Topic 8 — Growth experiments and channels

Goal: Learn how growth happens through disciplined experiments and the right channels.

Lesson 8.1 — Stop guessing, start hypothesizing

Imogen's first week on the growth team, Hollis pulls up a whiteboard and asks her to pitch one idea to lift sign-ups on Verdana's pricing page. She says, with the confidence of a radio host reading a hot take: "Let's add testimonials from studio owners. People trust other people."

Hollis nods, then says: "Good instinct. Now say it so we can prove you right or wrong."

That second sentence is the whole job. Imogen had a guess. What Hollis wanted was a hypothesis — a guess written so the result can settle the argument. The format the team uses, the one she'll write a hundred times:

If we [change], then [metric] will [improve], because [reason].

So Imogen's hunch becomes: "If we add member testimonials to the pricing page, then sign-ups will increase, because social proof builds trust." Same idea, but now it names exactly what she'll change, the one number she's watching, and the reasoning she's betting on. When the data comes back, there's no wiggling — either sign-ups moved or they didn't.

Growth runs on this. Nobody, not Hollis, not the famous growth blogs, can reliably predict what will move the numbers. So the team treats every idea as a claim to be tested, not a decision to be defended. The "because" matters as much as the rest: it forces you to have a real theory, so that even a failed test teaches you something about your customers.

Lesson 8.2 — The loop, and why you can't test everything

Imogen comes back the next morning with eleven ideas. Hollis reads the list, smiles, and says they'll run maybe two this month.

That stings until she understands the constraint. Every test costs engineering time, design time, and traffic, and a page only gets so many visitors before a verdict is in. The art is rarely in having ideas. It's choosing which few to spend on.

This is the growth loop the team runs, over and over:

1. Identify the biggest opportunity — usually the worst leak in the funnel (Topic 7). No point optimizing a step that's already fine.
2. Brainstorm ideas to fix that specific leak.
3. Prioritize — pick the best bets for the effort.
4. Test them.
5. Analyze and learn, then feed what you learned back into step 1.

For step 3, Hollis hands her a scoring sheet called ICE — created by Sean Ellis, the marketer who coined the term "growth hacking" and ran growth for companies like Dropbox and LogMeIn. You rate each idea 1 to 10 on three things and multiply them:

• Impact — if this works, how much does it move the metric?
• Confidence — how sure are you it'll work? What's the evidence?
• Ease — how cheap and fast is it to build?

ICE = Impact × Confidence × Ease

Imogen scores the testimonials idea: Impact 7, Confidence 6, Ease 8 → 336. A "rebuild the whole onboarding video" idea scores Impact 9, Confidence 7, Ease 2 → 126 (9 × 7 × 2). Notice the trap: the video has the higher Impact and Confidence of the two, yet its rock-bottom Ease drags the whole score down — that’s the multiply doing its job, since a slow, expensive bet has to clear a high bar to be worth it. The video might win eventually, but the testimonials are a far better use of this week. ICE is not precise science. Its value is that it forces the team to argue with numbers and stops the loudest voice from always winning.

Lesson 8.3 — The A/B test, honestly

Imogen's testimonials are live. How does she actually find out if they worked?

She runs an A/B test. Verdana's tool splits incoming visitors randomly: half see the old pricing page (the control), half see the new one with testimonials (the variant). Both versions run at the same time, and the tool counts which group signs up at a higher rate.

The random split is the part that makes it trustworthy. Because visitors are assigned by coin-flip, the two groups are alike in every other way — same mix of curious browsers and ready-to-buy owners, same time of week. So if the testimonial group converts better, the testimonials are the most likely cause, not luck or timing. That's what separates a real test from "we changed it and sign-ups went up that week" (when maybe a holiday promo was also running).

Two rules Hollis drills into her, and they're the ones every experimenter learns the hard way:

Change one thing at a time. Imogen wanted to add testimonials and recolor the button and reword the headline, all in the variant. Hollis stops her. If that combined version wins, which change won? You'd never know, and you'd carry three guesses forward instead of one fact.

One change per test. A test that proves five things at once usually proves nothing.

Lesson 8.4 — Small samples lie

Day two of the test, Imogen runs in grinning. The variant is winning, 8 sign-ups to 3. She wants to call it, ship it, and move on.

Hollis asks how many visitors that's out of. About forty.

"Flip a coin forty times," he says. "Sometimes you get twelve heads in a row. That's not a magic coin — that's just a small sample being noisy." Eight versus three feels like a landslide, but with numbers that tiny, a result that big can happen by pure chance. Tomorrow it could flip the other way.

This is the intuition behind statistical significance: a result is only believable once you've gathered enough data that random noise can't easily explain it. Imogen doesn't need to compute the math herself — Verdana's testing tool flags when a result is significant — but she does need the instinct to distrust small samples. The discipline is waiting. Let the test run until you've got enough visitors and conversions for the tool to call it, then look.

A week later, the verdict: with a few thousand visitors, the testimonial page converts about 11% better, and the tool marks it significant. Now she ships it. That patience is the difference between a growth marketer and someone who chases noise.

Lesson 8.5 — Channels, economics, and loops

Sign-ups are climbing, so Hollis turns Imogen to the harder question: where do new studio owners come from, and which sources are worth paying for?

A channel is a path that brings users in — paid search ads, Instagram, content, referrals, partnerships. The growth approach to channels is the same loop as everything else: test several, measure what each one delivers, double down on the winners. But here you measure two things, cost and quality, because some channels deliver lots of cheap signups who never pay and never stay.

The two numbers that decide everything:

• CAC (Customer Acquisition Cost) — what it costs, all in, to acquire one paying customer through a channel.
• LTV (Lifetime Value) — what that customer is worth to Verdana over the whole time they stay subscribed.

A channel only makes sense if LTV comfortably exceeds CAC. Spend $300 to win a studio that pays you $360 over its life, and you're barely breaking even after support costs. The common SaaS rule of thumb is an LTV:CAC ratio of about 3:1 or better — earn at least three dollars for every dollar of acquisition cost — with CAC payback ideally under ~12 months, so you're not waiting years to recoup the spend.

Imogen tests four channels for a month. Instagram ads bring cheap signups who churn fast — LTV:CAC barely 1.5:1. A partnership with a yoga-teacher training program brings fewer leads but they stick — 5:1. The lesson lands: don't spread thin. Most successful companies have only one or two dominant channels, not ten mediocre ones. She kills two channels and pours the budget into the partnership.

Then Hollis points at the prize: a growth loop. A one-off campaign spends money, gets users, and stops the day the budget runs out. A loop feeds itself. Verdana adds a referral: any owner who invites a fellow studio owner gets a free month, and so does the friend. Now happy customers bring new customers, who become happy customers, who bring more. That compounds; a campaign doesn't.

Campaigns add. Loops multiply.

One caution Hollis adds: match the loop to the product. A viral consumer app can grow on invites alone; an enterprise tool with a six-month sales cycle never will. Verdana sits in between, so referrals plus that one strong partnership is a realistic engine, not a fantasy.

Worked example — Imogen fixes the worst leak

Topic 7's funnel review left Imogen with a clear villain: of the studio owners who start a free trial, only 30% ever create their first class schedule — and almost nobody who skips that step sticks around. Devesh, the PM, calls scheduling the "aha" moment. The leak is the trial-to-activation step.

She runs the loop.

Identify: trial → first-schedule is the worst leak. Fix that, not the pricing page again.

Brainstorm + prioritize: five ideas. She ICE-scores them. A guided setup checklist scores Impact 8, Confidence 7, Ease 7 → 392 — cheap to build, and she's fairly sure it'll help. A full onboarding video scores higher on impact but Ease 2 → too expensive for now. The checklist wins.

Hypothesis: "If we add a 3-step setup checklist to the empty dashboard, then trial activation will increase, because new owners don't know scheduling is the first thing to do."

Test: A/B test, randomly split. Control sees today's blank dashboard; variant sees the checklist. One change only — she resists also tweaking the welcome email. After three days the variant looks ahead, but it's 90 trials in, so she waits.

Analyze: at two weeks and a few thousand trials, activation rises from 30% to 41%, flagged significant. Real win. She ships it, tells Brigid (her favorite Iron Fern customer, whose confusion inspired the checklist) — and feeds the result back in: the next leak is now trial-to-paid. The loop turns again.