By role · 4 tools
The web stack for a Head of Growth
A Head of Growth owns a number, not a website. The site belongs to somebody else, so the stack is picked for what can be changed without waiting on a release: a testing tool that runs on the live page, analytics that answer the question directly, and a written record of what has already been tried.
What is a Head of Growth actually accountable for?
You are measured on signups, pipeline or revenue, and the page that produces them is owned by somebody else. The constraint is not ideas. It is getting a change onto the live site this quarter, and being able to show afterwards which change did it.
Which tools should a Head of Growth run experiments with?
Four picks, each with the trade it makes. None of them is the right answer for every company, and a shortlist where everything is excellent is an advert.
Runs tests on the page as it already is, so you can ship an experiment without a release and without booking a slot in the web team's sprint.
Watch outClient-side tests can flicker on slow connections, showing the original page for a moment before the variant loads. Check that on a real phone before every launch, not once.
Built for marketing teams rather than engineers, so the person who writes the hypothesis is usually the person who can build the variant.
Watch outThat ease comes from a visual editor, and visual editors are tied to the markup underneath. A redesign invalidates tests built on the old page, and nothing warns you.
Server-side testing is a first-class option, which is what you need when the thing worth testing is pricing or the signup flow rather than a headline.
Watch outServer-side puts a developer back in the loop. You trade editing the page yourself for tests that do not flicker and cannot be blocked by an ad blocker.
The one that gets through procurement, and the feature-flag side means product and growth can run on a single system instead of arguing across two.
Watch outIt is priced for a company running tests continuously. If testing is a side project rather than somebody's job, you are buying a platform to do a simpler tool's work.
What can a Head of Growth hand to an assistant?
Each of these is one file you paste once, and then the job runs the same way every time.
Test hypothesisWrite an A/B test up properly before it gets built: who you are helping, what they are trying to do, what you are changing, why you think it will work, and what you will learn either way.Test iterationA test finished. Get ten specific things to try next, each with what changes, why it might work, and what success would look like, grounded in what the last one showed.Learnings logWrite a finished test up in four fixed sections, then decide what earns a place in the file everyone searches before proposing the next one.How Might We generatorTurn a customer problem into three to five How Might We statements, then into specific website ideas for each one, with where it goes and how you would test it.Which tools should a Head of Growth connect to their assistant?
Connecting these means asking a question about the real site instead of copying screenshots into a chat window.
PlausibleCommunity MCP server by the Sentry team, Stats API key, self-hosted or remote
AmplitudeRemote MCP server, mcp.amplitude.com/mcp, OAuth, EU endpoint available
VWORemote MCP server, mcp.wingify.ai/mcp, OAuth with dynamic client registration
OptimizelyRemote MCP server, exp.mcp.opal.optimizely.com/mcp, OAuthQuestions people in this role ask
Taken from what people actually ask about this, not written to fill the page.
How much traffic do I need before A/B testing is worth it?
Enough to reach roughly 100 conversions per variation inside two to four weeks. Below that a test either runs so long the season changes underneath it, or ends before the numbers mean anything. On a low-traffic page, make one large change and measure it as a before and after rather than splitting the traffic.
How long should a test run?
At least one full business cycle, and two to four weeks for most website tests. Weekday and weekend visitors arrive from different places and convert differently, so a Monday-to-Friday test has excluded part of your audience without telling you. Set the end date before you start, and do not move it because Wednesday looked good.
Why can I not stop the test as soon as it hits significance?
Because significance appears and disappears while a test runs. Most A/A tests, which serve the same page to both groups, will cross the line at some point if you keep checking. Stopping the moment you see a winner is how a difference that does not exist gets shipped. Fix the sample size first, then look.
Do I need a developer to run tests?
Not for client-side tests. VWO and AB Tasty give you a visual editor, so a marketer builds the variant and launches it without a release. You need a developer the moment the test touches what sits behind the page: pricing logic, a signup flow, anything server-side. That is the real line, not the tool's marketing.
What is flicker, and how much does it matter?
Flicker is a visitor seeing the original page for a moment before the variant loads. It matters because those people saw both versions, which pollutes the result. Modern setups load the script in the head so the browser holds the paint until the variant applies. Check it on a real phone on a slow connection.
Who owns the website if I own the number?
Usually whoever controls the budget owns the strategy, but the practical test is different. If every change needs somebody else to deploy it, they own the site regardless of the org chart. Worth settling explicitly rather than discovering it mid-quarter, and a client-side testing tool is the usual way to buy back some independence.
Not your job?
Every tool named here has a full entry in the stack, with the verdict and what it costs you.