AARRR

The older way of drawing a funnel ended at the purchase. For a product sold once, that was the right picture. For something free to start and billed monthly, the payment is the beginning rather than the end. A customer who pays this month and cancels next month leaves the company down whatever it cost to acquire them, and a funnel that stops at checkout doesn't show that loss at all.

So meetings turned into three people reading three different numbers. Marketing reports five thousand more signups this month, the product team reports that one in ten is still around after four weeks, and finance reports flat revenue. All three are true, but with no picture connecting them, nothing gets decided about next quarter — and the signup count, sitting at the front, usually wins the argument.

AARRR lays five stages in one row so the conversation fits on a single page. Dave McClure proposed it in 2007, and the name comes from the initials of the five stages, which happen to sound like a pirate. What matters is the use: once the five numbers sit side by side, the stage where people drop out fastest is visible, and the team spends a quarter on that one.

It starts with pinning each of the five stages to a specific behavior in your own product. You can borrow the stage names, but the conditions for entering each one have to be defined locally.

1. Decide what action counts as entering each stage. Acquisition is a completed signup, activation is creating a first document and sharing it with one person, retention is still visiting weekly four weeks after signup, referral is sending an invite link, revenue is converting to a paid plan. Leave these vague and every number downstream is vague. Activation is the one teams argue over longest, because it has to name the moment a user first gets something out of the product rather than the moment they finished setting it up.
2. Give each stage a numerator and a denominator so it reads as a rate. Activation rate is the share of signups that reach a first share. Raw counts hide problems, because a month with heavy ad spend lifts all five at once.
3. Count people grouped by when they arrived. Cohort analysis does this — measure the four-week retention of April signups separately from May's. A blended average makes retention look like it fell in any month with a surge of new users, since people who haven't reached four weeks yet are sitting in the denominator.
4. Put the five numbers in a row and pick the leakiest gap. A thousand signups, 220 activated, 90 retained at four weeks, 20 paying means 78 percent is lost between signup and activation. That gap is the quarter's target.
5. Make that one stage the goal and demote the rest to numbers you watch. Commit to raising all five and the team splits five ways, and none of the five moves visibly.

It fits products people return to after signing up, especially ones that start free and convert to paid later. The two stages after payment are the point of the model, so in a one-off transaction those stages sit empty. It also works as shared vocabulary in an organization where marketing, product and sales each track different numbers, since everyone can point at the same row and say which box they own.

There are three limits. First, it assumes people flow in one direction. Real users arrive through an invite and pay on day one, or cancel and return six months later. Second, it fits poorly when a product is sold to companies through a sales team, because one contract involves several people and the person using the product isn't the person paying, so moving one individual from stage to stage stops making sense. Third, attention drifts to the front. Acquisition responds to spending within a week while retention takes months, so teams keep pouring into the top of a funnel that leaks below. Keeping one north star metric for the whole team and using the five stages to explain why that number moved is the safer arrangement.

"The leakiest stage in our funnel is activation. 78 percent of signups leave without making a first document."

Naming the one stage this quarter's work goes into.

"We're not looking at referral this quarter. Activation comes first, then that."

A statement that the team is not raising all five at once.

"More acquisition just means more leakage. Let's talk ad budget after we fix the next stage."

Plug the lower stage before pouring money into the top one.

At Pagedeck the chart went up on the wall after Ravi tripled signups in three months through paid ads. Signups tripled and paid conversions barely moved. Mia recounted by monthly cohort and found activation had slid from 34 percent to 19 percent — most of the new arrivals never created a single document.

What Grace asked in that meeting wasn't whether to cut the ads. It was which stage the team would own for the quarter. They chose activation and left acquisition exactly where it was, neither raised nor cut. Ravi's summary was that the ads weren't the mistake; the stages below them couldn't absorb what arrived. Three months later, with spend unchanged, activation was back above 30 percent and paid conversions had roughly doubled.

The most common reading is that all five stages need managing. Five boxes are drawn, so filling all five looks correct, and a quarterly plan with five lines looks like a lot of work. The model was built to make you choose one. Attach an owner and a target to each stage and the team divides five ways, and the effort that would have moved one number gets spread until nothing moves. At the end of the quarter you have five numbers that each rose slightly and flat revenue.

The other mistake is assuming the stages happen in order. The picture runs left to right, so it reads that way. Real users arrive through an invite and pay the same day, or skip activation, use the free plan for months, and upgrade out of nowhere. Forcing the sequence produces numbers that don't match reality. The stages are better understood as five questions you want answered than as a path anyone walks. Read that way, a user sitting in two boxes at once stops being a measurement error.