Marketing has never had more ways to measure itself.
We have better data, better models and better techniques. Marketing mix modelling is faster and more accessible. Experiments can be run across channels that were once difficult to test. Platforms can process enormous amounts of data and return an answer while the campaign is still live.
There are now companies promising something even more attractive: instant incrementality.
Connect the data, look at the dashboard and discover how much revenue your marketing really caused.
That is a powerful promise. It is also a reminder that the hardest problem in marketing measurement may no longer be producing a number.
It may be getting anyone to trust it.
A better answer is useless if nobody will act on it
The measurement industry has spent years improving its methods. That work matters. A stronger model, a cleaner experiment and a more credible comparison should bring us closer to the truth.
But technical quality and organisational trust are not the same thing.
A model can be statistically impressive and still have no influence on a budget. The marketing team may not understand how it works. The CMO may not feel able to defend it. The CFO may see assumptions that cannot be reconciled with the numbers used elsewhere in the business.
When that happens, the analysis becomes an interesting presentation rather than a decision-making tool.
The question is not only:
Is this result accurate?
It is also:
Do the people who control the money believe it enough to make a different decision?
That is a much higher standard.
Trust does not mean accepting a result without challenge. In fact, measurement becomes more trustworthy when people can challenge it properly. They need to know what was measured, what was excluded, which assumptions mattered and how uncertain the answer is.
A precise number on a dashboard can look reassuring. A range, accompanied by a clear explanation of what could change it, is often more honest.
Instant incrementality still has to answer a difficult question
The appeal of incrementality is obvious. Marketers do not merely want to know what happened after somebody saw an advert. They want to know what happened because of the advertising.
That means estimating what would have happened without it.
No amount of speed makes that counterfactual disappear.
It might be estimated through a controlled experiment, a holdout group, a geographic test, a natural experiment or a model with carefully chosen assumptions. Technology can make each of these faster. It can automate data preparation, identify patterns and make techniques available to companies that could not previously afford them.
But the software still has to create a credible version of the world in which the marketing did not happen.
So when a company offers instant incrementality across every campaign, channel and market, the sensible response is not immediate disbelief. It is a set of questions.
What is the comparison?
Is this the result of a test or a modelled estimate?
Which assumptions have the greatest effect on the answer?
What uncertainty sits behind the number?
Can the result be checked against another source of evidence?
These questions do not make someone anti-technology. They are how a company works out what sort of claim it is being asked to trust.
The danger is not automation itself. The danger is that a simple interface can make an uncertain estimate feel like an observed fact.
Marketers, CMOs and CFOs are not asking the same question
Measurement is often discussed as though a business is one person looking at one number. It is not.
The marketing team wants an answer quickly enough to improve the campaign. It needs detail, direction and something practical to do next.
The CMO needs a view that works across channels and markets. The answer has to support a strategy, survive a board meeting and help secure the next budget.
The CFO wants consistency with the rest of the business. What counts as revenue? Have margins and fulfilment costs been included? Is the return genuinely incremental, or has marketing claimed demand that would have arrived anyway?
All three may be looking at the same analysis and judging it by different standards.
A directional result may be good enough to move spend between two adverts. It may not be good enough to add several million pounds to next year's marketing budget.
Trust grows when the level of evidence matches the size and reversibility of the decision. The bigger and less reversible the decision, the stronger the evidence should be.
Sometimes the problem is not the model
There is another, less comfortable part of this conversation.
Sometimes people say they want better measurement, but they are only ready to accept certain results.
A channel looks efficient until an incrementality test says much of its reported revenue would have happened anyway. A campaign that won awards produces no detectable commercial effect. A long-standing agency relationship appears less valuable than everyone believed. A senior leader's favourite investment performs poorly.
At that point, a technical discussion can become political very quickly.
The methodology that seemed perfectly reasonable before the result is suddenly examined from every angle. People ask for another cut of the data, a different time period or one more model. Standards of proof rise when the answer threatens someone's budget and fall when it supports what they already wanted to do.
This does not mean every uncomfortable result is correct. Measurement should be challenged. Models fail, tests are badly designed and data can be incomplete.
But there is a difference between scrutinising evidence and shopping for a more convenient answer.
Organisations often treat this as a measurement problem when it is really a governance problem. Who agreed the question? Who chose the method? What would count as a meaningful result? What decision would follow each possible outcome?
If those things are decided only after the number appears, politics has already entered the analysis.
One of the best ways to protect trust is to agree the rules before seeing the result. Write down the hypothesis, the success measure, the acceptable level of uncertainty and the decisions that different outcomes will trigger. It is harder to dismiss an inconvenient result when everyone approved the test in advance.
Trust is built before the dashboard appears
Trustworthy measurement is not a particular model or product. It is a way of working.
It starts with a clear question. It uses a method suited to that question. It makes assumptions visible and uncertainty impossible to miss. It distinguishes an observed result from a modelled estimate. It remains consistent when the answer is inconvenient.
It also uses more than one type of evidence.
Experiments can help establish causality, but they may cover only part of the business. Marketing mix models can provide a broader view, but they depend on data quality and modelling choices. Attribution can help teams understand journeys and optimise activity, but it should not automatically be treated as proof of incrementality.
The aim should not be to force every method to produce the same number. It should be to understand why the answers differ and what each one is qualified to say.
This requires a little humility from measurement providers too. A supplier that can explain when its product should not be used may earn more confidence than one that claims to answer everything.
There is strength in saying:
This is what the evidence suggests. This is how confident we are. This is what would change our view.
That sounds less exciting than instant certainty. It is much more useful in a serious budget conversation.
The real product is confidence in the decision
Marketing measurement will keep improving. Models will become faster. Experiments will become easier to run. Artificial intelligence will automate more of the analysis. Companies that once had very little evidence will gain access to techniques previously reserved for the largest advertisers.
All of that is good.
But a technically better answer does not automatically become an organisationally accepted answer. People still need to understand it, challenge it and be willing to live with what it says.
That last part matters. Trust is easy when measurement confirms the existing plan. Its real test comes when the evidence says that a favoured channel is weaker than expected, a celebrated campaign did not work or a budget should move away from the person defending it.
The best measurement systems do not remove disagreement. They make the disagreement more honest. They separate questions about the evidence from questions about status, ownership and preference.
In a market full of companies selling faster answers and instant incrementality, the winner may not be the one with the most impressive dashboard.
It may be the one whose answer a marketer can use, a CMO can defend and a CFO can trust, even when none of them particularly likes it.