AARRR vs RARRA
Same five stages, different order, and the order is the entire argument. AARRR lists them in the order a customer moves through them. RARRA lists them in the order a team should fix them, putting Retention first on the grounds that pouring acquisition into a product nobody keeps is filling a bucket with a hole in it. Nobody added or removed a stage — so the disagreement is about sequencing, and sequencing is an empirical question.
At a glance
The five stages
Acquisition, Activation, Retention, Referral, Revenue
Retention, Activation, Referral, Revenue, Acquisition
What the order means
The order a customer passes through
The order a team should work in
Where it comes from
Dave McClure's startup metrics talk, as a way to describe a funnel end to end
A reframing for mobile apps, where installs are bought and most of them never come back
What it assumes
That you have users moving through it
That you already have a cohort big enough to measure
The argument is with a misreading, not with the model
Every criticism RARRA makes of AARRR is a criticism of how AARRR gets used: as a to-do list read top to bottom, which puts acquisition first by accident of the acronym. The original framing never claimed to be a work order — it described the path, which really does start with someone finding you. Read as a description, AARRR says nothing about where to spend Monday.
That is why the two cannot really contradict each other. RARRA is a claim about priority dressed as a claim about structure, and the structure it borrows is the one it is arguing with. Anyone who says they have switched from AARRR to RARRA has changed what they work on first, not what they measure.
When retention-first is right, and when it is not
The case is strongest where acquisition is expensive and the product is free. An app paying for every install and losing three quarters of them inside a week is buying water for a bucket with a hole in it, and no amount of channel optimisation fixes the hole. In that setting, retention-first is not a doctrine, it is arithmetic.
It is weakest at the very beginning, and in anything sold rather than self-served. You cannot measure retention on a cohort of zero, and a product with fifty users has a retention curve made mostly of noise — the earlier honest signal there is activation. And a tool sold on an annual contract produces its first real renewal number a year after you needed it; starting with retention there means starting with a number you will not have for four quarters.
Let the numbers pick the order
The defensible version of RARRA's point is not "always start with retention". It is "do not start with acquisition by default". The way to settle which stage actually goes first is to measure all five and start with the weakest — which is neither acronym's order, and is the answer both of them are gesturing at.
It matters because the stages multiply rather than add: ten percent better at four of them is worth about forty-six percent more at the end, not forty. A weak stage does not subtract from the result, it scales everything downstream of it — which is exactly why the order is worth arguing about, and exactly why the argument should be settled with your numbers rather than with someone else's app.
Which to use, and when
Use AARRR to describe the journey: everyone knows it, and it is the right chronology. Borrow RARRA's instinct to resist acquisition-first reflexes. Then let your own five scores pick the stage you actually work on, because neither acronym knows whether your activation is at sixty percent or your retention is at thirty.
The terms behind this
The other comparisons
Whichever framework you settle on, the first question is the same one: which of the five stages is holding the others back? Fifteen questions, three minutes, no signup.