AARRR Metrics, Redrawn as a Loop for Product-Led Growth
The AARRR metrics framework — Acquisition, Activation, Retention, Referral, Revenue — is a five-stage lifecycle model that investor Dave McClure coined in 2007 to name the metrics that actually matter for early-stage products. Its weakness is quiet but real. It's almost always drawn as a one-directional funnel, and for product-led growth that sequence misleads you. Activation should sit earlier than "second," and referral should feed acquisition rather than wait politely at the end.
What AARRR metrics actually stand for
The framework comes from McClure's "Startup Metrics for Pirates" talk at Ignite Seattle in 2007. As the marketing team at DinMo puts it, he presented AARRR "as a simple and effective reference model to understand the five key stages of the customer lifecycle." The name is an acronym you can read as a pirate's growl, which is the whole joke.
Here's the plain-language version of each stage.
- Acquisition — how people first find you. A search result, an ad, a friend's link.
- Activation — the first genuinely good experience. Not "they signed up," but "they did the thing the product is for."
- Retention — do they come back? Once, twice, in week four.
- Referral — do they tell other people, and do those people show up?
- Revenue — do they pay you, and how much.
That's the whole vocabulary. Five words, one lifecycle. The trouble starts when you draw arrows between them and assume the arrows only point one way.
One canonical metric per stage
Most teams don't struggle to name the stages. They struggle to pick one number per stage and stop arguing. So I'll be opinionated here: each stage answers exactly one question, and there's one canonical metric that answers it honestly.
| Stage | The question it answers | One canonical metric |
|---|---|---|
| Acquisition | Are the right people arriving? | Visitor-to-signup conversion rate |
| Activation | Did they reach first value? | % of signups who hit the activation event |
| Retention | Do they come back? | Cohort retention at month 1 / 3 |
| Referral | Do users bring users? | Invites sent → new activated users |
| Revenue | Do they pay, and expand? | Free-to-paid conversion rate |
Notice what's missing: total signups. It's the metric everyone reaches for and the one that tells you the least. Feels great in a board deck, means nothing on a Tuesday, because it counts the people who created an account and then vanished like a bus you watched pull away from the curb. Report activated users instead. A signup who never activated isn't a customer. They're a tab someone closed.
For the connection between one headline metric and the inputs that move it, I've written separately about building a North Star metric tree — worth reading alongside this if you're deciding what to put on the wall.
A worked example: 1,000 visitors through the funnel
Let's run actual numbers, because the leaks only become obvious when you count.
Start with 1,000 website visitors and a freemium product. According to OpenView's 2022 Product Benchmarks, freemium products convert about 6% of visitors into signups, versus 3–4% for free-trial products. So our freemium funnel produces roughly 60 signups. A free-trial version of the same product would produce 30 to 40.
At first glance, freemium wins. More people through the door. But watch what happens next.
OpenView's data shows freemium signup retention decays steeply: 19% in month one, 11% in month two, 9% in month three. So of those 60 signups, about 11 are still around at month one. By month two you're near 7. By month three, roughly 5. The bus filled up nicely at the first stop and then quietly emptied at every stop after.
Then revenue. OpenView reports the median free-to-paid conversion rate is roughly 7% for freemium and about 14% for free-trial. Apply 7% to a generous read of your surviving freemium base and you're converting a small handful. The free-trial product started with fewer signups but converts them at double the rate.
So where do most freemium products bleed users? Not acquisition — it's activation and month-two retention. The visitor-to-signup step is the healthiest part of the funnel. The damage happens in the two stages the classic AARRR diagram tells you to worry about after you've celebrated your signup number.
Think of it like a subway line. Acquisition is the crowded platform. Activation is whether people actually board the train instead of standing there checking their phones. Retention is whether they're still on board three stops later. If you only count the platform crowd, you'll swear the line is thriving while the last car pulls in empty.
Why the linear funnel breaks for PLG
A funnel is a pour-through. You dump volume in the top, some fraction survives each stage, and whatever reaches the bottom is your yield. Nothing that comes out the bottom goes back to the top. It's a one-time pour.
In 2018, Brian Balfour and the Reforge team — Balfour, Winters, Kwok, and Chen — published "Growth Loops are the New Funnels," arguing that this shape misrepresents how the fastest products actually grow. Their claim, on July 31, 2018: "The fastest-growing products are better represented as a system of loops, not funnels. Loops are closed systems where the inputs through some process generates more of an output that can be reinvested in the input."
The cooking analogy I keep coming back to: a funnel is a single pour of batter. A loop is a sourdough starter. With a starter, part of tonight's output becomes tomorrow's input, and if you keep feeding it, it compounds. A product where each new activated user invites two friends who also activate isn't pouring batter. It's keeping a starter alive. AARRR-as-funnel can't show that, because referral sits at the end of the pour with nowhere to flow.
For freemium and self-serve products this isn't academic. When referral genuinely feeds acquisition, drawing it as a terminal stage means you never measure the compounding, and the thing you don't measure is the thing you quietly starve.
Redrawing AARRR as a loop for self-serve products
Here's my defensible take, and where I'll plant a flag. Keep the five words. Drop the strict order. Redraw it so referral loops back into acquisition and activation moves up front.
The old sequence reads like a receipt:
Acquire → Activate → Retain → Refer → Revenue
The loop sequence for self-serve freemium reads more like this:
Acquire → Activate (early, aggressively) → Retain → Refer → back to Acquire, with Revenue sitting alongside as a byproduct of the whole cycle rather than a gate at the end.
Why move activation early? Because the numbers demand it. Freemium wins the visitor-to-signup step but retention collapses to single digits by month three, per OpenView. You can't afford to treat activation as stage two of a leisurely sequence when most of your signups are gone before you'd even have gotten around to measuring it. Activation is the intervention that decides whether month one's 19% is a floor or a cliff.
Why treat referral as a loop input rather than a terminal stage? Two reasons. First, McClure never insisted on strict ordering. As DinMo notes, "unlike a classic funnel, these steps do not have to be taken in a precise order. The recommendation phase, for example, is highly beneficial for activating new customers." The recommendation stage activates people. That's a loop by any honest reading.
Second, revenue behaves like a byproduct in self-serve, not a gate. OpenView's 2023 SaaS Benchmarks report found product-influenced revenue is highest among self-serve freemium companies, at roughly 90% of total revenue. If nine-tenths of your revenue is influenced by the product experience itself, then revenue isn't a stage you push people into at the end. It falls out of a working loop. Fix activation and retention, keep the referral flow feeding acquisition, and revenue follows.
Which analytics surface measures each stage
Different stages live in different tools, and pretending one dashboard covers all five is how you end up with five half-answers. Here's the honest mapping.
Acquisition is web analytics territory — traffic sources, landing-page conversion, campaign attribution. Referral overlaps with attribution tooling, especially anything that tracks invite links and where new users came from. Revenue lives in billing and RevOps systems, because conversion and expansion are financial events, not product events. And the middle three — activation, retention, and cohort behavior — are the domain of product analytics: funnels, retention curves, cohorts, user flows.
For that product-analytics slot, several options apply the same rubric. Amplitude and Mixpanel are the established names, strong on event tracking, funnels, and cohort retention, with mature ecosystems. PostHog covers similar ground with an open-source, self-hostable option and session replay bundled in. Kixo belongs in the same row. It does product analytics — events, funnels, retention, cohorts, user flows — plus session replay and mobile attribution, and its differentiator is chat-first querying, where you ask a question in plain language and get a chart or dashboard back with a visible reasoning trail. The honest caveat is that Kixo is newer than the others, on per-project FREE/GROWTH/ENTERPRISE plans, so it hasn't got the track record of the incumbents. You can see the current capability set at Kixo.
One gentle warning, because I've watched teams learn this the expensive way. The tool doesn't fix the framework. A prettier retention chart — chat-generated or hand-built — still won't tell you why activation is late. It'll show you the cliff in higher resolution. The re-sequencing is your job. The tool just renders it. If you want to think harder about instrumenting the events behind these stages, the event-tracking fundamentals matter more than which vendor draws the curve.
The mistakes everyone makes with AARRR
The failures are consistent enough that I can almost predict them from the org chart.
The first is treating AARRR as a rigid pipeline — running it strictly left to right when McClure himself said the stages aren't ordered. The second is measuring acquisition volume as a win. Traffic and signup totals are the vanity metrics of this framework. They go up and to the right while your actual business quietly stalls, and nobody wants to be the person who points at the pretty line and asks "so what?"
Third, putting referral last for a freemium product. If your growth compounds through invites, referral is a loop input, and burying it at the end means you never fund the thing that scales you. Fourth, and most common, reporting signups instead of activated users. A signup count is the check-engine light you've covered with tape. It looks calm right up until the day it doesn't.
The short version
Keep the five stages as vocabulary. They're a genuinely good shared language for talking about growth. Drop the strict order. Move activation early, because freemium retention decays too fast to leave it for later. And treat referral as an input that feeds acquisition, not a courtesy stop at the end of the line.
So is AARRR outdated? No — it's mislabeled as a funnel. Read it as a loop with a starter you keep feeding, and it describes product-led growth about as well as any five words can.