Every guide on this term tells the same four stories. Dropbox gave away storage for referrals. Airbnb posted its listings to Craigslist. Hotmail added a line to the bottom of every email. PayPal paid people to sign up.
They are good stories. They are also fifteen to twenty five years old, and three of the four cannot be repeated today, which nobody mentions.
This page covers what growth hacking actually means, the frameworks that hold up, and then a section none of the ranking pages have: which of the famous hacks would still work if you tried them this year, and what replaced the ones that would not.
How this guide was put together
I read the pages currently ranking for this term, from Demand Curve, Stripe, GrowthHackers, Optimizely and Ward van Gasteren. They agree on the definition and on the case studies, and between them they cover the pirate funnel, the experiment loop, and the difference between growth hacking and marketing. This guide covers all of that, because a reader needs it.
What none of them do is check whether the famous examples are still available to you. That check is the second half of this page.
The definition
Growth hacking is a way of finding customers through fast, cheap, measurable experiments rather than through budget.
The phrase comes from Sean Ellis, who used it in 2010 to describe the kind of person he wanted to hire: someone whose only objective was growth, regardless of which department the work belonged to. He had done this job at Dropbox, LogMeIn and Eventbrite, and found that the usual marketing job description did not describe it.
Three things separate it from marketing as normally practised:
- The unit of work is an experiment, not a campaign. You are trying to learn something in a week, not to run something for a quarter.
- The scope crosses departments. A growth hacker changes the signup form, the onboarding email and the pricing page, which in most companies belong to three different teams.
- The constraint is time and creativity rather than money. The technique exists because startups had no budget. When you do have budget, most of this becomes ordinary paid acquisition.
It is not a synonym for cleverness, and it is not a set of tricks you can copy. It is a habit of running many small tests and keeping the ones that work.
Growth hacking against marketing
The distinction gets overdrawn in most explanations. They are not opposites. Growth hacking is a narrower activity that sits inside marketing and borrows from product and engineering.
| Traditional marketing | Growth hacking | |
|---|---|---|
| Time horizon | Quarter or campaign | One to two weeks per test |
| Main lever | Budget and reach | Product changes and mechanics |
| Success measure | Impressions, leads, brand lift | One metric, moved or not moved |
| Owns | Channels | The whole funnel |
| Fails by | Spending on the wrong audience | Optimising something that does not matter |
The last row is the one worth remembering. Marketing fails expensively and visibly. Growth hacking fails quietly, by running fifty clean experiments on a part of the funnel that was never the problem.
What a growth hacker actually does
The job is roughly four things, in a loop.
Finds the bottleneck. Before any test, work out where people are actually falling out. Most teams optimise acquisition because it is the most visible stage, when the leak is in activation.
Generates candidate experiments. Usually far more ideas than can be run, from support tickets, session recordings, competitor teardowns and customer calls.
Prioritises them. Every framework here is a way of scoring impact against confidence against effort. ICE and PIE are the common ones. The scoring is rough by design; its job is to stop the loudest person choosing.
Runs the test and reads it honestly. This is where most of it falls down, and the reason has nothing to do with creativity. See the section on sample size below.
The skill mix is unusual: enough analytics to read a result, enough technical ability to ship a change without waiting three weeks, enough copywriting to write the variant, and enough discipline to kill your own idea.
The pirate funnel
The AARRR framework, from Dave McClure, is the standard map. It is called the pirate funnel because of the acronym.
| Stage | The question it answers | A typical metric |
|---|---|---|
| Acquisition | Do people arrive? | New visitors, signups by channel |
| Activation | Do they get value the first time? | Percentage reaching the first useful action |
| Retention | Do they come back? | Week four retention |
| Referral | Do they bring others? | Invites sent per active user |
| Revenue | Do they pay? | Conversion to paid, average revenue per account |
The order matters more than the labels. Fixing acquisition while activation is broken means paying to fill a bucket with a hole in it, and the numbers will look like progress for about a month.
There is a widely repeated claim that improving retention is worth more than improving acquisition. It is true for subscription businesses and it is roughly the whole game there, but it is not a universal law. If you sell something bought once every four years, retention is the wrong stage to obsess over.
Tools for measuring this are ordinary product analytics, not anything specialised.
Any of the three will build the funnel view. PostHog has a free self-hosted tier, which matters when you are pre-revenue and the point of the exercise is not spending money.
The experiment loop
Every version of this process is the same five steps under different names.
1. State the belief. "New users do not understand what to do first." Not "let us test a new onboarding." 2. Predict the number. "Adding a checklist will take first-week activation from 22 percent to 30 percent." Write the number down before you run it. This single habit does more for honesty than any framework. 3. Build the smallest version. Not the good version. The version that tests the belief. 4. Run it long enough. Covered below, and it is the step everyone skips. 5. Decide, then write it down. Kept, killed, or inconclusive. A record of dead experiments is the actual asset a growth team builds, because it stops the same idea being re-proposed every six months.
The sample size problem nobody mentions
Most growth hacking content skips straight from "run a test" to "scale the winner". In practice, most small companies cannot run a valid A/B test on their conversion rate at all, and this is arithmetic rather than pessimism.
To detect a change from 3 percent to 4 percent conversion with any confidence, you need roughly 7,000 visitors per variant. At 500 visitors a week per variant, that test takes about seven months, by which point the seasons, the traffic mix and the product have all changed.
Two consequences follow, and they shape everything else on this page:
- Below a few thousand conversions a month, prefer changes big enough to see without statistics. Moving a rate from 2 percent to 6 percent is visible. Chasing a 0.3 point lift is not.
- Stopping a test the moment it looks significant is the most common error in the discipline. If you check daily and stop on the first good day, you will find winners in pure noise more often than not.
Optimizely and similar platforms will calculate the required sample for you before you start, which is the most useful thing they do.
The famous hacks, and whether they still work
Here is the part missing from the pages that rank for this term. Every guide retells these four stories as if they were still options.
Hotmail's email signature: still works
Adding "P.S. I love you. Get your free email at Hotmail" to the bottom of every outgoing message took the service from 20,000 to 1 million users in about six months in 1996. The modern equivalent of that trick is a well-written first email, and our sales email blueprint covers what still gets replies.
Still available. This is the ancestor of every "sent with" and "made with" badge in software today, and it still works because it is honest, it costs nothing per send, and the referral arrives with implicit endorsement. Loom, Calendly and Typeform all built substantial user bases this way. If you have any artefact that leaves your product and lands in front of someone else, this is the first thing to try.
Dropbox's referral programme: still works, with a caveat
Two-sided referrals, 500MB of storage for each side, roughly 60 percent more signups.
Still available, but the economics have changed. Dropbox's reward cost close to nothing to deliver and was the exact thing the user wanted more of. If your reward is a discount on a low-margin product, you are buying customers at a price you should compare against paid acquisition rather than treating as free. Referral works best when the reward is more of your own product and your marginal cost is near zero.
Airbnb's Craigslist integration: no longer possible
Airbnb built a tool that cross-posted listings to Craigslist, borrowing an audience it did not have. It is the most cited hack in the genre.
Not repeatable. Craigslist closed the hole, and doing this today means unauthorised automated posting against a platform's terms of service. Beyond the legal exposure, every large platform now detects and blocks this class of automation. Guides that still present this as a tactic are describing something that would get your accounts banned.
What replaced it: legitimate integrations and marketplace listings. Being in an app directory, a partner ecosystem or an aggregator is the same idea, borrowing someone else's distribution, without the part that gets you sued.
PayPal paying for signups: mostly not
PayPal gave 10 dollars for signing up and 10 for a referral, burning roughly 60 million dollars to buy its user base.
Available only if you can afford it, which makes it the opposite of growth hacking. It was venture-funded customer acquisition. It worked because PayPal had strong network effects and a clear path to making that money back per user. Copying the mechanic without the network effect is just paying strangers to open an account they will never use.
The pattern
The hacks that survive are the ones that create value for the person doing the spreading: a genuinely useful free tool, a referral reward the user actually wants, a badge on work they are proud of. The ones that died all borrowed an audience without permission, and platforms spent the last fifteen years closing those holes.
That is the honest state of growth hacking in 2026. The channels are all metered now. What remains is product mechanics and the search and content surfaces, which is where most of the durable work has moved.
Where the durable growth work sits now
With the loopholes closed, three areas still compound rather than requiring continuous spend.
Product-led mechanics. Anything that makes your existing users produce more users: shared documents, invitations that are part of the job, public output with your name on it.
Search and content. Slow, compounding, and the only channel where the asset keeps working after you stop paying. It has become considerably more competitive, and now includes being cited by AI assistants as well as ranking in search results. Our own tool sits here: Distribb publishes SEO articles on a schedule and runs a backlink exchange, which covers the mechanical half of this channel.
The honest limitation: it will not tell you whether search is the right channel for your business at all, and for many products it is not. If your customers do not search for the problem you solve, a content engine produces traffic that never converts, and running one is a slower and more expensive mistake than not running one. Work out the bottleneck first, which is the entire point of the pirate funnel.
Community and word of mouth. Unscalable by definition, which is why it keeps working when the scalable channels get crowded.
For the acquisition maths behind these choices, our breakdown of what leads actually cost by channel gives the benchmarks to compare against. If you want the experiment discipline applied specifically to conversion paths, the funnel hacking playbook goes deeper on that stage. And for one channel where borrowed audience is still legitimate, TikTok influencer marketing is the modern version of what Airbnb was doing on Craigslist, with the permission included.
Starting from zero: the first month
If you are setting this up rather than reading about it, this is the shortest useful sequence.
Week one: instrument the funnel. You cannot find a bottleneck you cannot see. Get the five AARRR stages into one dashboard, even a rough one. Most teams discover at this point that they do not know their activation rate.
Week two: find the leak. Compare each stage against a sensible benchmark for your model. The worst-performing stage is where every experiment goes until it stops being the worst.
Week three: list twenty ideas and score them. Impact, confidence, effort, one to ten each. Take the top three. The scoring is crude and that is fine.
Week four: run one properly. One test, with a written prediction, run to a pre-decided sample size or duration. One test run honestly teaches you more than five run badly.
Then repeat. The compounding comes from the loop running every week for a year, not from any individual clever idea.
Common questions
Is growth hacking still relevant in 2026? The mindset is. The specific tactics from the 2010s mostly are not, because the platforms they exploited closed the gaps. Treat it as a way of working rather than a list of moves.
Do I need to be technical? It helps a great deal and is not strictly required. What is required is being able to ship a change without a three-week queue, whether that comes from writing the code yourself or from no-code tools and a cooperative engineer.
How is it different from conversion rate optimisation? CRO is growth hacking applied to one stage, usually the page before signup. Growth hacking covers the whole funnel including retention and referral, which is generally where the larger wins are.
What is a reasonable success rate for experiments? At mature companies with strong experimentation programmes, roughly one in seven to one in ten tests produces a meaningful win. If most of your tests are winning, you are probably reading noise or testing things that were already obvious.
Can it work for a business without a product to modify? Partly. Service businesses and ecommerce stores can run the acquisition, activation and referral parts. The product-mechanic experiments that produce the famous results need a product with users inside it.
What tools do I need to start? Product analytics, a way to run a split test, and a spreadsheet for the experiment log. The spreadsheet is not the joke part; keeping the record is what separates a programme from a series of hunches.
What to take from this
Growth hacking is a loop: find the bottleneck, guess, test cheaply, keep what works, write down what did not. The frameworks are decoration on that loop. The tactics that survive that loop today are collected in our list of growth hacking strategies.
The famous stories are worth knowing as history and are mostly dead as tactics. What still works is the class of mechanic that gives your users a reason to spread your product, plus the slow compounding channels that nobody can lock you out of.
If the search and content half is the part you want handled mechanically, that is what Distribb does: scheduled publishing and a backlink exchange, so the compounding channel runs without a person driving it every week. Decide it is your bottleneck first.