How to get a marketing test approved when nobody wants to be wrong

You get a marketing test approved by removing the things that make a sceptic nervous. Cap the budget. Name the date the decision gets made. Name the person who makes it. Agree the failure conditions before launch, in writing. Say what happens to the learning if it dies. Do that and the ask changes shape into a fixed sum and a date in the diary, which is a far easier thing to say yes to.

Why has approval got harder as testing has got cheaper?

It has not got harder. The cost of being wrong fell, and the process that governs being wrong stayed where it was.

Most B2B approval chains were built around a version of marketing that has quietly stopped existing. A campaign meant a brief, a creative round, licensed imagery or a shoot, a build, a translation pass, and six weeks of production before a single human saw it. On those economics, caution paid for itself. Anything that expensive deserved four people looking at it, because the cost of a bad one was measured in a quarter of the team's scoped work.

That is no longer what a test costs. In our own delivery, a landing page variant, three ad concepts and the tracking behind them is now a matter of days rather than weeks, and the media behind a first read is often a three figure number.

The sign off chain did not move. So the ratio inverted. In a lot of businesses, getting a small test past a marketing lead, a brand owner, a compliance reviewer and a finance approver now consumes more salaried hours than building and running the thing they are reviewing. You are paying four people to protect you from a risk smaller than the meeting.

What are the four sign offs actually protecting?

Nothing dishonourable. That is the difficulty.

Your finance director is protecting against unbounded commitment. They have seen a £5,000 pilot become a £60,000 annual line with no decision point in between, and they are right to remember it. Brand is protecting consistency, because they are the ones who get asked why the German site says something different. Compliance is protecting the claim. The marketing lead is protecting the plan they already committed to at the start of the year.

Each of those positions is rational on its own. Put them in sequence and they become a filter, and the filter has a bias. Nobody in that room has an incentive tied to a test that worked. Several of them have standing tied to something visible going wrong. When the reward for being right is a slide and the penalty for being wrong is a conversation, the sensible individual choice is to slow everything down.

Which is how you end up with a marketing function full of competent people who all privately agree the plan is too safe.

Why does the chain select for the ideas least likely to work?

Because a business case is only demanded of ideas that are genuinely different.

Watch what actually happens. Repeat last year's trade show and it passes as continuation, on a line that was approved once and has never been reopened. Propose the same money on something nobody in the business has tried, and suddenly there is a template, a stakeholder review and a request for supporting evidence. The bar tracks novelty rather than money.

The consequence is the expensive bit. The ideas that survive your approval process are the ones that most resemble what you did last year. What you did last year is also what your three closest competitors did last year. It is in the auction, it is in the benchmark, and it is in everybody's plan. You are spending a full budget to arrive somewhere the market has already arrived.

The odd part is that this filter is sold internally as risk management, while it reliably funds the highest cost, lowest differentiation option on the table.

What does a test proposal that gets a yes contain?

Five things, on one page. Not a deck.

A ceiling, stated as a maximum loss. Open with the worst case in cash. "This costs £2,000, and it cannot become £6,000 without coming back to you." An approver's first silent question is how big this could get if it goes badly and nobody tells them. Answer it in the first line and most of the resistance leaves the room with it.

A named decision date. A day in the diary, three or four weeks out, when this either continues or stops. Not "we will review in due course". A date.

A named decision owner. One person, by name, who makes the call on that day. Committees do not kill things. People do.

Kill criteria agreed before launch. Which signal you will judge on, what the threshold is, and how many days after launch you will read it. Write it down before any data exists, because a threshold moved once you have seen the numbers is a negotiation with yourself. Add the guardrail alongside it, the thing that must not degrade for the test to count as a win. A variant that doubles form fills and halves lead quality is a loss wearing a medal.

What happens to the learning. Where the result gets written down, in what format, and who can find it next March. This is the part everyone skips and the only part that compounds.

How do you frame the ask so the sceptic stays in control?

Say their objection out loud before they do, then hand them the decision.

Open with the accusation they are already forming. "You are going to think this is a distraction from the Q4 plan, and for most of the budget you would be right." Naming it costs you nothing and takes the charge out of it. An objection you have voiced yourself is much harder to deploy against you.

Then ask rather than argue. "What would have to be true for this to be a bad use of two thousand pounds?" You will get the real objection, which is almost never the one in the meeting invite, and you will get it early enough to design around.

Then let them set the number. Ask what they would be comfortable losing entirely on this. Whatever they say becomes the cap, and because it is their figure it does not get relitigated in November. People defend their own numbers.

And give them a clean way to say no. "If this is the wrong quarter, tell me and I will bring it back in January." Someone who knows they can refuse without a fight tends to stop looking for reasons to refuse. What you are chasing is comfort with a bounded amount of exposure. Agreement with the idea itself is a separate purchase, and a much harder one to close.

Why is a clean kill a return on the money?

Because you bought an answer, and an answer removes a permanent item from the argument.

Consider the alternative. An untested idea stays a topic. It comes back every planning cycle, it consumes an hour of a leadership meeting each time, and eventually someone senior enough wants it and it goes live at full scale with no read on whether it works. Two thousand pounds spent finding out that it does not is a better trade than four years of debate followed by a £40,000 commitment.

So log the kill like an asset, because it is one. Date, hypothesis, spend, which signal you judged on, what it read, what you decided, who decided. After roughly fifteen closed tests, that register stops being a filing exercise and starts telling you which of your own early signals actually precede revenue in your business. Nobody else has that, and no vendor can sell it to you.

HMS Networks is the version of this that finance understands. Cost per qualified lead through trade events sat at around £760. The programme brought it to around £72, at 35:1 on media, with 201 confirmed leads into CRM and 1,588 sales ready leads from the website. The £760 line was the one nobody had to justify, because it had been approved for years. The £72 route needed arguing for.

TouchWood Play makes the same point over a longer horizon. £135k of cost, £927k of tracked return, £1.1m of confirmed funnel revenue and a 31.8% uplift in proposals linked to the funnel. None of that was available from a business case written in advance. It came from being allowed to run the thing and read it honestly.

What do you do if the answer is still no?

Three moves, in order of preference.

Shrink it below the line. Most businesses have a delegated spend limit that requires no committee. Get the test under it. A smaller read that happens beats a bigger read that waits until March.

Run it inside something already approved. A variant within a live campaign is usually a change, not a proposal, and changes go through a different door.

Get the finding another way. Add the "how did you hear about us" question to your forms. Ring five recent customers. Turn up with evidence rather than a hypothesis, and the second conversation is a different conversation.

Where to start this quarter

Pull out the last four ideas that were rejected, and sort them into two piles: rejected on merit, and rejected because nobody could see where the spending stopped. If the second pile is bigger, your problem is the shape of the ask.

Agree a standing test budget with your finance director once, with a cap and a monthly reporting line, so that approval becomes an arrangement rather than a negotiation you have to win every time.

Then write down the kill criteria for the next thing you launch, before you launch it, and put a date and a name against the decision. That single habit will change more about your marketing than any channel choice you make this year.

If the governance around this is the blocker, our marketing operations and governance work is largely about making that approval route survivable, and our B2B marketing strategy engagements start by finding out which of your ideas the process is quietly eating.

Frequently asked questions

What should a marketing test proposal include? Five things on one page: a capped budget stated as a maximum loss, a named decision date, a named decision owner, kill criteria agreed before launch, and where the learning gets logged if it fails. The proposal asks for a fixed sum and a date in the diary rather than for belief in the idea.

How do you write a business case for a marketing test? Lead with the worst case in cash rather than the upside. State the ceiling, the decision date, the person deciding, and the signal you will judge on before any data arrives. Then say what the business gains if it fails, which is a documented answer that removes the idea from every future planning argument.

How much budget should a marketing test have? Enough to produce a readable signal on the layer you plan to judge, and small enough that your approver can lose all of it without needing a conversation with their own boss. In our experience the second condition matters more. A test that sits under a delegated spend limit gets run this month, and a larger one waits for a committee.

Who should approve marketing experiments? One named person, with a standing cap agreed in advance. Sequential sign off from four functions filters for familiarity rather than for risk, because only genuinely different ideas ever get asked for a business case. Agree the limit once with finance, report against it monthly, and stop negotiating each test individually.

How do you prove a failed test was worth the money? By logging it properly. Record the date, the hypothesis, the spend, the signal you judged on, what it read, what you decided and who decided it. A killed test with that record removes a recurring debate from your planning cycle. After roughly fifteen closed tests the register also starts showing which of your early signals genuinely precede revenue.