Pirate Metrics (AARRR)

The five stages and the question each one answers

Each stage is a question about a different moment in a customer's life, and each has its own number. The names matter less than keeping the questions separate:

• Acquisition — how do people find you at all? Measured as visits or signups by channel, which is also what tells you which channel is worth doubling down on.
• Activation — do they reach a first useful result? This is the activation rate, and it is the stage founders most often skip measuring.
• Retention — do they come back? Usually a cohort curve rather than a single figure, and the closest thing the funnel has to a verdict on the product.
• Referral — do they tell anyone? Measured by invites, shares, or the share of new users who arrived through an existing one.
• Revenue — do they pay, and is it more than they cost to acquire? That comparison is customer acquisition cost against lifetime value.

One number per stage is the whole discipline. Five numbers you check weekly beat a dashboard of forty that nobody reads, and picking them forces the argument about what each stage actually means for your product — which is where most of the value shows up.

Why the funnel is a diagnostic, not a to-do list in order

The common misreading is to work the stages top to bottom, starting with acquisition. Acquisition is the most expensive stage to improve and the one that wastes the most when the stages below it leak: doubling traffic into a product that loses most of its signups in the first week doubles the waste rather than the growth. Read the funnel to find the worst stage, then work on that one.

The second misreading is treating the acronym's order as a priority order. Several practitioners deliberately reorder it retention-first, sometimes written RARRA, on the argument that retention is what makes every other stage worth improving. The reordering is not a different framework, it is an opinion about which stage usually deserves attention first in a product that already has users. Both versions agree on the underlying point: the stage you should work on is the one with the worst number, not the one at the top of the page.

AARRR and growth loops: funnel thinking versus loop thinking

The sharpest modern criticism of AARRR is structural rather than about the stages. A funnel is linear and ends at revenue, which quietly encourages you to treat a cohort as spent once it converts. A growth loop is a cycle where the output of one stage becomes the input of the next round — a user who invites a colleague, or a page that earns a link that ranks the page. In AARRR, referral is a stage at the bottom. In loop thinking, referral is the return path that makes acquisition compound without paying for it twice.

Keeping both is more useful than choosing. The funnel is good at telling you where you are losing people; the loop is good at telling you where growth could come from that is not a bill. Neither is a substitute for a single north star metric that says what the whole thing is for.

A first-party note from running this at AgentCeres: our own version splits activation rather than measuring it as one number. When activation was a single figure we could see it was bad and not where it broke, because everything between an account existing and the product doing something visibly useful sat inside that one number. Breaking that stretch into separate rungs, each with its own count, is what turned it from a bad number into a fixable one. If a stage in your funnel covers several minutes of a stranger's first experience, it is probably hiding more than one problem.