Try it on your own data
Every report in this article is in the free plan. One snippet, cookieless, up to 10 sites.
Start freeAARRR is twenty years old and still the most useful growth model, not because the five letters are clever but because they force a sequence. You cannot retain people you did not activate, and you cannot be referred by people you did not retain. That sequence is also the trap: most teams spend their energy at the top, on acquisition, because it is the easiest to buy. This article shows how to measure each stage in TapCub, how to compare the stages against each other, and how to pick the one stage where a single experiment moves the whole model.
What you will learn
- A concrete event definition for each of the five stages
- How to compare stages against your own last quarter instead of someone else’s benchmark
- Why retention is usually the first lever, and the one experiment per stage we recommend
What AARRR actually asks
Each letter is a question with a yes or no answer per person. Acquisition: did they arrive? Activation: did they do the one thing that shows they understood the product? Retention: did they come back? Revenue: did they pay? Referral: did they bring someone? The model works only if each question has a precise event behind it. “Activation” as a feeling is useless; “created a first site and saw a pageview within 24 hours” is a metric.
Write the five events down before you open a report. In TapCub the growth view on the insights page takes exactly these five definitions and draws the stage-to-stage rates for you; the hard part is agreeing on the definitions, not the chart.
Put a number on every stage
Once the events exist, each stage becomes a rate: the share of people from the previous stage who reached this one, in a fixed window. Below is the sample workspace we use in the product. Acquisition is the base; the other four are conversion rates from the stage before. The window is 30 days from first visit for activation and revenue, and “returned in week 2” for retention.
Acquired · 30 d
48,213
▲ 12.4%
Activated
38.1%
▲ 0.6 pts
Retained · week 2
24.3%
▼ 4.9 pts
Stage-to-stage conversion · this quarter vs. last
Humans onlyCompare against yourself, not a benchmark
Industry benchmarks are tempting and mostly useless: they average businesses that look nothing like yours. The comparison that works is against your own last quarter and against the stage next to it. A stage that fell while its neighbors held is the first suspect. A stage that has never moved is the second, because nobody has tried.
In the sample, activation held at 38%, retention fell from 29% to 24%, and revenue stayed flat. Revenue is downstream of retention, so a fix there inherits the retention loss. Retention is the lever.
Two cautions when you read the rates. First, make sure every stage is counted on the same population; a retention rate computed on all signups and a revenue rate computed on activated users are not comparable. Second, check seasonality before you call a drop a problem. If the quarter you are comparing against included a launch or a holiday, compare against the same quarter last year as well. The growth view keeps the previous period on the chart for exactly this reason.
Why retention is usually first
Retention multiplies everything after it. A 10% improvement in retention lifts revenue and referral by roughly 10% each with no other change, because more people are around to pay and to tell someone. The same 10% at acquisition lifts every stage too, but you have to pay for it every month. Retention improvements are bought once.
Retention is also the stage with the richest diagnostics. A retention matrix by signup week shows whether a change in week 12 actually moved the curve for later cohorts. The retention calculator will tell you how much a one-point change is worth over a year, and the user analytics view shows which segments retain and which do not.
One experiment per stage
Each stage has a natural first experiment. Pick the stage, run its experiment for a full cycle, and read the stage rate and the downstream rates before you judge it. The table lists the one we recommend starting with, the metric that judges it, and the common way it goes wrong.
| Stage | First experiment | Judge by | Common failure |
|---|---|---|---|
| Acquisition | Move budget from the lowest-ROI channel to the highest | Activated users per dollar, not clicks | Optimising for cheap clicks that never activate |
| Activation | Shorten the path to the first success event | Share activated within 24 h | Adding a tutorial instead of removing steps |
| Retention | A triggered message when a user stops after activation | Week-2 return rate by cohort | Sending to everyone, not to the ones who stopped |
| Revenue | Show the upgrade at the moment a limit is hit | Paid conversion within 30 days | Showing pricing before the user has felt the limit |
| Referral | Ask right after a success event | Invites sent per retained user | Asking at signup, before there is anything to recommend |
Start from a template, not a blank chart
Every industry has a different activation event and a different retention window. An e-commerce store activates on first add-to-cart and retains on a second order within 60 days; a SaaS product activates on first project and retains weekly. TapCub ships 10 industry templates with these definitions pre-filled, so the AARRR view is populated the day the snippet goes in. Pick the solution closest to your business, then edit the definitions where yours differ.
Finally, write down the one experiment you chose, the stage rate you expect it to move and by how much, and the date you will read the result. A model is only useful if it produces a decision, and a decision is only useful if someone checks whether it worked.
TapCub Team
The people who design, build and support TapCub. We write about what we measure on our own site and what customers ask us most.