View mode Color
Demo, synthetic data
Atreides: Product Analytics
Atreides is a fictional B2B work-management SaaS simulated from launch in July 2022 through today: about ... monthly active users on ... of recurring revenue, zero real companies or persons. Every pattern was planted on purpose, documented, and recovered by a validation gate before this page shipped (receipts at the bottom).
Data failed to load. Refresh to retry.
...MRR
...NRR (ttm)
...MAU
...DAU/MAU
...Month-3 retention
...Logo churn (ttm)
From
To
Plan group
Segment
Channel
Active users over time
Daily active, trailing-7-day, and trailing-28-day users. Switch the active definition to see how much the headline depends on it. Hover or tap a day for the breakdown.
Active =
View SQL
Growth accounting
New and resurrected actives above the line, churned below, net as a line. Quick ratio = (new + resurrected) / churned. Hover or tap a month.
View SQL
Stickiness by plan
DAU/MAU each month, one line per plan group. Higher plans run tighter daily routines. Segment, channel, and date filters apply; the plan filter highlights its line. Hover or tap a month.
View SQL
Metric definitions
- MRR
- Monthly recurring revenue: the sum of paying accounts' current monthly subscription value.
- NRR (ttm)
- Net revenue retention: take the accounts paying one year ago, sum their MRR today, divide by their MRR then. Above 100% means expansion outran churn and contraction.
- MAU
- Distinct users active in the latest complete calendar month, under the active definition selected on the trend.
- DAU/MAU stickiness
- Mean daily active users in a month divided by that month's active users: the fraction of the monthly base that shows up on an average day.
- Month-3 retention
- Of a signup cohort, the share active (any activity) in the third calendar month after signup. Shown for the latest cohort old enough to measure.
- Logo churn (ttm)
- Paying accounts that fully churned in a month over accounts paying at the prior month end, averaged across the trailing 12 months.
- Growth accounting
- Each month every active user is new (first ever active), resurrected (back after a gap), retained (active last month too), or churned (active last month, gone this month). MAU(t) = MAU(t-1) + new + resurrected - churned.
- Quick ratio
- (new + resurrected) divided by churned. Above 1 means more users gained than lost.
- DAU / WAU / MAU
- Distinct users active on a day, in the trailing 7 days, and in the trailing 28 days.
- Active: any activity
- The user had any session in the window. The loosest, most flattering count.
- Active: core action
- The user did the core job: at least one task or doc.
- Active: qualified
- Engaged use: at least 3 events across at least 2 different features.
How this was built
A seeded generator simulates Atreides day by day and writes plain CSVs (accounts, users, milestones, the subscription ledger, and per-user daily activity). A DuckDB transform aggregates them into one JSON file this page fetches, pre-rolled to a small set of plan-by-segment-by-channel cells. The build refuses to ship unless every planted effect is recovered within its documented tolerance AND the cell sums reproduce those numbers, so the figures above are checked, not just plausible.
Validation report: planted vs recovered
View SQL (full transform)
Cohort retention curves
One line per signup-quarter cohort: the share still active (any activity) by months since signup, over the first year. Cohorts that signed up after the March 2025 onboarding launch are in color, earlier cohorts in gray, and young quarters run only as far as every signup has fully reached. Hover a line for its numbers; click it to trace that cohort against the average below.
Cohorts
View SQL
Cohort vs the average
The pooled average retention curve for the selected pool, with any cohort line you click above traced on top of it. The average at each month age pools every cohort old enough to have reached that age; it stops once fewer than 50 users back a point.
View SQL
Retention by plan at signup
Pooled curves by the plan group the user signed up into, first 24 months. One hue: a fuller line is a higher tier.
View SQL
Retention by channel
The same pooled curves cut by acquisition channel, first 24 months. Sales-led accounts onboard whole teams; self-serve signups arrive alone.
View SQL
The onboarding launch, in retention
Paid signup cohorts from the year before the March 2025 onboarding launch against the year after, traced over the months both have fully reached. Same pooling, same definition; the only thing that changed between the two lines is the launch.
View SQL
How these numbers are computed
- Cohort
- Every user who signed up in a given calendar month. Pool = the account's plan group on the user's signup date: paid means team, business, or enterprise; free means the account had not converted yet.
- Retention at month N
- The share of the cohort active (any activity) in the Nth calendar month after the signup month. Month 0 is the signup month itself.
- Mature ages only
- A cohort shows only calendar months that have fully elapsed at the data cutoff. No projections, no partial months, no future cells.
- Small-n suppression
- Cohorts under 50 users show an explicit marker instead of percentages: user-level rates on a handful of people are noise. They still feed the pooled curves at their honest (tiny) weight.
- Pooled curve
- At each month age, the sum of active users divided by the sum of cohort sizes, over every cohort old enough to have reached that age. Computed in the pipeline; the browser only divides the shipped integers.
- Launch comparison
- Paid cohorts from March 2024 to February 2025 (before) vs March 2025 to February 2026 (after). Both windows are fully mature through month 3, where the gated lift is measured.
Activity depth, last 28 days
How many of the last 28 days each active user touched Atreides. Drag the threshold to set your own power-user bar; the readout reads build-time cumulative shares.
Plan
View SQL
Event concentration
Users sorted by how much they do; the curve shows the share of May 2026 events generated by the bottom share of users. The dashed diagonal would be perfectly even usage.
View SQL
Feature adoption vs intensity
Each feature: the share of paying accounts using it (5+ events in the last 28 days) against how hard adopters use it (median weekly events). Quadrants split at half of accounts and 10 events a week.
View SQL
Feature breadth per account
Distinct features (of 7) each paying account touched in the last 28 days. Zero means a paying account with no activity at all in the window.
View SQL
Activation funnel
Signups advancing through nested milestones: completed setup within 14 days, plus a teammate invite within 7 days, plus a week-4 habit (active in 3 of the first 4 weeks). Cohorts through Feb 2026, so every stage is fully observable. Hover or tap a bar for the drop-off.
Signup quarter
View SQL
Time to the aha moment
Days from signup to the first teammate invite, among users who ever send one. Bright bars fall inside the 7-day aha window used everywhere else on this tab.
View SQL
Aha explorer: does the action cause retention?
Pick a candidate aha action and compare two honest numbers, the same play as the
active-user definition toggle. Raw lift answers "how different are users who did this":
month-3 retention of exposed vs unexposed signups. Stratified lift answers "what does
the action itself buy": the same comparison made only within matching plan-at-signup and
signup-quarter cells, then averaged. When adoption concentrates in plans that retain
anyway, the raw number inherits that selection effect and the stratified one strips it.
Action
Raw lift ...
Stratified lift ...
Who takes the action: adoption by plan at signup
View SQL
Activation rate over time
Share of each signup month completing setup within 14 days. The dashed step marks the March 2025 onboarding revamp; reference lines show the pooled rate over the year before and the year after. Hover or tap a month.
View SQL
How these numbers are computed
- Activation
- Completed onboarding setup within 14 days of signup.
- Week-4 habit
- Active on at least 3 of the first 4 weeks after signup.
- Aha window
- 7 days for the teammate invite, 14 days for dashboards and integrations: the action must happen early to count as exposure.
- Raw lift
- Month-3 retention of exposed signups divided by unexposed. Honest about how different the two groups are; silent about why.
- Stratified lift
- The same ratio computed inside each plan-at-signup by signup-quarter cell with at least 25 users on both sides, then averaged weighted by exposed count. Computed with numpy in the pipeline and shipped as a number; the browser never re-derives it.
- Selection effect
- When the users who take an action would have retained anyway (enterprise plans adopt integrations 20x more than free), the raw lift credits the action with their retention. Stratification removes the part explained by plan and cohort mix.
- Cohort universe
- Signups through Feb 2026, the latest cohort whose third month is fully observable at the data cutoff.
MRR waterfall
Monthly recurring revenue gained (new, expansion, reactivation) above the line and lost (contraction, churn) below it, with ending MRR on the right axis. The walk is penny-exact: every month's ending balance equals the prior balance plus these five flows. Hover or tap a month for the full bridge.
View SQL
Net and gross revenue retention
Accounts paying a year ago: their MRR today over their MRR then (net), and the same with growth capped so only losses count (gross). Grouped by the plan each account was on at the cohort start. Hover or tap a bar for the cohort dollars.
View SQL
Logo vs revenue churn by plan
Monthly account churn (logo) against the share of MRR it took (revenue), trailing 24 months, one panel per plan on a shared scale. Hover or tap a month.
View SQL
Churn early warning
View SQL
At-risk watchlist
View SQL
How these numbers are computed
- MRR waterfall
- Signed subscription-ledger events summed per calendar month: new = first conversion, expansion = seat adds or upgrades, reactivation = churned accounts returning, contraction = seat drops or downgrades, churn = the full remaining balance. Ending MRR is the running ledger balance at each month end.
- NRR (ttm)
- Take every account paying on 2025-05-31; divide the cohort's MRR on 2026-05-31 by its MRR then. Above 100% means expansion outran churn and contraction. Plans group by the plan at the cohort start.
- GRR (ttm)
- The same cohort with each account capped at its starting MRR, so growth cannot hide losses. GRR can never exceed 100%.
- Logo churn
- Churn events in a month over accounts paying at the prior month end.
- Revenue churn
- MRR lost to churn events in a month over MRR at the prior month end.
- Trailing baseline
- An account's mean weekly sessions over weeks -16 to -9 relative to the week being judged: its own normal, measured before any decline starts.
- Early-warning rule
- On the 2-week smoothed weekly series: 4 consecutive declining values AND current usage under 75% of the trailing baseline. Evaluated only for accounts with 26+ weeks of paid tenure and a baseline of at least 20 sessions a week (percent rules need a volume floor).
- Flag lift
- Of the weeks the rule flags, the share where the account churned within 8 weeks, divided by the same share over all eligible account-weeks. Weeks too close to the data cutoff to observe an outcome are excluded from both sides.
- Watchlist
- Accounts where the rule fired this week or last and usage is still below 75% of baseline, steepest decline first, capped at 15 rows. Decline % compares current smoothed usage to the trailing baseline.
What this means for your business
Same suite, your product
Atreides is fictional; the method is not. Pointed at your real product events and billing data, the same pipeline ships a validated engagement mart and an interactive suite your team can actually explore, with every number backed by SQL you can audit.