We've got artificial intelligence, predictive models and more marketing data than any generation before us. So why are we still struggling to answer the most basic commercial question?

AI can write code, produce films, diagnose patterns in medical images and solve problems that would take humans decades. It can remix low-budget Amanda videos into dancehall classics more convincing than half the case studies on LinkedIn. Yet ask a marketing director how much additional profit the business generated from its last £100,000 of advertising, and watch what happens.

Google reports one number. Meta reports another. GA4 has its own interpretation (don't get me started on what GAT6 thinks). The agency has a dashboard explaining why everything is broadly going to plan. Finance has a completely different set of numbers. Everyone's confident. Nobody agrees. And somehow we're all in the same meeting, trying to decide where the next £100,000 should go.

It's quite extraordinary when you think about it, particularly when you consider how much everyone in that meeting is being paid. At least the meeting itself is generating some economic activity.

We've become very good at measuring things

The marketing industry has spent the last twenty years building increasingly sophisticated ways to track performance. We can measure impressions, clicks, engagement, conversions, customer journeys and dozens of variations of return on advertising spend. We can tell you which campaign received the last click before a customer purchased. We can even tell you which creative someone saw three weeks before converting. But can we tell you whether that customer would have purchased anyway? That's a rather different question, usually followed by someone suggesting we revisit the attribution methodology. More Teams meetings? Thought you'd never ask!

A campaign reporting £500,000 of revenue from £100,000 of spend looks impressive. There'll probably be a case study about it. Someone might even win an award - then post the photos on LinkedIn because what doesn't scream I'm happy in my job more than "hey everyone, I've won an award, look how employable I am". But if £350,000 of those sales would have happened without the advertising, the commercial picture changes considerably. And that's before we consider margin, customer lifetime value or the opportunity cost of spending the money somewhere else. The award, I'm told, is non-refundable.

That's the difference between marketing performance and marketing effectiveness. One tells us what happened around our marketing activity. The other attempts to establish what happened because of it. They're related, but they're not interchangeable, although you'd be forgiven for thinking otherwise after sitting through enough quarterly business reviews. Particularly the ones where every channel is apparently responsible for 40% of the revenue. With everyone pretending they don't see it.

Surely this is exactly what AI should be doing?

This is the part I find most interesting. Consider the work involved in a typical marketing effectiveness review. We gather data from advertising platforms, analytics tools, CRM systems and financial reports. We reconcile inconsistent numbers, investigate performance changes, identify possible explanations, evaluate the evidence and recommend what to do next. We then repeat much of the process the following week or month. Sometimes we even change the PowerPoint template so it feels like a fresh piece of work. New colour palette, same conclusions.

That sounds like an enormous amount of work that AI should be capable of supporting, if not largely automating. And increasingly, it is. We already have marketing mix modelling, incrementality testing, attribution models, predictive analytics and automated bidding. AI can help connect these approaches, investigate patterns and evaluate alternative explanations far faster than a person manually working through multiple reports. Which raises an awkward question about the billable hours involved in producing them.

So why hasn't the problem been solved? Partly because it's genuinely difficult. We can't observe what would have happened in an alternative reality where a customer never saw an advert. We have to estimate it using experiments, statistical models and other evidence. The quality of those estimates depends on the data, the assumptions and the methods used. AI won't eliminate uncertainty. But uncertainty has never been a good excuse for avoiding a decision. Although it has been an excellent excuse for commissioning another study, ideally one that takes us comfortably into the next financial year.

But I think there's another reason. We've become so obsessed with proving marketing works that we've forgotten the point is to make it work better. That's a fairly damning admission for an industry that's spent so much money telling everyone how measurable it is.

Nobody gets paid for having the most accurate dashboard

Well, actually, quite a few people do. And perhaps that's part of the problem. The advertising platforms have their measurement systems. Analytics vendors have theirs. Agencies have reporting frameworks. Consultants have methodologies (although mine are actually very good and worth every penny!). Everyone has a proprietary approach, preferably with an acronym that sounds expensive. If it can be turned into a three-letter framework and trademarked, even better.

We've built an entire industry around explaining marketing performance. I'm increasingly interested in how much of that explanation actually changes what a business does next. Some of that work is genuinely valuable. Some of it is essential. Some of it involves changing the date range on a dashboard and sending an invoice. Occasionally, there's a new chart. That's when you know you've got the premium package!

But the commercial purpose of measurement is to help a business make better decisions. Should we increase investment in paid search? Reduce spending on social? Invest more in brand? Change our pricing? Improve conversion before spending another pound on traffic? Those decisions matter far more than whether we can attribute another 3% of revenue to a particular channel. Although I appreciate that 3% can be the difference between an agency hitting its bonus and having an awkward conversation with HR.

Yet how often does a marketing performance meeting end with a clear decision, an expected commercial outcome and an agreed method for establishing whether we were right? Too often, we finish with a list of things to investigate and a promise to monitor performance. Which is a remarkably sophisticated way of saying we'll have another meeting next week. And possibly invite more people. Someone from Data Science, perhaps, just to make sure we can disagree at a higher level of statistical confidence.

AI might make parts of our profession nervous

There's an obvious implication here for marketing professionals, myself included. If AI can retrieve the data, identify the patterns, perform the analysis and produce a reasonable recommendation, what exactly is the value of the person doing that work today? I think that's a question worth confronting, preferably before someone in Finance asks it on our behalf. Particularly if they've just discovered Claude...

There will always be a need for commercial judgement, customer understanding, creative thinking, experimentation and accountability. But I'm less convinced there's a long-term market for people whose primary value is knowing how to extract numbers from platforms and turn them into presentations. The technology is getting too good, too quickly. And I suspect "advanced Excel skills" is a less compelling justification for a six-figure salary than it used to be.

And frankly, I don't think many business owners will mourn the loss of another 40-slide monthly performance deck, particularly one where the executive summary is on slide 37 and the recommendation is to "continue optimising". I say this as someone who has produced more than a few of those decks. Many of them were excellent. And I'm sure the people who received them remember them fondly. The real question is whether we'd pay for that work ourselves if it were coming out of our own bank accounts. I'm going to assume the silence means you're all thinking about it.

What if we've been solving the wrong problem?

Marketing effectiveness is usually approached as a measurement challenge. Better attribution. Better tracking. More sophisticated modelling. More experiments. All important. All capable of consuming an impressive amount of time and money. But the ultimate objective is better commercial decision-making. A fairly simple concept that somehow requires a surprising number of consultants to explain.

Imagine a system that continuously reviews marketing performance alongside business outcomes, market demand, competitor activity and customer behaviour. It identifies where performance is changing, investigates possible causes and evaluates the available evidence. It then recommends an action, estimates the potential commercial impact and explains how confident we should be in the recommendation. Ideally without needing a 90-minute ExCo meeting to agree that revenue going down is generally undesirable.

Sometimes the answer is to increase investment. Sometimes it's to reduce it. Sometimes it's to run an experiment because we simply don't know enough. And occasionally, the answer might be that marketing isn't the problem. I appreciate that's a controversial suggestion in a room full of marketers discussing marketing performance. Especially if the sales director is also in the room.

The important part is what happens next. The decision is recorded. The action is taken. The outcome is measured. The system learns from the difference between what was expected and what actually happened. That sounds much closer to what a marketing effectiveness function should be doing. It might even mean the next meeting starts with whether last month's recommendations actually worked. Fun times!

It also sounds like something AI could make dramatically more accessible to businesses that could never justify an expensive team of analysts and consultants. Which, as someone who runs a marketing consultancy, is a slightly awkward business proposition to be advocating. But I'd rather be part of building it than spend the next decade explaining why it can't be done. There's only so long you can charge people to tell them the future is coming.

The question marketing needs to answer

I don't think AI is about to solve every marketing effectiveness problem. Causality is difficult. Data quality matters. Models can be wrong. And someone still needs to be accountable for the decisions, preferably someone who can't blame the algorithm when the results are disappointing. "The AI recommended it" is unlikely to go down particularly well at the next board meeting.

But I do think we're approaching a point where much of the analytical work involved in marketing effectiveness can be automated. Which leaves a difficult question for the industry. If AI can analyse the evidence, recommend where the next pound should go and demonstrate whether the decision improved commercial outcomes, how much of what we currently sell as marketing measurement will businesses still need?

I suspect the answer will make some people very unhappy, especially those selling dashboards. If the main output of your marketing effectiveness function is a report explaining what happened last month, I think you should be worried about what AI will do to your business model. And if your response is to add an AI-generated commentary box to the dashboard, I suspect you've rather missed the point.

(This is probably the point I come full circle and crack a joke about the love letter to the toaster. I'll be honest, AI wrote that headline and it forgot to add that in. So use your imagination!)

We've built machines capable of extraordinary things. Surely we can do better than another presentation explaining why ROAS was down last Tuesday. And if the answer is another dashboard, I might genuinely lose the will to live. Although I'm sure someone could build a dashboard to track that too.


GV Marketing helps businesses understand where their marketing investment should go next, using commercial analysis, measurement and evidence-led decision-making.

Know where your marketing budget should go next.

If you're trying to understand what's driving performance, where growth opportunities sit or how to allocate your budget more effectively, let's talk.

Get in touch