The Power User Curve: Measuring SaaS Engagement Depth

The power user curve is a histogram of engagement frequency — for each user, how many days they were active over a trailing window, usually the last 28 days (hence "L28"). It exists to show you the one thing an average cannot: the shape of engagement across your base. As Andrew Chen describes it, the curve is "the best way to understand your most engaged users," because it reveals whether you have a committed core returning almost daily or a base that is broad but shallow. Two SaaS products can post the same DAU/MAU and have entirely different curves — and entirely different futures.

That is the case for using it. A single engagement number, however carefully chosen, throws away the distribution behind it, and the distribution is where the strategy lives. A product with a heavy right tail of near-daily users has found a habit and can build on it; a product whose curve slides to nothing has an occasional-use tool and needs a different retention playbook. This post covers how to build the curve, how to read its three canonical shapes, what it tells you that ratio metrics cannot, and how to instrument it without a warehouse project.

What the curve actually plots

Build it by counting, for every user, the number of distinct days they performed a chosen action within the trailing window, then bucketing users by that count. The x-axis runs from 1 day to the full window length; the y-axis is the number or percentage of users in each bucket. A user active on 3 of the last 28 days lands in the "3" bucket; a user active on 27 of 28 lands near the far right. The result is a frequency distribution of engagement — the same underlying data your DAU/MAU stickiness ratio summarises into one figure, but shown in full.

The window length is a real choice. Twenty-eight days is favoured over thirty precisely because it is four clean weeks, so weekly rhythms — the fact that a B2B tool is quiet on weekends — do not distort the buckets. Use L7 for products people are meant to touch daily and where a week is a fair test of habit; use L28 or L30 for products with a weekly or monthly natural cadence. The point is to match the window to your product's expected rhythm so the curve measures habit formation rather than the calendar.

Logins lie; value actions tell the truth

The single most important instrumentation decision is which action defines "active." A curve built on logins or app opens can look robust while telling you almost nothing, because a login is not value — users open apps out of habit, from a notification, or to check a thing and bounce. Build the curve on the action that represents core value delivered: the query run, the message sent, the report published, the deploy shipped. That ties engagement frequency to value, which is the quantity you actually want to grow, and it aligns the curve with the same value logic behind your aha-moment definition.

This is the frequency counterpart to the activation question. Activation asks whether a user reached value once; the power user curve asks how often they keep returning to it. Using a value action for both keeps the two metrics telling a consistent story — a user who activated on a real value event and then shows up in the right tail of the value-action curve is the genuine article, not someone padding a login count. If you track a login-based curve as well, treat it as a vanity reference and make decisions from the value-action version.

Reading the three canonical shapes

The healthiest shape for a habitual product is the smile: a tall left side (many light, occasional users), a dip through the middle, and a clear rise on the far right (a core active most days of the window). The right-side bump is the whole point — it is direct evidence that a real segment has made your product a habit. Most durable consumer and daily-use B2B products show this barbell, because engagement is rarely uniform: a broad base contributes lightly while a committed minority carries the load. If your smile is growing its right side over time, your product is deepening.

The two unhealthy shapes are the slide and the spike. A slide falls monotonically from left to right with no terminal rise: engagement decays with frequency and virtually no one is active most days — the mark of a product that is occasionally useful but has no daily hook. A spike concentrated on the far left (active just one or two days) with nothing behind it is worse still, signalling that users try the product and never build frequency. Both shapes say the same thing DAU/MAU would only whisper: there is no committed core yet, so retention work should focus on creating a reason to return, not on acquiring more top-of-funnel.

What the curve reveals that ratios hide

DAU/MAU answers "how sticky is the average user," but averages conceal bimodality, and engagement is almost always bimodal. Two products at a 20% stickiness ratio can differ completely: one has a fat middle of moderately engaged users, the other a barbell of daily power users and one-time tourists. The strategic implication diverges sharply — the first should push moderate users toward daily habit, the second should figure out why so many churn after one visit while a core thrives. Only the full histogram exposes which situation you are in.

The curve also changes how you read retention. A cohort retention chart tells you how many users come back; the power user curve tells you how intensely the ones who come back engage. Reading them together — cohort analysis for breadth of return and the power user curve for depth of engagement — gives a two-dimensional picture that either alone flattens. A cohort that retains well but sits entirely in the low-frequency buckets is fragile; one that retains modestly but concentrates in the right tail has found real product love in a niche worth expanding.

Instrumenting it without a data project

You do not need a warehouse to produce a power user curve. The raw material is an event stream with a user identifier, an event name, and a timestamp — which any competent product analytics setup already has. Count distinct active days per user over the trailing window for your chosen value action, bucket the counts, and plot. Refresh it weekly and, critically, watch it move over time rather than reading a single snapshot: a right tail that thickens month over month is the clearest evidence that your engagement is deepening, and a thinning right tail is an early warning that ratio metrics will confirm only later.

Once it is running, connect it to the rest of your measurement so it is not an island. The share of users in the right-tail buckets makes a strong candidate engagement input beneath your north-star metric, and the curve pairs naturally with your weekly retention benchmarks — retention tells you who returns, the curve tells you how hard they lean in. Track both and you stop mistaking a broad, shallow base for a healthy one, and start seeing exactly where your product has earned a place in someone's day.