Marketing teams have access to more data than ever. Most have a dashboard showing click-through rates, cost per acquisition, ROAS by channel, and revenue by campaign. That is a lot of measurement. But it often leaves the most important questions unanswered.
Here are five questions that rigorous marketing measurement should be able to answer, and what it takes to actually get there.
1. What did marketing actually contribute?
This is the foundational question, and it is harder than it sounds. When your revenue goes up in October, how much of that was your autumn campaign, how much was seasonal demand that would have happened regardless, and how much was driven by that competitor closing down?
Most measurement tools do not separate these factors. They record that revenue went up and attribute it to whatever campaigns were running. Marketing Mix Modelling is designed specifically to untangle these effects, estimating how much each factor contributed independently.
The right benchmark for marketing is not total revenue. It is incremental revenue: the revenue that would not have happened without the marketing.
2. Which channels created incremental value?
Your paid search campaign might show an excellent ROAS. But if the people clicking those ads were already searching for your brand name and would have found you organically, the campaign is not creating value. It is capturing demand that already existed.
Incremental value means new customers, new revenue, or new behaviours that would not have existed without the marketing investment. Identifying which channels actually deliver this, rather than which channels appear in the conversion path, requires either a holdout experiment or an MMM that separates media effects from baseline demand.
3. How confident should you be in the result?
Most marketing reports present numbers as facts. Paid social drove £420,000 in revenue last quarter. But that is not a fact. It is an estimate, and like all estimates it has uncertainty around it.
Good measurement quantifies that uncertainty. Instead of saying a channel drove £420,000, it says the channel drove somewhere between £280,000 and £560,000, with the middle of that range being the most likely figure. The width of that range tells you how reliable the estimate is.
When you are deciding whether to double a channel's budget, the confidence interval matters as much as the point estimate. A result you are 90% confident in justifies a different decision than a result you are 55% confident in.
4. Where should the next budget go?
Budget allocation is rarely done with precise data. Most marketing teams allocate budget based on last year's split, adjusted for whatever performed well in recent months, with some negotiation between channel owners.
A rigorous measurement approach gives you something better: the marginal return of each channel. This tells you not just which channel has performed well historically, but where the next additional unit of budget would generate the best return right now, accounting for the fact that channels become less efficient the more you spend on them.
- Which channels are close to saturation and will generate diminishing returns on additional spend
- Which channels still have headroom and would respond well to increased investment
- What the expected revenue impact would be of moving budget from one channel to another
5. What should you test next?
Measurement is not just about understanding the past. It should inform what you do next to improve your evidence. If your MMM shows high uncertainty around paid social contribution, that is a signal to run an incrementality test on paid social. If your holdout experiment shows strong TV lift, that is a signal to test increasing TV weight.
Good measurement generates a testing roadmap. The questions you cannot currently answer reliably become the experiments you design to answer them. Over time, the loop of model, challenge, experiment, improve makes each measurement cycle more accurate than the last.
Measurement is not a report you produce once a year. It is a system that gets better over time as you feed experimental evidence back into your models.
Why most measurement does not answer these questions
Platform analytics tools are built to measure activity within the platform, not to give you an independent view of your full marketing contribution. They are not designed to answer these questions, and they have a commercial interest in showing strong results.
Getting answers to these five questions requires an independent approach: data that is not controlled by any single platform, methods that account for factors outside your marketing, and a willingness to present uncertainty honestly rather than hiding it behind a single confident-looking number.
What if my data is not good enough to answer these questions?
Start with what you have and identify the gaps. Data quality issues are usually solvable with better tracking, consistent naming conventions, and more complete historical records. The measurement process itself often reveals where the data needs to improve.
How long does it take to build rigorous measurement?
An initial MMM project can be completed in four to eight weeks. Building a full measurement system with ongoing experimentation is a longer process, typically six to twelve months to establish a reliable rhythm of model, test, and improve.
Do I need a large budget to do this properly?
The cost of good measurement is almost always small relative to the media budgets it helps optimise. Even a modest improvement in budget allocation, reallocating 10% of spend from low-incrementality channels to high-incrementality ones, typically generates a return many times the cost of the measurement.