What twelve tests a quarter does to a marketing team

Moving from three big bets a year to twelve tests a quarter changes your meetings before it changes your numbers. Reviews get shorter because there is a decision to make rather than a performance to give. Being wrong stops being career threatening because it happens weekly, to everyone. Junior people start running things. And within about a year the team's log of what worked becomes the most valuable thing it owns.

What changes in the first month?

The meeting.

A marketing team running three big bets a year holds review meetings that are essentially presentations. Someone has been working on the thing for eleven weeks, the deck exists to show that the work was done well, and the honest question underneath, which is whether to keep going, gets raised in the last four minutes by whoever is bravest.

A team running twelve tests a quarter holds a different meeting. There is a list. Each item has a decision date that has either arrived or not. Where it has arrived, someone says kill or scale and the reason, and the register is updated in the room. Where it has not, the item is skipped. Meetings that used to run ninety minutes tend to settle at thirty inside a month.

The change people notice first is that nobody presents. The change that matters is that the awkward question stops being awkward, because it is now the entire agenda.

Why does being wrong stop being dangerous?

Because it stops being rare.

When a team runs three significant pieces of activity a year, each one carries a share of someone's reputation. There are only three data points, everybody knows whose idea each was, and the incentive to keep a struggling piece of work alive until it improves is enormous. Nobody is being dishonest. They are behaving sensibly given how they are being judged.

Run forty eight tests a year and the arithmetic changes. No single result carries a career. Most tests fail, everyone knows most tests fail, and the person whose idea it was is already three tests further on. What gets judged is the quality of the hypothesis and whether the decision was made on the day, which are both things a marketer can control.

The tell is what people volunteer. In a team running three bets a year, nobody brings you bad news in week two. In a team running twelve tests a quarter, someone tells you on Tuesday that the variant they built is losing, because saying so is the job rather than a confession. That difference is worth more than any individual test result.

One rule makes it hold. A clean kill is logged as a completed test, with the same status as a winner. The moment killed tests are counted as failures in anybody's review, the whole thing reverts within a quarter and you will not be told.

What does it do to junior people?

It gives them something to own that a business can actually afford to hand over.

A test with a capped budget, a fixed decision date and a named decision owner is real authority with a bounded downside. That combination is rare in marketing. Ordinarily the only way to give a junior marketer decision rights is to give them something small enough not to matter, which teaches nothing, or something large enough to matter, which nobody sensible does.

Put a two year marketer's name in the decision owner column and several things happen at once. They read the numbers differently, because it is their call. They form an opinion before the meeting rather than during it. And they find out what it feels like to kill their own idea on a date they agreed to, which is the single most useful thing a marketer can learn early.

The honest part: some need coaching through the first few. The instinct to escalate rather than decide is strong, and the correct response is to hand the decision straight back with the question of what the register says. Two or three rounds of that and it sticks.

Why does the register become the team's most valuable asset?

Because it is the only part that compounds.

Everything else in a testing programme is consumed. Media spend is gone. Creative gets tired. Landing pages get rebuilt. The register, if it carries a hypothesis, a kill tier, a decision, a date, an owner and a 30 day read for every test including the ones killed at day four, is the one output that gets more useful the longer it exists.

After roughly fifteen closed tests it starts telling you something you could not have bought. It tells you which of your own early signals actually preceded qualified pipeline in your business, which is different from the general answer and different from your competitor's answer. At that point you stop borrowing somebody's testing model and start reading your own data.

Three secondary effects are worth having on their own. New starters read the register instead of asking four people what has already been tried. Agency changes stop resetting institutional memory. And the annual planning argument gets shorter, because the thing people used to argue about is now written down with dates on it.

A competitor can buy the same tools tomorrow. They cannot buy your log.

What does the compounding look like commercially?

TouchWood Play manufactures bespoke playgrounds and sells into a long B2B cycle with committee decisions and public sector procurement in the mix. Over two years the funnel produced £1.1m in confirmed revenue, £927k of tracked return against £135k of cost, and a 31.8% uplift in proposals linked to the funnel. That is 1200% ROI across the two years.

The shape matters more than the size. It came from the same funnel being worked continuously rather than rebuilt each year, across paid search, SEO, email automation, the website, the martech and call tracking. The 31.8% proposal uplift is the number that tells you what happened: a programme that got better at the same job repeatedly, rather than three separate attempts at a new one.

The other pattern worth naming is what small tests find that plans do not. On the Arrow ECS amplification during the Broadcom transition, the cheap reach layer that carried the programme turned out to be YouTube Shorts, with one piece, Containers Need Full Stack, reaching 169,660 views. The seven week programme delivered 454,176 views against a 75,000 floor, at £0.0019 per view against an £0.08 benchmark.

Nobody was going to argue for Shorts in a planning meeting about enterprise virtualisation. It was cheap enough to try, which is the only reason anybody found out.

What does it actually cost you?

More than the people selling testing culture tend to admit.

The first quarter is heavier. Tracking, CRM fields, tagging conventions and the register itself all have to exist before test one, and that work has no output anybody outside the team can see. Expect the first quarter to feel like it produced less than a normal quarter, because in visible terms it did.

It is more meetings, not fewer. They are shorter, and they replace longer ones, but the rhythm is weekly rather than quarterly and it does not skip when someone is on holiday.

Some people genuinely dislike it. The marketers who are best at the big set piece, who do their strongest work over eight weeks and want it finished properly, find a system that stops work on a date to be a poor fit. Some of the best people I have worked with found the first quarter of this genuinely unpleasant. A few adapt and become the strongest testers in the team. A few decide they would rather run campaigns, and going somewhere that runs campaigns is the right answer for everybody.

It exposes weak measurement immediately. Three bets a year lets you get away with reporting that cannot separate channel from segment from week. Twelve tests a quarter does not, and the discovery usually arrives in week three.

How does it fail?

Three ways, and the third is the one to watch for.

Test theatre. The twelve tests are the twelve things that were going to happen anyway, relabelled as tests. New landing page, refreshed email, another webinar. The diagnostic is quick and unwelcome: if nothing in the round would make a sensible person in your business slightly uncomfortable, what you have is a plan with better paperwork. Reject it and generate twelve more.

Testing trivia. Button colours, subject lines and image swaps whilst the offer, the audience definition, the pricing page and the qualification criteria go untouched for two years. The rule that fixes it is a single question at the point a test enters the round. If the best possible result would not change what you do next quarter, it does not get a slot.

Twelve tests and nothing learnt. The team runs the tests, makes the calls on time, kills promptly, and never logs the 30 day qualification read. The register fills up with hypotheses and same week decisions and an empty column where the part that compounds should be. This is the most common failure and the least visible, because from outside the team looks fast and disciplined. A year later it can tell you what it did and nothing about what it knows.

How many tests is actually the right number?

Enough that no single one matters.

For a marketing team of three to six people, twelve a quarter works out at roughly one going live a week with slack for the weeks that fall apart. Below about six a quarter you lose the effect entirely, because each test starts carrying weight again and the old behaviour returns. Above roughly twenty, small teams stop logging properly, which removes the only part worth having.

The number that does not work, in any size of team, is three a year.

Where to start

Three things you can do this quarter without buying anything.

Run one round of twelve ideas against a single real problem, and apply the uncomfortable test to the round as a whole. If none of the twelve would make anyone in the business shift in their seat, throw the round away and do it again. That exercise alone tells you a great deal about the team you have.

Put the 30 day read column in your register before you run a single test, and make it the field that has to be filled for a test to count as closed. Adding it later almost never happens.

Give one test to your most junior marketer with the decision right attached, not the delivery. Then let them make the call on the day, including the wrong one.

The full method is written up in Kill Perfect, which I presented at Webflow Conf in Boston on 2 September. The governance around it, the register, decision rights and review rhythm, is what our marketing operations and governance work puts in place, and the wider programme design sits inside B2B marketing strategy.

Frequently asked questions

How many marketing tests should a team run? For a B2B marketing team of three to six people, around twelve a quarter works, which is roughly one going live each week with room for the weeks that collapse. Below six a quarter the cultural effect disappears, because each test starts carrying enough weight for people to defend it. Above twenty, small teams stop logging results properly.

How do you build a testing culture in a marketing team? Start with the accounting. Log a clean kill as a completed test with the same status as a winner, and judge people on the quality of the hypothesis and whether the decision was made on the agreed date. If killed tests are counted as failures in anybody's performance review, the behaviour reverts within a quarter and nobody will tell you.

What is the biggest risk of running lots of small marketing tests? Testing trivia. Teams end up optimising button colours and subject lines whilst the offer, the audience definition and the qualification criteria go untouched for years. The filter is one question at the point a test enters the round: if the best possible result would not change what you do next quarter, it does not get a slot.

Does a testing programme mean more work for the team? Yes, particularly at the start. Tracking, CRM fields, tagging conventions and the register all have to exist before the first test, and that work is invisible from outside the team. Expect the first quarter to feel less productive than a normal one. From the second quarter the setup cost falls away and the register starts paying back.

What should a marketing test register contain? A hypothesis, the kill tier, the guardrail that must not degrade, the budget cap, the decision date, the named decision owner, the decision taken, and a 30 day qualification read logged even on tests killed in the first week. The 30 day read is the field teams skip and the only one that compounds into something worth having.