How to find growth where everyone else misses it

Every founder says they run experiments. Most don't.

They swap a headline, recolour a button, wait a week, then crown a winner off the back of a 3% lift. That isn't experimentation. It's guessing with extra steps.

The single most useful idea I took from Hacking Growth by Sean Ellis and Morgan Brown wasn't a clever tactic. It was a reframe: growth is a process for learning faster than your competitors. The companies that own their categories rarely have better ideas. They generate better evidence.

And after building SaaS products myself, I've learned that the biggest wins almost never come from the homepage. They come from fixing leaks all the way down the funnel.

The mistake almost everyone makes

Nearly every founder is obsessed with acquisition. More traffic. More Meta spend. More Google budget. More SEO.

Meanwhile their onboarding converts at 18%.

Think about that. Pouring another £50,000 into ads while 82% of new users vanish before they ever feel the product work. That's not a growth strategy. It's a leaky bucket with a bigger tap.

Ellis frames growth as improving every stage of the customer journey, not just the top of it. That means running experiments across acquisition, activation, retention, revenue, and referral. If the only thing you're testing is your landing page, you're ignoring roughly 80% of the opportunity in front of you.

The question I ask instead

Most people ask, "How do I increase conversions?"

I ask, "Where are people deciding not to continue?"

A funnel is just a sequence of decisions. Someone decides to click the ad, create an account, verify their email, finish onboarding, connect their site, experience value, subscribe, and, eventually, stay. Every one of those decisions can be influenced. But you can't influence what you haven't found.

So I map the funnel before I touch anything

Before I write a single experiment, I lay out every event in the journey:

Ad click → landing page → signup → email verification → intent selection → onboarding → website connected → first value → checkout → payment → day 7 active → day 30 active

Now I can see exactly where people disappear. If 80% reach onboarding but only 15% connect their website, I don't have a traffic problem. I have an onboarding problem. More visitors would just mean more people quitting at the same step.

The trick nobody talks about: test bigger things

Most A/B tests fail because they're too timid. Button colours, font sizes, rounded corners. These almost never move revenue.

Instead, I look for the moments where users are making a meaningful decision, and I test the decision itself.

Not blue button vs. green button, but AI onboarding vs. manual onboarding.

Not two pricing headlines, but asking for payment before onboarding vs. after.

Not two hero images, but showing the product immediately vs. asking questions first.

Experiments like these can shift conversion by double digits. Button colours shift it by rounding error.

My favourite onboarding experiments

The highest-leverage tests are often the simplest.

Remove fields. Every extra field costs you users. Make company name optional, drop phone number, kill the password confirmation, add social login.

Delay friction. Don't demand five integrations up front. Let people feel value first. Dropbox won partly because users experienced the product almost immediately after signup, and it sold itself.

Reduce decisions. Choice creates hesitation. Instead of "What do you want to do?", show one obvious next step and guide them to it.

Add a progress bar. People like finishing things. Completion rates climb for no reason other than users understanding how much is left.

Celebrate the win. "Connected" is a status update. "Your first report is ready" is a payoff. Tiny wording changes can transform perceived value.

The framework behind every test

Every experiment I run follows the same shape:

• Observation: Users abandon onboarding right after connecting their website.

• Hypothesis: They don't understand what happens next.

• Experiment: Generate a report immediately instead of showing a loading screen.

• Success metric: Activation increases by 15%.

The metric matters. "Let's see what happens" isn't an experiment. It's a vibe. Every test should carry a prediction you're willing to be wrong about.

ICE scoring saves you from yourself

Ellis popularised ICE for prioritising the backlog: score each idea 1 to 10 on Impact, Confidence, and Ease.

Guess which one I run first. It's the top row, every time. Fast, cheap learning beats a perfect idea that takes a month to ship.

Optimise for learning velocity, not win rate

Here's the reframe that stuck with me most: winning experiments are nice, but learning is the real product.

If I run 20 experiments and only 3 win, that's a great month, because now I know 17 things that don't matter. Fast-growing companies don't necessarily have higher hit rates. They simply test more, and they build the machinery to keep testing: a cross-functional team, a regular cadence, and a backlog so the learning compounds week over week.

But don't stop tests early

This one stings. You launch a test, Variant B is up 18% after three days, and every instinct screams ship it. A week later, the two variants are identical.

This happens constantly. "Peeking" at results before enough data has accumulated is one of the fastest ways to fool yourself, which is exactly why modern experimentation platforms lean on sequential methods designed to guard against false positives from early stopping. Patience isn't a virtue here. It's a P&L line.

The metrics I actually watch

I rarely care about click-through rate. I watch the transitions that explain revenue:

Visitor → signup, signup → activation, activation → subscription, subscription → week 1 retention, month 1 retention, expansion revenue, LTV:CAC, and payback period.

Revenue is the metric. Everything else is just a story about why revenue is what it is.

The backlog never ends

Every time I spot friction, it becomes a line in the backlog: remove the pricing page entirely, auto-fill company details, generate a report before signup, replace onboarding with an AI chat, swap the demo for an interactive tour, move social proof later, ask one question instead of five, put testimonials beside checkout, let people skip setup completely.

Most will fail. A couple will change the company. You almost never know which in advance, and that is the whole point of testing.

The real lesson

The biggest thing Hacking Growth taught me wasn't how to run a better A/B test. It was that growth isn't a single magic trick waiting to be discovered. It's a system that keeps discovering what works, and keeps compounding what it learns.

The best SaaS companies aren't smarter than everyone else. They just run more high-quality experiments and learn faster than the competition can react.

If I could give a founder one piece of advice, it's this: stop trying to be right, and start trying to learn.

Because in growth, the company that learns fastest usually wins.