AARRR framework: pirate metrics in practice

What AARRR is and why pirates

AARRR is a five-stage model of a user's life inside your product: Acquisition, Activation, Retention, Revenue, Referral. Read the acronym out loud and it sounds like a pirate — which is why the framework is universally called pirate metrics. Dave McClure of 500 Startups proposed it in 2007 as a way to stop teams from staring at a single number (DAU, MRR, signups) and missing where the funnel is actually broken.

The framework's real value is not the acronym. It is the discipline of forcing every product conversation to name which stage a problem lives in. A PM who says "growth is slowing" without naming a stage is asking the team to debug a black box. A PM who says "activation rate dropped from 42% to 31% on the iOS cohort after the v4.2 release" is asking the team to fix something concrete. AARRR is the vocabulary that makes the second sentence possible.

One important framing note: AARRR is not a strategy. It is a diagnostic. It tells you where the leak is, not what to build. The teams that get the most out of it pair the funnel with a North Star metric and a quarterly bet on one stage — usually the one with the worst conversion relative to a credible benchmark.

Acquisition

Acquisition is the stage where a stranger first touches the product — a landing page visit, an app store impression, a signup. The metrics that matter at this stage are not just volume. Volume without quality is the most expensive mistake in growth.

The trap here is conflating cheap traffic with good traffic. A channel that delivers signups at $2 CAC but converts to activation at 4% is almost always worse than a channel at $18 CAC that converts at 38%. Always carry the channel dimension down at least one stage further than you think you need to.

Activation

Activation is the moment a user gets the first real value from the product. Not "signed up", not "verified email" — value. Defining activation well is the single highest-leverage analytics decision a product team makes. A bad definition makes every downstream metric noisy.

Examples of credible activation events:

• A SQL-drill app: 3 problems solved in the first 3 days.
• A B2B SaaS for teams: 2 teammates invited and 1 project created within 7 days.
• A consumer social app: followed 5 accounts and posted one reaction.
• A marketplace: completed a search, opened a listing, sent a message in the same session.

Metrics to watch:

• Activation rate — share of signups who hit the activation event within a window
• Time to value (TTV) — median time from signup to activation
• Onboarding step conversion — drop-off between each onboarding screen

Activation is the most under-invested stage in most companies because the work is unglamorous — copy edits, empty-state design, default settings — and the wins are quiet. Compounding wins, though.

Retention

Retention asks: do users come back after they activate? It is measured by cohorts, not by aggregate active-user counts, because aggregate counts mix new and old users and hide the trend.

The classic chart is a cohort retention curve: for users who signed up in week N, what share are still active in week N+1, N+4, N+12, N+24? Three shapes matter:

Useful retention metrics:

• D1, D7, D30 for consumer apps
• Weekly active rate (WAU/registered) for B2B
• Stickiness (DAU/MAU) — anything above 20% is good, 50%+ is exceptional
• Power-user share — users with 4+ sessions per week

Retention is the most honest signal of whether the product is worth building. A team can fake acquisition with paid spend; a team can fake activation with onboarding gimmicks. Retention is unfakeable on a long enough horizon.

Revenue

Revenue covers everything from free-to-paid conversion to expansion. The mistake here is staring at averages. The mistake is so common it deserves its own callout.

Core revenue metrics for a PM dashboard:

• Free-to-paid CR — share of activated users who pay
• ARPU and ARPPU — averages, with median and p90 alongside
• LTV — historical or modeled lifetime value
• CAC payback period — months until cumulative gross profit ≥ CAC

The benchmark teams cite most is LTV/CAC ≥ 3, but that ratio is a heuristic, not a law. At seed stage, a ratio of 1.5 can be fine if growth is fast and gross margin will improve. At Series C, a ratio under 3 with a long payback (>18 months) is a warning signal. The right comparison is to your own trend: is the ratio improving cohort over cohort?

Referral

Referral is the question of whether users bring in other users. The headline metric is K-factor — the number of new users one existing user generates over a defined window.

• K-factor > 1 — viral growth: each user brings more than one new user. This is rare and usually transient (Dropbox, Hotmail, Snapchat in their early years).
• K-factor between 0.1 and 0.4 — normal for most products. Worth optimizing because it cuts blended CAC.
• K-factor at 0.0 — referral is dead; revisit the share/invite UX before declaring it a product property.

Related metrics: invite send rate, invite-to-signup CR, and NPS as a soft proxy for willingness to recommend. NPS is not a referral metric in itself — it predicts referral, it does not measure it.

Putting the funnel on one dashboard

A practical AARRR dashboard is five panels, 2-3 metrics per panel, with the cohort dimension carried through every stage. Most teams build the first version in a week or two, then spend a quarter cleaning event definitions until the numbers stop wobbling.

Here is a minimal SQL skeleton for the spine of an AARRR dashboard — a single CTE that produces one row per signup with every stage flag attached. From here, every panel is a group-by.

-- one row per signup, every stage flag attached
WITH signups AS (
  SELECT user_id, signup_ts, acquisition_channel
  FROM dim_users
  WHERE signup_ts >= DATE '2026-01-01'
),
activations AS (
  SELECT user_id, MIN(event_ts) AS activated_at
  FROM fact_events
  WHERE event_name = 'problem_solved'
  GROUP BY user_id
  HAVING COUNT(*) >= 3
),
revenue AS (
  SELECT user_id,
         MIN(paid_ts) AS first_paid_at,
         SUM(gross_amount_usd) AS ltv_to_date
  FROM fact_payments
  GROUP BY user_id
),
referrals AS (
  SELECT referrer_user_id AS user_id,
         COUNT(DISTINCT referred_user_id) AS invited_signups
  FROM fact_referrals
  WHERE status = 'signed_up'
  GROUP BY referrer_user_id
)
SELECT
  s.user_id,
  s.signup_ts,
  s.acquisition_channel,
  a.activated_at,
  a.activated_at IS NOT NULL                              AS is_activated,
  DATE_DIFF('hour', s.signup_ts, a.activated_at)          AS time_to_value_hr,
  r.first_paid_at,
  r.ltv_to_date,
  COALESCE(rf.invited_signups, 0)                         AS invited_signups
FROM signups s
LEFT JOIN activations a USING (user_id)
LEFT JOIN revenue     r USING (user_id)
LEFT JOIN referrals  rf USING (user_id);

From that base table, every AARRR panel is one aggregation away — acquisition_channel for the top of funnel, is_activated for activation rate, DATE_TRUNC('week', signup_ts) for cohorts, and so on.

Common pitfalls

The most common AARRR failure is fixation on one stage. A team pours money into Acquisition while activation sits at 18% — meaning 82% of paid users vanish before they ever see value. Three months and a million dollars later, the team realizes the leak was never at the top. The fix is procedural: every quarterly review names the worst-performing stage relative to benchmark and assigns one team to it.

A second pitfall is a fuzzy Activation definition. If activation is "completed signup", you are measuring registration, not value. Activation must be a behavioral event the user performs after signup that correlates with long-term retention. The fastest way to find a credible activation event is to plot D30 retention against early-session behaviors and look for the variable with the sharpest cliff.

A third trap is average-based Revenue analysis. Reporting ARPU without ARPPU, median, and a top-decile breakdown hides whether the business is monetizing the middle or surviving on whales. The fix is to publish a paying-user distribution alongside every revenue chart — even just three numbers (p50, p90, p99) makes the difference visible.

A fourth pitfall is aggregate Retention without cohorts. Month-over-month "active users" can rise even as every new cohort is dying faster, because the denominator is dominated by long-tenured users. Always look at retention by signup-week cohort, and prefer the curve shape over a single number. Aggregate retention is the easiest metric in analytics to accidentally lie with.

The last pitfall is AARRR without instrumentation. The framework only works if the underlying events are reliably logged with consistent user IDs across web, mobile, and server. A team that adopts pirate metrics without first auditing event coverage will be debating definitions for six months and shipping no improvements. Spend the first sprint on event QA.

Related reading

• A/B testing for product managers
• How to calculate activation rate in SQL
• How to calculate cohort retention in SQL
• How to calculate LTV in SQL
• How to calculate CAC payback in SQL
• Guardrail metrics in A/B testing

If you want to practice the SQL behind every one of these AARRR metrics on realistic product schemas, NAILDD is launching with 500+ drills modeled on real PM and analyst interview questions.

FAQ

Does AARRR work for B2B?

Yes, but the stage definitions shift. Acquisition becomes inbound leads and demo requests, Activation becomes successful team onboarding (seats filled, first workspace artifact created), Retention becomes contract renewal and seat expansion, Revenue splits into new ARR and expansion ARR, and Referral often appears as customer references rather than user-driven invites. The mechanics are identical; the events are different.

How is AARRR different from a generic funnel?

AARRR is a funnel — just one with five specific, named stages that tend to recur across product types. Some teams add a sixth stage, Awareness, in front of Acquisition for products with long consideration cycles, or rename Retention as Engagement when daily use is the goal. The vocabulary matters more than the count.

How often should we revisit our Activation definition?

Every 6-12 months, and always after a major release that changes the onboarding path or core flow. As new features ship, the behavior that best predicts long-term retention drifts — what activated users in v1 may not activate them in v3. A stale definition makes the entire funnel quietly wrong.

Activation or Retention — which should we fix first?

Activation, almost always. Retention is a measurement of users who already activated, so if activation is broken there is nothing to retain. Fix activation until the rate plateaus, then move budget to retention experiments, then — and only then — scale paid acquisition.

Can we use AARRR during MVP stage?

Yes, in a compressed form. At MVP the priority is verifying that any conductivity exists between Acquisition and Retention — that strangers can find the product, get value, and come back at least once. Spending MVP energy on Referral optimization or revenue tuning is premature. Use AARRR as a checklist of "have we proved this stage exists yet", not as a five-axis optimization problem.

What does pirate metrics look like in practice inside a real team?

Usually a single dashboard with five panels, 2-3 metrics per panel, broken out by acquisition channel and weekly signup cohort. Every quarter the team picks one stage as the bet, writes an OKR for it, and assigns ownership. The dashboard does not change; the focus does. Teams that rotate the bet thoughtfully tend to compound — the ones that change focus every two weeks tend to stall.