AARRR (Pirate Metrics)
Acquisition to Revenue: five numbers that describe a whole business model, from first click to payback.
On this page
The gist
- →AARRR = Acquisition, Activation, Retention, Referral, Revenue — a funnel of conversion rates, not counts; the lowest step-to-step % is your bottleneck
- →Define each stage in the client's language (e.g. Activation = first order in 7 days), track one cohort forward, then divide each stage by the one above it
- →Compute CAC per RETAINED user (spend / users who survive to retention), not per install — cheap installs often hide an expensive business
- →If retention is weak, fix it before spending more on acquisition (RARRA logic) — more traffic into a leaky bucket just buys a bigger leak
The framework at a glance
When to use it
Reach for AARRR whenever the case is about a consumer app, a marketplace, a subscription product or any business where the same customer transacts repeatedly — quick commerce, food delivery, edtech, fintech, OTT, ride-hailing, D2C with repeat purchase, SaaS. Typical prompts: "Our app has 5 million downloads but revenue is flat, what's going on?", "A Swiggy-style client is burning cash on discounts, where should they cut?", "Our edtech client's paid conversion dropped 30 percent this quarter", "How would you grow a new social app in Tier-2 India?", "We're spending 40 crore a year on performance marketing — is it working?" It is also the right skeleton for product-management and growth rounds in competitions run by Flipkart, Zomato, Nykaa and Meesho. Do not use it for a one-off big-ticket purchase (a house, a wedding venue, industrial capex) where there is no retention loop, and do not use it for pure cost, supply chain or org-design cases — a profitability tree, value chain or 7S serves you better there.
What it is
AARRR stands for Acquisition, Activation, Retention, Referral and Revenue. It was coined by investor Dave McClure (500 Startups) in a 2007 talk called "Startup Metrics for Pirates" — the name stuck because the acronym sounds like a pirate's growl. The idea was a reaction to startups drowning in dashboards full of vanity numbers (page views, app installs, "10 million registered users") that told nobody whether the business actually worked. McClure's argument: there are only five things worth measuring, they happen in order, and each one is a conversion step from the one before it.
Keep reading ↓Show less ↑
Read it as a funnel where every stage is a percentage, not a count. Acquisition asks how a stranger first lands on you and what each landing costs. Activation asks whether that stranger reached the "aha" moment — the first time the product visibly did its job (first ride booked, first order delivered, first lesson completed). Retention asks whether they came back a week and a month later, which is the only stage that proves you built something people actually want. Referral asks whether they pulled other people in for free. Revenue asks whether the money they generate over their lifetime exceeds what you paid to get and serve them. The diagnostic power comes from stacking the conversion rates side by side: the lowest step-to-step conversion is your bottleneck, and that is where a growth team should spend the next quarter.
Two things to know before you use it in a case. First, AARRR is a measurement and diagnosis framework, not a strategy framework — it tells you where the leak is, not what to build. You still need something else (jobs to be done, customer journey mapping, pricing analysis) to generate the fix. Second, the ordering is contested. Practitioners popularised RARRA — Retention, Activation, Referral, Revenue, Acquisition — arguing that in saturated markets, pouring money into acquisition while retention is broken is simply buying a leaky bucket faster. That critique is highly relevant to India, where cheap Android installs make Acquisition look healthy while Activation and Retention quietly collapse. Naming this tension in an interview is an easy way to sound like you have actually run a growth team.
How to apply it, step by step
- 1
Define the five stages in the client's own language
Generic labels win no points. Before touching numbers, translate each stage into a concrete, countable user action for this specific business. For a quick-commerce app: Acquisition = app install, Activation = first order delivered within 7 days, Retention = at least one order in month 3, Referral = a referral code redeemed by a new user, Revenue = contribution margin per user per month. State these definitions out loud so the interviewer can correct you early rather than at the end.
- 2
Get the absolute numbers for each stage
Ask for the count of unique users at each stage over a fixed window, usually a month or a single cohort. If the interviewer has no data, estimate and say you are estimating. Keep it to five or six stages — the point is to see the shape of the funnel, not to build a 20-row dashboard. Always work with one cohort of users moving through time, not five unrelated monthly snapshots, or your conversion rates are meaningless.
- 3
Convert to step-to-step percentages
Divide each stage by the one above it: installs to activated, activated to retained, retained to referring, and so on. This single act separates a candidate who has memorised the acronym from one who can use it. Write the percentages in a column down the page. Absolute numbers hide problems; percentages expose them.
- 4
Name the bottleneck — the lowest conversion
The weakest step-to-step conversion is where you focus, because a 10-point improvement there is worth far more than 10 points anywhere else. Sanity-check against benchmarks: for Indian consumer apps, install-to-first-transaction of 25 to 35 percent is normal, and month-3 retention below 20 percent usually means the product has no habit. Say it explicitly: 'Acquisition is not the problem here, retention is, and here is why that changes where the money goes.'
- 5
Ask why the bottleneck exists, using qualitative evidence
Numbers tell you where users leave; they never tell you why. Propose specific diagnostics: session recordings and drop-off screens for onboarding, exit surveys for churned users, five customer interviews per segment, and a cohort split by acquisition channel, city and device. A common finding is that the bottleneck is not uniform — one paid channel or one city drags the average down while the rest are healthy.
- 6
Segment the funnel before you conclude
Rerun the same five percentages by channel (organic vs paid vs referral), by geography (metro vs Tier-2), and by cohort month. A blended 30 percent activation can easily be 55 percent organic and 12 percent from a cheap install campaign. This is the step that turns a generic answer into a recommendation: switch off the bad channel, rather than vaguely 'improve activation'.
- 7
Close the loop with unit economics
Tie the funnel back to money. Compute effective CAC per retained user (total acquisition spend divided by users who survive to your retention stage, not by installs), then LTV as contribution margin per order times order frequency times expected lifetime, then CAC payback period in months. A funnel fix is only worth doing if it moves LTV to CAC in the right direction — that is the sentence the interviewer is waiting for.
- 8
Prioritise two or three interventions and state what you would test
Rank fixes by expected impact against effort, pick the top two or three, and define the experiment: what you change, the metric it should move, the size of move that counts as success, and how long the test runs. Quantify the prize — 'lifting month-3 retention from 20 to 30 percent adds X thousand active users at zero extra marketing spend' — so the recommendation carries a number.
Worked example
GharSe is a quick-commerce grocery app operating in Jaipur and Indore. It has raised a Series A and spends 5 crore rupees a month on performance marketing. Leadership is proud of 20 lakh cumulative app installs, but the board is alarmed: monthly revenue has been flat for two quarters and cash runway is 11 months. The CEO's proposed fix is to raise the marketing budget to 8 crore. You are asked whether that is the right move.
Step 1 — Define the stages for GharSe
Acquisition = app install. Activation = first order successfully delivered within 7 days of install. Retention = at least one order placed in month 3 after install. Referral = a referral code redeemed by a brand-new user. Revenue = contribution margin generated per user per month, after delivery cost and discounts.
Step 2 — Pull the numbers for one monthly cohort
Installs: 2,00,000. Activated (first order within 7 days): 60,000. Retained (ordered in month 3): 12,000. Referring users: 720. Economics: average order value 450 rupees at 12 percent contribution margin, so about 54 rupees of margin per order; a retained user orders 4 times a month, giving roughly 216 rupees of margin per retained user per month. Marketing spend behind this cohort: 5 crore rupees.
Step 3 — Convert to step-to-step percentages
Install to activation = 60,000 / 2,00,000 = 30 percent. Activation to month-3 retention = 12,000 / 60,000 = 20 percent. Retention to referral = 720 / 12,000 = 6 percent. Install all the way to retained customer = 12,000 / 2,00,000 = 6 percent. Blended CAC per install = 5 crore / 2,00,000 = 250 rupees, which looks cheap — and is exactly why leadership believes acquisition is working.
Step 4 — Find the bottleneck and restate CAC honestly
Activation at 30 percent is roughly in line with Indian quick-commerce norms, but month-3 retention of 20 percent is the weakest link and well below the 35 to 40 percent a habit-forming grocery app should reach. Now recompute the real cost: 5 crore divided by 12,000 genuinely retained users is 4,167 rupees per retained customer. Against 216 rupees of monthly contribution, payback is about 19 months — longer than the runway. The bucket is leaking faster than the tap can fill it.
Step 5 — Ask why, and segment
Splitting the cohort reveals two things. Users who redeemed the 99-rupees-off coupon on staples (atta, milk, oil) retain at 34 percent, while users whose first basket was impulse snacks retain at 9 percent — the staples buyer forms a weekly habit, the snack buyer does not. Separately, users whose first three deliveries averaged over 20 minutes retain at less than half the rate of users under 15 minutes, and both slow pincodes sit inside one under-supplied Indore dark-store catchment.
Step 6 — Recommend, quantify, and answer the CEO's question
Do not raise the budget to 8 crore. Three moves instead: steer first-order coupons toward a staples basket rather than blanket discounts; open one more dark store in the slow Indore catchment to pull delivery time under 15 minutes; and launch a weekly staples subscription to convert habit into a default. If month-3 retention moves from 20 to 32 percent, the same 5 crore yields about 19,200 retained users instead of 12,000, cutting effective CAC from 4,167 to about 2,600 rupees and payback from roughly 19 months to about 12 — with zero extra marketing spend.
Takeaway: Cheap installs disguised an expensive business. Because AARRR forces you to read the funnel as conversion rates rather than counts, the true cost of a customer turned out to be 4,167 rupees, not 250 — and the highest-return investment was fixing retention, not buying more traffic. More acquisition on top of a broken retention rate just buys the same leak at a larger scale.
More worked examples
Worked example: Duolingo, why the growth team reads the funnel backwards+
Duolingo is a freemium language-learning app with a very large free user base and a small paying minority on Super/Max subscriptions. It is unusual among consumer apps because most of its installs are not bought: they come from branded and category search in the app stores plus an enormous amount of earned social attention (the green owl's meme accounts, streak-shaming jokes, year-in-review screenshots). Imagine you are on its growth team and the mandate is to double subscription revenue over two years without a big jump in paid user-acquisition spend. All figures below are publicly reported orders of magnitude or clearly illustrative estimates, not audited numbers.
Monthly active users (approx, public)
~130 million
Daily active users (approx, public)
~40 million
DAU/MAU stickiness (approx)
~30 percent
Paying subscribers (approx, public)
~10 million
Paid as percent of MAU (illustrative)
~8 percent
Acquisition: define the stage, then split the channel mix before judging it
Define acquisition as a new app install or web signup, counted once per person, not as an ad impression or a site visit. Split installs into four buckets and price each one: branded organic (someone searching the word Duolingo), non-branded organic or ASO (someone searching learn Spanish app), earned social and PR, and paid UA. Duolingo has publicly said the large majority of its installs are organic, which means blended CAC is close to zero and rising paid spend has very little headroom to move the total. The correct diagnostic conclusion at this stage is uncomfortable but important: acquisition is not a lever here because it is already both cheap and near-saturated in its main search terms, so anything you do to grow must come from the four stages below it.
Activation: the aha moment is the first finished lesson, not the created account
The wrong definition of activation is account created; the right one is first lesson completed and first XP awarded, because that is the first moment the product visibly does its job. Duolingo publicly restructured onboarding so a new user can complete a lesson before being asked to register, which reverses the usual order of cost and value. The arithmetic is why it matters: an email wall placed before the first lesson typically converts roughly 40 to 50 percent of installs (illustrative), while letting the user answer five questions first and asking for the account afterward converts a much larger share, because by then the user has something to lose. Also track time-to-value: a three-minute first lesson is a far better activation gate than a fifteen-minute placement test, which is why the placement test is optional and buried.
Retention: this is the engine, and the maths explains why
Define retention as returning and completing a lesson on day 7 and day 30, then read the cohort curve rather than a single number. Every famous Duolingo mechanic is a retention mechanic and not an acquisition one: the daily streak, streak freeze, push notifications timed to the user's habitual lesson slot, weekly leagues that reset, and friend quests. The visible outcome is a DAU/MAU ratio around 30 percent, which is roughly two to three times what a typical consumer education app achieves, and it is the reason DAU has grown faster than installs. Model it as steady-state DAU approximately equals daily installs multiplied by average days retained: with installs flat, lifting D30 retention from 12 to 14 percent lifts steady-state DAU by roughly 17 percent, and that increase compounds every month at zero marketing cost.
Referral: measure it honestly and admit it is the weak stage
Referral is structurally weak for Duolingo compared with a Dropbox-style loop, because there is no high-value reward that a referrer can only get by inviting someone. The nearest analogues are friend streaks, leaderboards that are more fun with people you know, family-plan seat sharing, and the shareable year-in-review and streak-milestone cards. Measure it properly: invite send rate times invite accept rate gives the viral coefficient k, and for an app like this k is almost certainly well under 0.2, meaning each user brings in fewer than one fifth of a new user and growth is nowhere near self-sustaining from referral alone. The candidate-differentiating observation is that earned brand social does the job referral does elsewhere, so recommending a generic refer-and-earn programme would be adding a weak loop on top of a strong one instead of feeding the strong one.
Revenue: freemium conversion, and the constraint that pricing must respect
Revenue has four lines: subscriptions (Super and the higher-priced AI tier), ads served to free learners, in-app purchases such as gems and streak freezes, and the separate Duolingo English Test. Paid conversion of roughly 8 percent of MAU is genuinely strong for consumer freemium, where 2 to 5 percent is more common, but ARPU across the whole base is only a few dollars a year, so the model only works because the marginal cost of one more free learner is near zero and retention keeps the base engaged for years. This creates the central trade-off: every monetization lever that could lift conversion (hearts that block lessons, tighter paywalls, more aggressive ad load) taxes the very retention engine that produces the audience. Any pricing test must therefore be judged on retention-adjusted revenue over a 90-day window, not on the first-week conversion lift, because a paywall that adds 5 percent conversion and costs 3 points of D30 is value-destroying.
Diagnose and act: rank the levers against the mandate
Stack the five stages against the goal of doubling subscription revenue: acquisition is capped (already organic and saturated), activation is already well optimised by the lesson-before-signup order, and referral is structurally limited by the absence of a strong incentive. That leaves two real levers, retention depth and revenue per retained user, and they multiply rather than compete. On retention, deepening the streak and league loop to move D30 from 12 to 14 percent raises steady-state DAU roughly 17 percent, and because paid conversion happens as a rate on engaged users, most of that flows straight into subscriptions. On revenue, tiering upward (moving a slice of Super subscribers to a higher-priced AI tier) and expanding seats per payer through family plans both raise ARPU without touching the free experience, which is the only way to grow money without damaging the engine.
Takeaway: Duolingo is the textbook argument for running AARRR in RARRA order. Read top-down, the funnel looks like it needs more acquisition; read bottom-up, acquisition turns out to be an output of retention, because streaks and leagues are what generate the memes, the branded search and the near-zero CAC in the first place. The framework's real output here is a resource-allocation rule: fund the retention loop and the price ladder, do not fund a referral programme or more paid UA, and judge every monetization experiment on retention-adjusted revenue rather than conversion lift.
Worked example: PaisaPath, a Tier-2 SIP investing app burning cash on installs+
PaisaPath is a mutual-fund investing app targeting first-time investors in Tier-2 and Tier-3 cities (Nashik, Raipur, Jalandhar, Warangal). It runs vernacular video ads on short-video and social platforms with a 100 rupees cashback hook, and spends 6 crore rupees a month on user acquisition. The founders are pitching a Series B on the strength of 5 lakh installs a month at a 120-rupee blended cost per install, which they call the cheapest acquisition in Indian wealthtech. A board member asks you a simpler question: how many of those installs are actually earning the company money, and when does the 6 crore come back? All figures below are illustrative case numbers, not real company data.
Installs in the cohort (illustrative)
5,00,000
Signup to KYC completion
42 percent
SIP still live in month 6
33,000 users
Effective CAC per retained user (est.)
~1,818 rupees
Trail revenue per retained user (est.)
~4 rupees a month
Step 1: define the five stages in PaisaPath's own language
Generic labels are worthless here because Indian fintech has a compliance step that most funnels do not. Acquisition equals app install. Signup equals mobile OTP verified and PAN entered. Activation is a two-part gate: KYC completed (video KYC or Aadhaar-based eKYC, plus bank penny-drop verification) and then a first SIP mandate live within 30 days, because a KYC-complete user with no mandate earns the company nothing. Retention equals the SIP still successfully debiting in month 6, which is the industry's real test given how visible SIP stoppage ratios are. Referral equals a referral code redeemed by a new user who themselves completes KYC, and Revenue equals distribution trail earned per user per month. State the two-part activation gate out loud, because splitting KYC from first SIP is what will locate the leak.
Step 2: pull one cohort's absolute numbers
Take the January install cohort and follow the same people forward rather than comparing five monthly snapshots. Installs 5,00,000 against 6 crore rupees of spend. Signups (OTP plus PAN entered) 2,25,000. KYC fully completed 94,500. First SIP mandate live within 30 days 58,000. SIP still debiting in month 6, 33,000. Referral codes redeemed by a new KYC-completed user, 2,000. Average SIP ticket 1,200 rupees a month, which is typical for a first-time Tier-2 investor, and the app earns roughly 0.6 percent a year of trail on regular-plan assets under management.
Step 3: convert to step-to-step percentages and name the bottleneck
Install to signup is 2,25,000 divided by 5,00,000, which is 45 percent. Signup to KYC complete is 94,500 divided by 2,25,000, which is 42 percent. KYC to first SIP is 58,000 divided by 94,500, which is 61 percent. First SIP to month-6 survival is 33,000 divided by 58,000, which is 57 percent. Referral is 2,000 divided by 33,000, which is 6 percent, and install all the way to a paying, retained investor is 33,000 divided by 5,00,000, or 6.6 percent. The lowest conversion by a clear margin is signup to KYC at 42 percent: more than half the people who have already typed in their PAN, which is a high-intent act, never finish. Say it plainly to the interviewer: acquisition is not the problem, a compliance-and-onboarding leak is.
Step 4: ask why the KYC leak exists, then segment before concluding
Numbers say where, screen recordings and exit calls say why. Two diagnostics do most of the work here. First, segment by acquisition channel: the 100-rupees-cashback campaign on short-video apps supplies roughly 40 percent of installs but converts signup to KYC at only about 19 percent, while organic app-store search on terms like SIP investment app converts at about 71 percent, so a single channel is dragging the blended number down and the cashback is selecting for people who want 100 rupees, not a SIP. Second, segment by device and screen: the video-KYC liveness step fails disproportionately on entry-level Android front cameras and in poor evening light, and the two highest-abandonment screens are the PAN-versus-bank name mismatch error and the nominee plus FATCA declaration form. A third of stuck users are simply blocked on penny-drop verification because their account is not net-banking enabled.
Step 5: close the loop with unit economics, and find the real problem
Effective CAC is not 120 rupees. It is 6 crore divided by the 33,000 users who actually retained, which is roughly 1,818 rupees per genuinely retained investor, fifteen times the headline number. Now compute what a retained investor earns: at a 1,200-rupee SIP, average AUM in year one is only about 7,800 rupees, and 0.6 percent trail on that is roughly 47 rupees for the whole year, or about 4 rupees a month. Even letting AUM compound, cumulative trail over three years is roughly 47 plus 140 plus 259, about 446 rupees, against an 1,818-rupee acquisition cost, so the company recovers barely a quarter of CAC in three years. This is the moment the case turns: the funnel has a bad step, but the business has a revenue-stage problem that no amount of funnel repair can fix.
Step 6: quantify the funnel fix, then show why it is not sufficient
Fix the KYC stage with four cheap moves: switch off the cashback channel and redeploy that spend, prefill name and address from the PAN and CKYC record so the mismatch error disappears, offer an Aadhaar OTP eKYC fallback whenever liveness capture fails twice, and put a same-day vernacular callback on every user stuck for more than 24 hours. If signup to KYC rises from 42 to 65 percent, the same 2,25,000 signups yield about 1,46,000 KYC-complete users, about 89,000 first SIPs at the existing 61 percent, and about 51,000 retained investors at the existing 57 percent. Effective CAC falls from about 1,818 rupees to about 1,180 rupees per retained user, a 35 percent improvement at zero extra marketing spend, which is a genuinely large win. But 1,180 rupees against 4 rupees a month of trail is still nowhere near payback, so the funnel fix alone does not make the business investable.
Step 7: recommend, with the revenue redesign as the headline
Three recommendations, ranked. One, fix KYC as above, because it is fast, cheap and worth 18,000 extra retained investors per cohort. Two, redesign the revenue stage, which is the actual constraint: nudge step-up SIPs and goal-based defaults to lift the average ticket from 1,200 to about 2,000 rupees, and cross-sell a 99-rupee-a-month premium plan (research, model baskets, tax reports) to the roughly 20 percent of retained users who already log in weekly, which together take blended revenue per retained user from about 4 rupees to roughly 26 rupees a month and pull payback from effectively never to somewhere around 45 months. Three, cut paid spend by half and redirect it to referral and campus or employer payroll channels, where fully loaded cost per acquired investor is closer to 350 rupees, and where the 6 percent referral rate says the product already has advocates who are simply not being asked. Tell the board the honest version: PaisaPath should not raise money to buy more installs until revenue per retained user has been fixed, because right now every additional crore of spend enlarges a loss.
Takeaway: The 120-rupee CAC the founders were pitching was a vanity number; the real cost of a retained investor was about 1,818 rupees, and the app earned roughly 4 rupees a month from that person. Running AARRR located a genuine and fixable leak at KYC, worth a third off CAC for almost no money, but its more valuable output came from the last stage: when lifetime revenue is that far below acquisition cost, the correct recommendation is to fix monetization and shift to cheaper channels, not to fund more growth. The general rule a student should carry away is that you have not finished an AARRR analysis until you have divided total acquisition spend by retained users rather than by installs, and compared that to lifetime contribution.
Common pitfalls
- •Reciting the five stages as a list instead of computing the conversion between them. The framework has no diagnostic value until you divide each stage by the one above it and point at the lowest number. A candidate who names all five and then discusses all five equally has added nothing.
- •Using absolute counts as evidence of health. 20 lakh installs, 50,000 signups, 1 crore registered users — these are vanity metrics and interviewers know it. Always ask what percentage of the previous stage that number represents, and over what time window.
- •Mixing snapshots instead of following a cohort. Comparing this month's installs against this month's retained users blends people acquired years apart and produces conversion rates that are simply wrong. Say explicitly that you want a single cohort tracked forward in time.
- •Reporting a blended average and stopping there. Activation of 30 percent can be 55 percent organic and 12 percent from one bad paid channel. Without segmenting by channel, city, device and cohort you will recommend a product fix when the real answer is to switch off a campaign.
- •Treating the order as sacred in a saturated market. Pouring acquisition budget into a product with 20 percent retention is the classic leaky-bucket error — the RARRA critique exists precisely for this. Check retention before you agree to spend more on acquisition. And never skip Revenue: an answer that never computes contribution margin, LTV to CAC or payback is incomplete.
- •Forcing AARRR onto cases with no repeat relationship. For a one-time high-ticket purchase, a B2B tender, a cost-reduction problem or a supply-chain question, this framework makes you look like you are pattern-matching rather than thinking.
Interview tips
- •Open by defining the stages in the client's own vocabulary — 'for this app I'll treat Activation as first order delivered within 7 days' — and ask the interviewer to confirm. It takes 20 seconds, shows business judgment, and stops you from analysing the wrong thing for ten minutes.
- •Write the funnel vertically with two columns: the count on the left, the conversion percentage on the right. Circle the lowest percentage as soon as you have it. Interviewers explicitly reward the candidate who narrows to one bottleneck instead of spreading effort across all five stages.
- •Restate CAC in terms of retained users, not installs. Total spend divided by the number of users who survive to your retention stage is a number most candidates never compute, and it usually reframes the entire case in one line.
- •Say the word 'cohort' and mean it. Mentioning that you would compare the month-1 and month-6 cohorts, or split by acquisition channel and city, signals you have thought about how the data is actually pulled and not just about the acronym.
- •Bring up the RARRA counterargument once, briefly, when retention is weak: 'in a market this competitive I'd fix retention before adding acquisition budget, otherwise we are just filling a leaky bucket faster.' One sentence, not a lecture — it reads as practitioner experience.
- •Finish with an experiment, not an opinion. Name the change, the metric it should move, the target size of the move, and the read window. 'Steer first-order coupons to staples, target month-3 retention from 20 to 30 percent, read the cohort at 90 days' beats any amount of qualitative recommendation.
Test yourself
Best video explainers
Pirate Metrics for Startups - AARRR Framework by Dave McClure
Headway
The cleanest short overview of the original framework, walking through each of the five stages the way McClure defined them — start here if you have never seen AARRR.
Key Metrics for Pirates and Product Managers: explaining the AARRR framework
Product Teacher
Framed for product-management interviews, with guidance on which metric to pick at each stage and how PMs are expected to talk about them.
What are Pirate Metrics? AARRR framework in 3 minutes
Product Chat
A three-minute refresher — the right thing to watch the night before an interview when you just need the structure locked in.
AARRR Framework: Improve Your Growth Funnel [With Pirate Metrics]
Oskar Bader | DesignWithValue
Goes past definitions into how to actually improve each stage of the funnel, which is where most case answers need to land.
AARRR Framework Product Management with Example
Yash Thakker
An Indian PM working through the framework on a concrete product example — useful for hearing the vocabulary in an Indian consumer-internet context.
Go deeper
Startup Metrics for Pirates (original deck)
Dave McClure on SlideShare
The 2007 presentation where AARRR was coined, still free to read. Citing the primary source in an interview is a cheap credibility win.
AARRR: Come Aboard the Pirate Metrics Framework
Amplitude
Written by a product-analytics company, so it names the exact metrics to track at each stage — CAC, activation rate, time-to-activate, churn, NRR, ARPU.
AARRR Pirate Metrics Framework — Glossary
ProductPlan
A tight, jargon-free definition of each of the five stages with example metrics. Good for a fast revision pass the day before a round.
AARRR vs RARRA: Pirate Metrics Explained
Mind the Product
Explains the retention-first critique of AARRR and when each ordering applies — the nuance that separates a good case answer from a textbook one.
What is the Pirate Funnel (AAARRR) + How to apply it in 5 quick steps
Grow with Ward
The most practical guide to the bottleneck method: fill in the numbers, compute step-to-step conversions, attack the lowest one. Exactly the mechanic you use in a case.
Now use it on a real case
Reading a framework isn't the same as applying it under pressure. Practise with an AI interviewer that pushes back.
Practise a case free