AARRR Pirate Metrics: Spot Growth Leaks in 90 Days

AARRR is a framework Dave McClure introduced in 2007 for thinking about a startup's customer lifecycle in five stages: Acquisition, Activation, Retention, Referral and Revenue. Said out loud it sounds like a pirate, hence the name.

Its value is not the acronym. It is the idea that growth is a funnel with leaks at specific points, and that fixing the biggest leak usually beats pouring more people into the top. This post turns that idea into a 90-day plan for a small SaaS team, using one running example so you can see how the stages are defined in practice.

The running example

Imagine a small invoicing and quoting app for sole-trader tradespeople: electricians, plumbers, tilers, handymen. They sign up on a phone, create a quote, turn it into an invoice after the job, and get paid by card.

The numbers used below are invented for illustration. The point is the method, not the values.

Step 1 (weeks 1 to 2): Define each stage for your product

The framework only works if every stage is a specific, observable event in your product. Generic definitions ("user is engaged") are where most AARRR dashboards go wrong.

| Stage | Generic meaning | Definition for the invoicing app | |---|---|---| | Acquisition | Someone arrives | A new visitor lands on the site or app store page | | Activation | They get the first real value | They send their first real invoice to a real customer within 7 days of signup | | Retention | They keep getting value | They send at least one invoice in weeks 3 to 6 after signup | | Referral | They bring others | A tradesperson signs up after receiving an invoice or a direct invite from an existing user | | Revenue | They pay | They move to a paid plan or pay a transaction fee |

Two things to notice. Activation is defined by the core job, not by finishing onboarding: a user who fills in their logo and business details but never sends an invoice has not activated. And retention uses a window that matches how often the job happens. A tradie invoices weekly. A tool for annual tax returns would need a completely different window.

Some people reorder the stages (Revenue before Referral, or Retention first in the thinking). The order matters less than having clear definitions.

Step 2 (weeks 3 to 6): Get a baseline

Instrument each event and collect a few weeks of data. For a small product you do not need an analytics platform to start. A spreadsheet updated weekly from your database is enough.

Suppose a four-week baseline looks like this:

| Stage | People | Conversion from previous stage | |---|---|---| | Visitors | 2,000 | | | Signups | 160 | 8% | | Activated (first real invoice within 7 days) | 40 | 25% | | Retained (invoice in weeks 3 to 6) | 26 | 65% | | Paying | 12 | 46% of retained | | Referred signups | 5 | |

Again, illustrative numbers only.

Step 3: Find the biggest leak

Look for the stage where the drop is both large and fixable. In this example the obvious leak is activation: three out of four people who sign up never send a real invoice.

Compare that with the alternatives. Doubling traffic would double everything, including the people who sign up and leave, and it costs money. Improving retention from 65 percent is possible but it is already the healthiest stage. Fixing activation from 25 percent to, say, 40 percent would increase every stage below it by more than half without a single extra visitor.

This is the general pattern for early products: fix activation and retention before spending on acquisition. Paying to fill a leaking bucket is the most common way small SaaS teams burn their runway.

Step 4 (weeks 7 to 12): Fix one leak properly

Pick the one stage and give it the rest of the quarter. For activation in the invoicing app, that might look like:

1. Watch it happen. Ask five recent signups who did not send an invoice to share their screen or just tell you what happened. You will hear things like "I needed to add my bank details and did not have them on me" or "I was not sure it would look professional to the customer".
2. Shorten the path. Let people send the first invoice before completing their full profile. Show a preview of what the customer will see.
3. Nudge at the right moment. A reminder the evening after signup, when the day's jobs are done, rather than at a random time.
4. Measure the same definition. Do not quietly change what "activated" means halfway through. That is how teams convince themselves a fix worked.

At the end of the quarter, re-run the baseline table. If activation moved, move on to the next biggest leak. If it did not, you have learned that your guess about the cause was wrong, which is also useful.

When the numbers are too small

With twenty signups a month, percentages swing wildly and a single user can move a stage by five points. Do not over-read small data. At that scale AARRR is better used as a list of questions than as a dashboard:

• Of the last ten people who signed up, who sent an invoice, and why did the others not?
• Of the people who sent one, who sent another three weeks later?
• Did anyone pay, and what made them decide?
• Did anyone tell a friend?

You can answer all four by talking to people. The framework still tells you which question to ask first.

Mistakes to avoid

• Treating signups as activation. A signup is a promise. Activation is the promise kept.
• Measuring referral with a share button. Most B2B referral happens through conversation and the product itself (the customer receiving an invoice). Track where new signups actually came from by asking them.
• Five dashboards, zero decisions. One table, updated weekly, with one leak you are working on.
• Optimising revenue before retention. Pricing experiments on a product people leave after a month are mostly noise.

Using a beta cohort for the early stages

Activation and retention are the hardest stages to understand from numbers alone, because the reasons are inside people's heads. A structured beta helps here. On LetsBeta, businesses apply to try builds that are still in development and send a mid-trial report and an end-of-trial report: what worked, what broke, whether they would pay and at what price. A mid-trial report that says "I have not sent an invoice yet because I could not find where to add GST" is an activation leak described in plain words, weeks before your dashboard would show it.

For the wider set of numbers to watch alongside the funnel, see Early-Stage Startup Metrics: Seven Numbers That Matter. If your revenue stage runs through a trial, Free Trial to Paid Conversion goes deeper on that one leak.

Define the five events, get a baseline, fix one leak. Then list your build so real businesses can show you where the next one is.