activated users
% of activated users who leave a Web Store review. This is the number that drives your CWS SEO review-velocity signal. Range assumes a well-timed in-product prompt per the growth playbook; without a prompt, multiply by ~0.2.
Daily-driver tools earn reviews through repeated value. Niche/entertainment users like the product but don't form the gratitude moment that drives a review.
The math compounds: a productivity extension with 60% activation × 11% review rate generates 6.6 reviews per 100 installs. The same extension with no prompt generates ~1.3. Five-fold lift from one product change.
A composite "growth health" score
Single-number summary that captures the four most informative benchmarks. Sum the points:
- Activation rate ≥ category median: 25 pts
- D7 retention ≥ category median: 25 pts
- D30/D7 ratio ≥ 0.75 (habit forming): 25 pts
- 30-day uninstall rate ≤ category median: 25 pts
Reading the score:
- 0–25 — activation problem dominates. Fix onboarding before anything else.
- 26–50 — activation OK, retention leaking. Look for single-session and never-formed-habit patterns.
- 51–75 — competitive. Focus on the lowest unscored quadrant.
- 76–100 — moat tier. Now go invest in growth (review rate, listing CVR, localization) — the retention base will multiply the gains.
Most extensions land between 25 and 50 on first measurement. Moving to 50–75 is achievable in a quarter with focused work per the recipes linked above. 75+ takes a year of compound.
FAQ
Where do these benchmarks come from?
Aggregate ranges derived from observed patterns across customer extensions and public Web Store data, with the category split applied retroactively. Ranges, not absolute numbers, because cross-extension variance is meaningful and any single number would be misleading.
My category isn't in the five buckets — what do I do?
Pick the bucket your use case patterns most resemble: how often does the user encounter the moment your extension solves? Once per session → daily-driver; once per week → productivity; once per dev day → developer; once per month or per workflow → niche; rarely with strong recall → entertainment.
How are activation events defined per category?
Per the DAU/MAU guide: a translator's activation is translation.run; a clipper's is clip.saved; a blocker's is block.matched; a dev tool's is panel.opened. Pick the smallest event set that, if all are zero for a user-day, you'd honestly say they didn't use the extension.
D7 retention drops below D14 for me — is that wrong?
Almost certainly an instrumentation bug. Retention is monotonically non-increasing — D14 can never be higher than D7 from the same cohort. If you see this, you're likely mixing cohorts or counting a user who returned after a gap as "retained" on the later day but not the earlier. Check the cohort definition.
My D1 is >90%. Is that real?
Probably overcounted. Either you're counting ext.installed as engagement (it shouldn't be — exclude it from the activation event set, see the DAU/MAU pitfalls), or your activation event fires on script injection rather than on real interaction.
What benchmark should I use for paid users?
Paid users typically retain 2–3× their free-tier counterparts on D30 and beyond. If your paid D30 is below 50%, the upgrade path is selling something that doesn't fit the retained-user need. Look at the activation event distribution of paid users vs. free — they're usually different.
How often should I re-measure these?
Monthly is enough — the underlying patterns move slowly. Look weekly if you're running a specific experiment (onboarding change, permission migration, listing redesign); otherwise the noise overwhelms the trend.