Every test, whatever the channel, follows one shape: a hypothesis with a number, a fixed budget and an end date, one change at a time, a read against the number, and a decision. Startups that keep that shape find their channel; startups that chase tactics do not.
The order matters: the pages and tracking first, then channel tests two at a time, then funnel tests on the winning channel's traffic, then retention. Our startups page says it in one line: small test budgets on the channels where your buyers already look, scaled only when the numbers hold.
What a growth test is, and what it is not
A growth test is an experiment with a written hypothesis: “If we run search ads on [problem] terms to [first buyers] with a $4,000 budget for 30 days, we will win at least six paying customers at $700 or less each.” It has a fixed budget, an end date, one change under test and a pass mark decided before it starts. When it ends, it is read against the pass mark, and the decision is made: scale, repeat with one change, or stop.
A growth test is not “trying LinkedIn for a bit”, a hack copied from another company's blog, or a campaign judged on clicks. Those produce opinions. The test produces a number and a decision, which is the only thing a startup's marketing can afford to produce in its first year.
Before the first test: the pages and the tracking
A channel test lands somewhere, and if that page does not convert, the test tells you about the page rather than the channel. Before the first test, build the smallest site that converts and the tracking that counts sign-ups, trials, demos and paying customers by source, end to end. Check that a test customer shows up correctly in the numbers before a dollar is spent. Our guide to startup website design covers the pages.
Which channels to test first
List every channel where the first buyers already look, and rank them by how directly they reach those buyers and how fast a result can be read. The ranking differs by startup, and these are the usual candidates.
- Search ads on the problem. Fast to read, expensive per click, and the clearest signal that the message matches a need. Suits products people already search for.
- Paid social to a defined audience. Reaches people who are not searching, by job title, interest or a list; needs creative that shows the product doing the job. Our paid social page covers the campaigns we run.
- Founder-led outreach. Direct messages and emails to a named list of first buyers, within CAN-SPAM in the USA and CASL in Canada. Slow to scale, fast to learn from, and the source of the words the ads will later use.
- Communities and partners. The forums, groups, newsletters and complementary products the first buyers already use. Cheap, slow, and often the best early channel for a narrow buyer.
- Marketplaces and app stores. Where the product is a plug-in or an app, the listing itself is a channel to test.
- Search and content. The slowest and the one that compounds; started early at a small pace rather than tested for a month.
How to run a channel test
Six steps, the same for every channel:
- Write the hypothesis with the pass mark: channel, buyers, budget, days, customers expected, cost per customer the business can carry (the budget guide shows the arithmetic).
- Set the budget as a fixed monthly amount. For Google Ads, the platform's help on budgets says daily spend can run to twice the daily budget and the month is capped at 30.4 times it, so set the daily figure from the monthly one and judge at month end, not day three.
- Change one thing at a time inside the test: one audience, one message, one landing page. A second variable means the result has two causes and no lesson.
- Send the platform the right conversion. Automated bidding learns from whatever you count; if you count free sign-ups, it finds people who sign up and leave. Count paying customers or qualified demos where the data allows.
- Read weekly for errors and at month end for the decision. Mid-month numbers are incomplete; platforms attribute conversions over days.
- Decide and record: scale with a larger fixed budget, repeat with one change, or stop. Write the result in the test log with the date, whatever it was.
After the channel: funnel tests
Once a channel brings visitors at a known cost, the funnel decides what a customer costs. Read it in four steps: visitor to sign-up or demo request, sign-up to activated (the first real result in the product), activated to paying, and paying to retained at month two. The step with the largest drop is the next test.
Funnel tests follow the same shape as channel tests, with the pass mark set on the step's rate. Typical first tests: the message and proof on the landing page, the number of fields in the sign-up, the first screen after sign-up, the pricing page's clarity, the onboarding email that gets a user to the first result. Run one at a time, for long enough to cover a full buying cycle and enough conversions to be sure; a startup with few customers tests big changes and skips button colors. Our conversion rate optimization page explains how we read a funnel and pick the test.
Retention is a growth test too
A customer lost in month two makes every acquisition dollar more expensive, so retention tests start the month the first customers arrive. The hypothesis is the same shape: “If we add a week-one check-in call for customers who have not reached the first result, month-two retention will rise from X to Y.” Measure cohorts by sign-up month, change one thing at a time, and treat a retention gain as worth more than an equal acquisition gain, because it compounds.
Content and search inside a growth program
Search and content do not fit a 30-day test, so they run alongside the tests at a small steady pace from month one. Write for the first buyers' exact problem, in their words, showing the product solving it. Google's helpful content guidance asks for first-hand expertise, original information and reasons to trust the author, and names trust as the most important quality; a founding team has those by definition.
Two cautions. Pew Research Center analyzed 68,879 Google searches by 900 US adults in March 2025 and found that 58% of them saw at least one AI summary that month and that users who saw one clicked result links less often, so broad explainer posts are the least likely to bring a visit; write the specific, product-adjacent pages instead. And treat search traffic as its own channel in the weekly sheet, with its own cost per customer counted in hours, so that it is judged like everything else.
When to scale, and how fast
A channel has earned scale when its cost per customer has held at or below the pass mark for three consecutive months at a volume that matters to the business. Scale it by raising the fixed budget in steps, such as a third at a time, and pause the increase the month cost per customer slips, because most channels get more expensive as they grow. Keep one new test running at all times, so that the business never depends on a single channel, and keep the weekly sheet described in the strategy guide: spend, customers and cost per customer by channel, the funnel rates and runway.
If you want it run with you
PromotEdge International is an international strategy, technology, marketing and design company, and our delivery team is in India. Our marketing for startups page explains how we run small test budgets on the channels where your buyers already look and scale only when the numbers hold, and our guide to the startup marketing budget covers the money behind the tests.
Facts we rely on: see our company facts for what we state about ourselves and where each fact comes from.