More visitors may not fix the growth problem

The hypothetical practice app Cadence doubles its advertising budget because paid subscriptions are growing slowly. Its dashboard shows more visits and more accounts, but the team does not know how many people finish a useful practice session or come back.

AARRR groups product-growth questions into acquisition, activation, retention, referral, and revenue. Dave McClure’s Startup Metrics for Pirates presentation describes the five-category model. Cadence and every number in this example are invented.

The framework helps the team decide where to investigate. It does not make every product follow the same funnel or establish which behavior causes long-term value.

Define what one useful experience looks like

Cadence treats acquisition as a first visit from a relevant campaign audience. It proposes activation as completing a guided practice session within seven days of signup. Creating an account alone does not demonstrate that a person experienced the product’s value.

That activation definition is a hypothesis to examine. A completed session might reflect curiosity rather than durable benefit. The team should investigate what users find useful and compare later behavior across meaningful groups.

Five lenses: acquisition · activation · retention · referral · revenue

Arriving

Acquisition

How do relevant people reach the product?

First value

Activation

What behavior indicates a useful first experience?

Returning

Retention

Do people return for the product’s core value?

Recommending

Referral

Do users bring other suitable people?

Paying

Revenue

Does use produce revenue, with sustainable economics?

These are measurement categories, not a universal chronological journey. Define events, units, windows, and eligible populations before comparing rates.

Put denominators beside the numbers

For one invented weekly signup cohort, Cadence records 200 new accounts, of which 80 complete a guided session within seven days. The activation rate is 80 / 200 = 40%.

A different report lists 60 returning users during a calendar week. Dividing 60 by 200 would not produce a meaningful retention rate unless those 60 belong to the same eligible cohort and the return window is defined consistently.

Suppose instead that 24 of the 80 activated users complete another session during days eight through fourteen after signup. Cadence could report 24 / 80 = 30% retention among activated users for that window. Among all 200 signups, the corresponding rate would be 24 / 200 = 12%. Both calculations are valid when labeled; they answer different questions.

The numbers do not establish a benchmark or show whether the app is healthy. They expose how easily an unlabeled dashboard can confuse the discussion.

Define each remaining category deliberately

For referral, Cadence counts referred people who start a useful session, alongside the number of eligible users who made referrals. Invitation clicks alone could reward spam or ineffective prompts. A useful report separates invitations, arrivals, and meaningful participation.

For revenue, it records paid customers and revenue over the same stated period. Revenue is not profit: refunds, acquisition cost, payment fees, service delivery, and support still matter. A campaign can increase sales while worsening the economics.

Acquisition should also distinguish channels and audiences. A high-volume channel that attracts people outside the intended use case can depress downstream rates without implying that onboarding broke.

Use the categories to pick an investigation

Question Proposed check What it would not prove
Why do signups stop before a first session? Observe relevant new users attempting the first session A long-term retention improvement
Why do activated users not return? Investigate recent use and abandonment episodes That a reminder will solve the problem
Which channels bring suitable users? Compare consistently defined cohorts by source That the channel caused every behavioral difference

Cadence’s first proposal is to investigate incomplete first sessions before increasing spending further. A broken flow, unclear promise, or unsuitable audience would each require a different response.

Segment carefully enough to explain differences without creating tiny groups whose rates are dominated by chance. Record tracking changes and missing events so an instrumentation problem does not become a product conclusion.

Build a small measurement contract

Choose one meaningful event per category to begin. Write down its unit, eligibility, time window, data source, and owner. Review whether the event captures customer value and whether incentives could encourage an unhelpful substitute behavior.

The categories are not compulsory sequential gates. Customers can pay before returning, recommend before subscribing, or arrive through several channels. Represent the real business rather than forcing every journey into five arrows.

Use a Sales Funnel to inspect the sales journey and stage handoffs. Use Jobs to Be Done when the team needs to understand why people seek or abandon the product. AARRR highlights a measurement question; customer research helps explain the behavior behind it.

Cadence’s next decision is where to investigate first, with cohort definitions clear enough that the next dashboard can inform that decision.

Sources and further reading