Ten reasons marketing effectiveness projects fail (it’s rarely the numbers)

Econometricians find it fascinating that a business will spend £25m+ a year on marketing but won't put a small fraction of that toward understanding how to spend it better. Instead of saying "let's do more TV", why not prove where the money works hardest first?

It may well be TV, but we need to know.

Plenty of businesses that do invest come away disappointed. The report is filed but the media plan goes ahead unchanged, and the verdict is that econometrics doesn't work. In my experience, the analysis is rarely the problem.

I've evaluated a lot of other econometricians' work over the years, and the numbers are seldom wrong. What goes wrong is everything around them: whether a brand starts at all, what question it asks, who it brings in, and who it trusts to explain the answer.

Here are the ten mistakes I see most often when it comes to measuring marketing effectiveness.

1. "It's too complex so we won't start"

Some brands have a genuinely complicated sales picture. Think of an insurer selling many different types of policies, or a car brand. Faced with all of it, they decide it's too difficult to measure and fall back on gut feel.

But complexity is a reason to measure, not a reason to wait. The more moving parts you have, the more you need to see where the problems are and where the value is really coming from.

2. Data analysis paralysis

Other brands have the opposite problem. They gain twenty answers from twenty data sources and don't know which to believe. I've seen senior people with four different answers from four different teams, some with a vested interest in their own data set. The result is paralysis.

Platforms add to it. You used to rely on your agency; now Facebook, Google, and others offer their own analysis, too. Those relationships are useful, but treat them with care. Google will tell you how to do best on Google. Effective measurement needs a framework that pulls every source into one coherent story.

3. Sales vs marketing, and speaking the Board's language 

Sales teams often position themselves as the numbers guys. They've come up from the shop floor and already believe they know what drives sales. It has improved over the last decade, but some marketing teams still struggle to push back because sales often holds more sway across the business.

In some companies, marketing has less pull than it should with the people who set budgets and plans. That's usually about language rather than value. Marketing talks about campaigns, content, sponsorships, and channels. Boards want to hear about growth, profit, market share and future performance. Effectiveness work is how you translate one into the other.

4. Underinvesting in understanding

We regularly see brands spend upwards of £25m on marketing but hold back on analysis because they don't see it as working capital. Often that's because they assume serious econometrics costs far more than it does.

Cost follows complexity, not the size of your spend. A £1.1bn business with a stable model might need one project a year, while a small direct-to-consumer brand testing multiple channels and audiences may need more frequent analysis to maintain speed of analysis. And smaller brands can get just as big wins. Knowing what's working, what isn't, and where to put next year's budget is a strong start for any team that wants to make confident marketing decisions.

5. Selling it to the wrong person, alone

Sometimes an econometrics project idea doesn't reach the CMO in the strongest way. It needs to be clear how it will help the wider business, otherwise it's hard for the CMO to sell it to the CFO and the rest of the business.

Be clear on what analysis of your marketing data can and can't do. Set the parameters and outcomes in advance and show the value to the business. As soon as one key person misunderstands the project's purpose, it loses traction.

6. Not understanding the "exam question"

This is the biggest one. Everyone says the analysis looks great and is very interesting, and then it sits on a shelf and is never looked at again.

Take a client who asked for an ROI analysis when the real issue was over-reliance on one channel in a constrained market, and they needed proof that something else would drive long-term growth. Nobody is being difficult - they just haven't boiled their needs down into the questions the analysis has to answer. A good effectiveness partner makes sure those questions are asked; digs into what the brand is trying to achieve and shows how the data can point the way.

7. Surprises, and staying too small

Before any work starts, there should be no surprises. Everyone relevant should know the project exists, what it's answering and what the outcome will look like, and they should have put in their tuppence worth.

Skip that step and important parts of the business will say "it's not a project for me". It becomes an interesting project for marketing and nothing more. If your project doesn't cross silos and departments, it won't deliver the outcomes you want. Go high enough in the business, and you'll go wide enough.

8. Siloed analysis

It's amazing how much separate analysis happens inside one business at any one time. Part of going wide is bringing in different divisions' views, so effort isn't duplicated. Teams often tell us they're already doing work that overlaps with our brief, and joining it up gives a clearer, deeper picture of what drives sales and delivers the best return.

On one project, loyalty analysis and pay-cert work were already happening elsewhere in the client, so we could add another piece of the jigsaw. Fact-finding means following the data: "You want my data? You should probably talk to that guy over there." It takes time, but it can really affect the outcome.

9. Using AI the wrong way

Understand AI's pros and cons, but don't delegate your thinking to it. Use it to turn PhD-level methodology papers that Google and others publish into plain English, explain why two numbers disagree, or poke holes in your thinking. Don't expect it to replace judgement, or even to analyse your data correctly.

Always check AI output with someone who understands how marketing mix modelling (MMM) works. Otherwise, you risk sending your brand down a path built on incorrect modelling.

10. The wrong partner

Given everything above, choose a partner you trust, and one who can stand up in front of your CEO. If you wouldn't be happy putting them in front of the boss, you've probably chosen the wrong one.

You need to be able to present your analysis and plans to the C-suite. A partner who can clearly explain how the models work, and why they recommend a course of action, makes securing budget much easier. They should help you understand what your data means, and support you in working out how to bring the awkward person on side.

A quick check before you commission any marketing effectiveness analysis:

• Do we know the decision this work needs to inform?
• Have we written down the questions it must answer?
• Does everyone who will be affected know about it, including finance and sales?
• Have we found the other analysis already happening in the business?
• Would we put our partner in front of our CEO?

It's rarely the marketing effectiveness analysis

It's what happens around it: the question nobody asked properly, or the department that wasn't included. Ensure the exam question is right, bring in insight and data from across the business so everyone owns the answer, and your project won't end up on a shelf. Then pick a partner you'd happily put in front of your CEO. Do that, and the analysis stops looking like a cost and starts looking like the best money in your marketing budget. 

If you want to talk through the questions your marketing data should be answering, contact the MetaMetrics team.

Article by Sam Watts | Operations Director, MetaMetrics