Pirate Metrics (AARRR)

In brief

In 2007 Dave McClure, then an angel investor and later founder of the accelerator 500 Startups, gave a talk called "Startup Metrics for Pirates." His joke was that the five things a founder should track spell AARRR, the noise a pirate makes: Acquisition, Activation, Retention, Referral, Revenue. Behind the joke is a serious claim. A business is a funnel. Strangers arrive, some try the product, some come back, some tell friends, some pay. Each arrow is a conversion rate, and the whole system is only as healthy as its weakest link. McClure's argument is that founders drown in dashboards and fixate on one impressive-looking total, usually raw signups or page views, while the real problem sits at a single stage where users quietly leak away. Measure the five stages separately, find the leak, fix that.

The full treatment

The problem it answers

Early-stage companies have almost no data and too much of it at once. A founder can watch traffic climb, announce a milestone, and still be dying, because traffic that never activates, never returns, and never pays is not a business. McClure's target is the "vanity metric," a number that goes up and to the right and makes everyone feel good while explaining nothing you can act on. Cumulative registered users only rises; it cannot fall, so it can never tell you that something broke last week. Pirate Metrics replaces one flattering total with five diagnostic rates, each of which can go down, and each of which points at a specific thing to fix.

How it works: the five stages

Acquisition is how people first find you: search, ads, social posts, word of mouth, a link from another site. The metric is volume and, paired with cost, how expensive each arrival is.

Activation is the first good experience, the moment a visitor becomes a user who has felt the product work. It is not signing up; it is reaching the point where the product delivered on its promise at least once. Teams often define an "activation event" (uploaded a first file, invited a teammate, completed a booking) and measure the share of new arrivals who reach it.

Retention is whether they come back. This is the stage McClure and later analysts treat as the true test, because a product no one returns to has no future no matter how cheap acquisition is. It is usually measured as the fraction of a weekly or monthly cohort still active a set time later.

Referral is whether users bring other users. This is the stage that can compound, and it is treated separately below.

Revenue is whether, and how much, they pay: conversion to a paying plan, average revenue per user, the money that finally justifies everything upstream.

The stages are sequential. You cannot retain someone who never activated, and you cannot earn referral revenue from someone who never came back. So the funnel has an order, and a conversion rate lives on each step: of everyone acquired, what share activate; of those, what share retain; and so on.

The leaky-stage logic

The model's real payoff is arithmetic. Suppose you acquire well but only 20 percent of arrivals activate, then 40 percent of those retain, then 10 percent of those pay. The end-to-end conversion is the product of the rates, so 0.20 times 0.40 times 0.10 gives 0.8 percent reaching revenue. Now double your worst honest lever. Buying twice the traffic doubles the input but leaves every leak in place. Fixing the 20 percent activation to 40 percent, by contrast, doubles the whole downstream funnel at no acquisition cost. The stage with the lowest rate, relative to what is achievable, is the binding constraint, and effort spent anywhere else is largely wasted until that stage improves. This is the same discipline that a factory learns from the theory of constraints (see bottlenecks): the throughput of the system is set by its narrowest point, so you improve the system by improving that one point, not by speeding up everything at once.

Referral as the compounding stage

Four of the five stages are linear: work harder and you get a proportional return. Referral is different, because a referred user can themselves refer. When each user brings on average more than one new user who stays, growth stops being additive and starts to multiply, the mechanism behind viral products. This is where Pirate Metrics touches network-effects: a referral loop that runs above the break-even point produces exponential rather than paid-linear growth, and it lowers the effective cost of acquisition because your existing users are doing the acquiring. It is also the hardest stage to fake. People refer products they actually value, so a strong referral rate is usually evidence that retention is genuinely good underneath it.

Lineage

Pirate Metrics did not appear from nowhere. Direct marketers had modeled the "purchase funnel" for a century, and the general idea of a customer moving through awareness to purchase is old. McClure's contribution was to compress it into five memorable stages aimed squarely at software startups, add retention and referral as first-class stages (not afterthoughts to a sales funnel), and give it a name that stuck. It arrived alongside a wider movement. Eric Ries's The Lean Startup (2011) made the vanity-versus-actionable-metrics distinction famous and shares the same target: measure what you can act on. Alistair Croll and Benjamin Yoskovitz's Lean Analytics (2013) extended the stage-by-stage approach into a full method. The three form a loose canon of the metrics-driven startup era.

The strongest case for it

Its virtue is focus. A first-time founder handed a hundred possible numbers will optimize the reassuring one and go broke. Five stages, in order, force the question that matters: where are we actually losing people. The framing is also honest about causation. Because the stages are sequential and multiplicative, it makes plain that acquisition spending is pointless when activation or retention is broken, a mistake companies make constantly by pouring money into ads to fill a bucket with a hole in it. And it is memorable enough to survive contact with a busy team, which is why it spread far beyond the talk that launched it.

The strongest case against it

The sharpest objection comes from within the lean-analytics camp itself. Croll and Yoskovitz argue that the five metrics are not equal and should not be tracked in parallel: a startup should identify its single most important metric for its current stage, the "One Metric That Matters," and ignore the rest until that one moves. On this view AARRR risks becoming its own dashboard of vanity, five numbers watched at once and none of them driven.

A second objection is that the funnel imposes a linear customer journey that many products do not have. Users churn and return, discover the product through a friend before ever being "acquired" by marketing, or pay before they are retained. Sequencing the stages can hide these loops. Ries's own emphasis on cohort analysis and retention is partly a correction: retention, not the tidy funnel, is where he and many practitioners now say the truth lives.

Third, the model says nothing about whether the economics close. A funnel can convert beautifully at every stage and still lose money on every customer if the cost to acquire exceeds the lifetime value earned. That question belongs to unit-economics, and a founder who tracks AARRR without tracking the money per customer can optimize a machine that runs faster into the ground. Pirate Metrics tells you where users leak; it does not tell you whether the ones who stay are worth more than they cost.

Where it stands now

The five stages are near-universal vocabulary in startups, growth teams, and product analytics tools, which frequently ship dashboards organized exactly as Acquisition, Activation, Retention, Referral, Revenue. It is a standard teaching frame in accelerators and business courses. The mature view treats it not as a scorecard to admire but as a diagnostic map: use the five stages to locate the binding constraint, pick the one metric that matters for right now, fix it, and re-check the funnel, always alongside the unit economics that decide whether the whole thing is a business. The joke in the name has outlived most of the serious frameworks of its era, which is its own kind of evidence that the underlying idea was sound.

Test yourself

Think of a product you tried once and never opened again. Which of the five stages did it lose you at, and was it because the first experience never landed (activation) or because nothing pulled you back (retention)? Now ask the harder question a founder must ask: if that company had spent more to acquire people like you, would it have grown, or just leaked faster?