Feature Adoption Metrics: Breadth, Depth, Time-to-Adopt
Feature adoption is how much a specific feature gets used, and one number can't capture it. You need three: breadth (what share of users touch the feature), depth (how often the ones who touch it come back), and time-to-adopt (how long from first exposure to real use). A feature can score high on one and fail the other two, which is exactly why teams argue about whether a feature is "working."
The short version: measure all three, plot breadth against depth, and let the quadrant tell you whether to promote the feature or deepen it.
The three dimensions, with formulas
Most dashboards report a single "feature adoption rate" and stop. That's fine as a headline, but it hides the shape of the thing. Here's each dimension on its own.
Breadth of adoption is the percentage of your active users who use the feature at all in a window.
Breadth = (users who used the feature) / (active users) × 100
If 2,000 of your 8,000 weekly-active users open the new export panel, breadth is 25%. Simple, and it's the number most people mean when they say "adoption."
Depth of adoption measures how intensively the adopters engage once they're in. Two common shapes:
Depth (frequency) = feature events / adopters
Depth (stickiness) = feature DAU / feature MAU
If 200 users generated 3,000 interactions with a dashboard filter last month, average depth is 15 interactions per user. A feature can have thin breadth and enormous depth (a power-user tool that 8% of people live inside) or wide breadth and paper-thin depth (everybody clicks it once, nobody returns).
Time-to-adopt is the lag between first exposure and the first meaningful use. How quickly the feature "lands."
Time-to-adopt = timestamp(first real use) − timestamp(first exposure)
Report it as a median, not a mean, because a handful of users who adopt six weeks late will drag an average into nonsense. Median time-to-adopt tells you whether the feature is discoverable and self-explanatory or whether people bounce off it and only come back after a nudge.
Benchmark bands (and a warning about them)
Numbers help, so here are real ones. Pendo's feature-adoption benchmarking puts the median feature adoption rate at 6.4%, with top-10% products hitting 15.6%, roughly 2.5x the median. For enterprise digital products specifically, Pendo reports an average of 6% and a best-in-class figure of 16%. Media products sit lower, around 4.9%.
Those numbers look brutally low until you realize they're computed across every feature a product ships, including the long tail nobody was ever going to use. When you narrow to core features, the ones onboarding actually pushes people toward, the bands rise sharply. Artisan Growth Strategies' 2025 benchmark review pegs average core-feature adoption at 24.5%, with the top quartile above 45%.
So which benchmark is "right"? Honestly, the industry disagrees here, and I'll pick a side: the 6.4%-style figures are the more useful gut-check for a whole product, because they include the features you keep forgetting exist. If you only benchmark your three hero features, you'll feel great and never notice that 40% of your codebase is dead weight.
| Dimension | Weak | Okay | Strong |
|---|---|---|---|
| Breadth (single feature, core) | < 15% | 15–35% | 35%+ |
| Breadth (any feature, whole catalog) | < 4% | 4–10% | 10%+ |
| Depth (feature DAU/MAU stickiness) | < 15% | 15–30% | 30%+ |
| Time-to-adopt (median, from exposure) | > 14 days | 3–14 days | < 3 days |
Treat these as directional. Your category, your pricing, and how aggressively you onboard will move every row. A compliance feature that gets used once a quarter isn't broken because its stickiness is 5%. It's just quarterly.
The breadth-vs-depth 2x2
This is the part I actually use in reviews. Plot breadth on one axis, depth on the other, and every feature lands in one of four quadrants. Each quadrant is a different decision.
| Low depth | High depth | |
|---|---|---|
| High breadth | Broad but shallow: everyone tries it, nobody stays | Core hit: protect and expand |
| Low breadth | Struggling: few try it, few return | Niche-loved: small crowd, obsessed |
The two interesting corners are the off-diagonal ones.
Broad-but-shallow (high breadth, low depth) is the trap that looks like a win. A big adoption number, and a churn-out-of-the-feature you won't see unless you measure depth separately. The move here is almost never "promote it harder," because it already has attention. The move is to fix the second use: figure out why people don't come back, because pouring more traffic onto a leaky feature just wastes exposure.
Niche-loved (low breadth, high depth) is the opposite mistake waiting to happen. A small group depends on this feature completely. The temptation is to kill it for low breadth. Don't, at least not before you check who those users are. Niche-loved features are often where your highest-value accounts live. The move is promotion: this thing works, more people just don't know it exists.
That's the whole point of splitting breadth from depth. One number would've told you the broad-shallow feature is thriving and the niche-loved feature is failing. Both readings are wrong.
The adoption funnel: exposed → tried → retained
Breadth, depth, and time all fall out of one underlying object: the feature adoption funnel. It has three stages that matter (some frameworks split them finer, but three is enough to act on).
- Exposed. The user encountered the feature. They saw the button, got the tooltip, landed on the panel. Exposure is the denominator people forget; if only 10% of users ever see a feature, low adoption isn't an adoption problem, it's a discovery problem.
- Tried. The user took a first meaningful action inside the feature. Not a hover, not an accidental click. The actual thing the feature does.
- Retained. The user came back and did it again, days or weeks later. This is where "adoption" actually lives. Appcues frames the same shape as exposed, activated, used, and used-again, and the value is identical: funnel analysis shows you exactly which step is bleeding.
Userpilot's diagnostic guide makes the same case. Feature adoption rate is the share of active users engaging a feature, and mapping the path from exposure to consistent use tells you where the drop-off is. Breadth is the exposed→tried conversion times exposure. Depth is what happens after retained. Time-to-adopt is the clock on exposed→tried. Same object, three angles.
If your funnel building needs a home for these features in the wider metric system, they usually hang off an engagement branch of your north star metric tree rather than sitting as standalone vanity counts.
A worked example (12 users, 3 weeks)
Small datasets make this concrete, so here's a real-shaped one. You ship a "saved filters" feature and watch 12 beta users for three weeks.
| User | Exposed (day) | First use (day) | Uses in wk 3 |
|---|---|---|---|
| 1 | 1 | 1 | 6 |
| 2 | 1 | 2 | 4 |
| 3 | 1 | 9 | 1 |
| 4 | 2 | n/a | 0 |
| 5 | 2 | 3 | 5 |
| 6 | 3 | n/a | 0 |
| 7 | 3 | 4 | 0 |
| 8 | 4 | 5 | 3 |
| 9 | 4 | n/a | 0 |
| 10 | 5 | 6 | 2 |
| 11 | 6 | 18 | 1 |
| 12 | 6 | n/a | 0 |
Now the three numbers.
Breadth: 8 of 12 users tried it, so breadth is 67%. Looks great for a beta.
Depth: of those 8, only 6 used it again in week 3, and the counts are lopsided. Two users (1 and 5) account for 11 of the 22 week-3 uses. Average depth is 2.75 uses per adopter, but the median is 2, and the distribution screams "a couple of people love this, the rest tried it and drifted."
Time-to-adopt: the gaps between exposure and first use are 0, 1, 8, 1, 1, 1, 1, 13 days for the eight who adopted. Median is 1 day. But two users took 8 and 13 days. Those are your late adopters, and they barely used it afterward (1 use each). Late adoption predicted shallow depth here, which is a pattern worth watching at real scale.
The one-number version of this feature is "67% adoption, ship it." The three-number version is "strong breadth, thin and top-heavy depth, and a bimodal time-to-adopt where slow adopters don't stick." That second reading tells you to work on the second-use experience before you roll out wider. It's the difference between a review that congratulates you and a review that tells you what to do.
Where this goes wrong
A few failure modes I've watched teams walk into, more than once.
Counting exposure as adoption. If your event fires when the feature renders rather than when someone acts, your breadth number is fiction. A panel that mounts on every page load will show 100% "adoption" and mean nothing. Instrument the meaningful action, not the impression.
Averaging time-to-adopt. I said it above and I'll say it again because everyone does it anyway. One user who adopts eight weeks late turns a 2-day median into a 9-day mean, and now your dashboard says the feature is slow to land when 90% of people got it on day one. Use the median. Always.
Benchmarking against the wrong catalog. Comparing your core-feature breadth (25%) against the whole-product 6.4% median makes you feel like a genius. Comparing a quarterly compliance tool against a daily-use stickiness band makes you feel like a failure. Match the benchmark to the feature's job.
Killing niche-loved features on breadth alone. The low-breadth column of the 2x2 has two very different tenants. One is genuinely failing. The other is quietly load-bearing for your best accounts. Check depth and check who before you cut.
There's a broader version of this mistake, which is measuring only the features you're proud of. Artisan's review makes the point that breadth of adoption across features predicts retention better than depth on any single one, because users who adopt more features stick around longer because they've wired more of their workflow to you. So the whole-catalog breadth number, ugly as it is, is often the one that correlates with the metric you actually care about.
Tooling: computing per-feature funnels
None of this needs anything exotic. Any product analytics platform that supports funnels over your own events can compute exposed→tried→retained per feature, then slice breadth and depth from the same data. Amplitude and Mixpanel do it with event funnels and stickiness reports; Pendo and Userpilot bundle it with in-app guidance so you can act on the drop-off; Kixo lets you build the same per-feature funnels by asking for them in plain language and gets a visible reasoning trail with the chart. The honest trade-off is the usual one: guidance-first tools make the "so now nudge these users" step easier but lock you into their onboarding layer, while pure-analytics tools give you cleaner data and leave the intervention to you.
Pick based on whether your bottleneck is seeing the funnel or acting on it. If you already know exposure is your weak step, a tool with in-app messaging pays off. If you don't yet trust your numbers, get the measurement clean first and worry about nudges later.
FAQ
What's a good feature adoption rate? Depends on the denominator. Across a whole product catalog, the median is about 6.4% and top performers hit 15.6% (Pendo). For a core feature that onboarding pushes people toward, expect 25%+ and treat 45% as excellent (Artisan Growth Strategies).
Is breadth or depth more important? Neither, alone. But if you're forced to pick a single retention predictor, breadth across features tends to win, since people who adopt more of your product churn less. Depth matters most for judging one specific feature's health.
How do I measure time-to-adopt? Median gap between a user's first exposure event and their first meaningful-use event. Report the median, segment by acquisition source, and watch for late adopters who also show shallow depth.
Should I delete a feature with low adoption? Not on breadth alone. Check depth and check which users depend on it. Low breadth plus high depth is a niche-loved feature, usually a promotion problem and not a deletion candidate.