AARRR tells you retention is falling. It cannot tell you why.
Your dashboard is doing its job. Acquisition is holding, activation is fine, and retention has been sliding for two quarters. The number is red. Everyone in the room can see it is red.
Now what?
Someone will say "we need to improve retention," which is not a plan, it is the metric said out loud in a different tone of voice. Someone will propose a win-back campaign, someone else an onboarding revamp, and the strongest opinion in the room will win, because nothing in your instrumentation can settle the argument.
That is the gap. The funnel told you where the number moved. It has no way to tell you why, and you cannot fix a where.
Credit first: AARRR is very good at what it is
Dave McClure published Startup Metrics for Pirates in 2007: Acquisition, Activation, Retention, Referral, Revenue. It gave a generation of founders a shared vocabulary, it made growth legible to people who had never run a marketing function, and it is still the fastest way to instrument a business from scratch. I would not want to run a company without it.
Notice what he called it. Startup metrics. Not startup strategy.
That distinction is the whole story, and the failure is not McClure's. AARRR is a measurement framework that got promoted to a strategy framework while nobody was looking. An entire discipline grew up around optimising the five letters, and growth hacking inherited an assumption it never examined: that the product, the positioning, and the proposition are already right, and the remaining work is throughput.
Optimise the pipes without checking what is flowing through them, and you will get very efficient at moving the wrong thing.
Five letters, three levers, one equation
Start by mapping AARRR onto the arithmetic, because the funnel is not wrong, it is just incomplete in a way that only becomes visible when you write the maths down.
Revenue = AOP × NT × ATV × 12. Your average active base across the year, times how often they buy, times what they spend, times twelve. And that base only moves three ways: EOP = BOP + GA − CHURN. You start with a base, you add customers, you lose customers.
This is where most strategy tools stop short. The Business Model Canvas maps revenue as a flat block, a box labelled Revenue Streams that you fill in and move on from. AARRR does better, giving you five stages to watch. But an actionable marketing strategy needs neither a box nor a sequence. It needs an active arithmetic equation, because only arithmetic tells you which term dominates.
So the five letters collapse:
The funnel is not wrong. It is incomplete in a way that only becomes visible when you write the maths down.
Revenue = AOP × NT × ATV × 12
EOP = BOP + GA − CHURN
Your average active base across the year, times how often they buy, times what they spend, times twelve. The base moves three ways only: customers you add, customers you lose, and what each one is worth.
| AARRR stage | In the equation | The lever |
|---|---|---|
| Acquisition | GA (gross adds) | GET |
| Activation | Whether a gross add ever becomes a real customer | GET / KEEP boundary |
| Retention | CHURN | KEEP |
| Referral | GA at a lower cost of acquisition | GET |
| Revenue | NT × ATV | GROW |
Five letters. Three levers. One equation. The funnel lays five stages side by side as equal concerns. The equation shows they are terms in a product, and terms in a product are not equal. Some dominate.
Now look at what the equation says that the funnel cannot. AARRR presents five stages as a sequence of roughly equal concerns, laid left to right. The equation says they are terms in a product, and terms in a product are not equal: some dominate. If CHURN is running hot, no realistic increase in GA can compensate, and that is arithmetic, not opinion. The funnel shows you acquisition and retention side by side with no exchange rate between them. The equation gives you the exchange rate, and the exchange rate is what tells you which lever carries your year.
That alone reorders most growth roadmaps I see.
The thermometer problem
Here is the sharper limit, and it is structural.
Retention is falling. AARRR has told you everything it is capable of telling you: there is a leak. It is a thermometer. It reports a fever. It cannot tell you whether you have flu or sepsis, and the treatments are not remotely the same.
So take the same red number to the Marketing Canvas Method, which breaks marketing into 24 dimensions, the specific parts of your strategy you can score and fix, each scored from −3 to +3. Retention has at least six plausible causes, and they live in different meta-categories:
110 Job To Be Done at −2. You are serving a job the customer does not actually have. They leave because they hired you for the wrong task and eventually noticed.
120 Aspirations at −2. The product works, but it does not move them toward the person they are trying to become. They stay, unmoved, then go.
130 Pains & Gains at −2. You solved a pain they did not have, and the one they did have is still there.
420 Experience at −2. The service broke after the sale. They believed you, they bought, and you let them down.
330 Pricing at −2. The value is real but the price is misaligned to it, and every renewal is a fresh negotiation with themselves.
140 Engagement at −2. They disengaged months before they churned, and nobody was watching the leading indicator.
Six causes. One metric. Every one of those produces the same falling retention line, and every one demands a different fix, a different budget, and a different team. A −2 in Experience is an operations problem. A −2 in Job To Be Done is an existential one. Your onboarding revamp fixes exactly one of the six, and you have a one-in-six chance of having picked it.
You cannot allocate budget against a symptom. That is the entire reason a diagnostic layer has to exist beneath the dashboard.
What growth hacking assumes without saying so
Now the systemic point, which is bigger than any single metric.
Funnel optimisation is a multiplier. It takes what you have and moves it through faster. Which means it silently assumes the sign is positive: that your positioning is clear, your proposition believed, your job correctly named. Grind the funnel with 220 Positioning at −2, where three competitors say your sentence and buyers cannot tell you apart, and you will fight for a five percent conversion lift while a fifty percent problem sits upstream, invisible to every letter in AARRR.
The method has a name for that upstream problem. Some dimensions are Fatal Brakes: weaknesses severe enough to cancel the value of everything you are doing right. And it has a rule: the No-Scale Gate. You do not fund a growth bet while a Fatal Brake sits unfixed. Not for prudence. Because the arithmetic cancels the bet before the money leaves the account.
AARRR has no equivalent. It cannot, because there is no field on a funnel where a number goes to say nothing downstream of me matters until I am fixed.
Airbnb deleted $800 million from the top of its funnel
If you want the argument in one company, take the most celebrated growth-hacking story ever told.
In 2020, with bookings down 72% in eight weeks, Airbnb cut roughly $800 million of marketing spend. On a pure AARRR reading, that is a company setting fire to its own acquisition engine at the exact moment it needs volume. Every letter should have collapsed.
By 2022 it had the most profitable year in its history.
Read it through the method and the reason is not mysterious. Airbnb's scored strengths were never the funnel: Media (530) at +2, and cutting the spend is what proved it, because the organic demand was the engine all along. Engagement (140) at +2, the host community. User Lifetime (630) at +2. Identity (240) at +2, "Belong Anywhere." And the Fatal Brake for its archetype, Experience (420), was moved from +1 to +2 during the crisis: the refund mess handled, cleaning protocols written, the Airbnb Clean standard shipped.
Airbnb did not out-optimise its funnel. It fixed its Fatal Brake, leaned on brand and community, and let the funnel take care of itself. The $800 million was not the engine. It was a tax the company had been paying because nobody had scored the difference.
That is the most expensive lesson available in growth marketing, and it is invisible to a framework whose five letters are all downstream of the thing that actually mattered.
Keep the dashboard
Be straight about the boundary, because this is not an argument for turning off your analytics.
Keep AARRR. Instrument everything. You cannot manage what you cannot see, and a business without a funnel dashboard is flying without instruments. If your retention number were not red, you would not be reading this, and that red number is AARRR doing its job perfectly.
What it cannot do is tell you why. Detection and diagnosis are different instruments, and they run in that order. The thermometer is not competing with the doctor.
Do one thing
Take your worst AARRR number. The one that has been red for two quarters and keeps generating meetings that end in the strongest opinion winning.
Now name the dimension underneath it, and put a score on that dimension, from −3 to +3, with evidence from outside the building. If you cannot do that, you do not yet have a growth problem. You have a diagnosis problem, and every euro you spend optimising the funnel before you solve it is a euro spent making the wrong thing move faster.
The funnel tells you the number moved. Only the dimension underneath it tells you what to do on Monday.
If you want to see which dimensions sit under your levers, the quick assessment takes a few minutes. The full architecture, all 24 dimensions and the six steps, is on the method.
