Do They Love You? The Gap Between Brand Preference and Purchase in MMM

Does it matter for sales?

It is a bit like asking:

Do you know the time?

Yes.

But that is not what you wanted to know.

Correct answer. Wrong question.

That mismatch, between the answer you receive and the answer you actually need, shows up more often than we admit.

Preference and purchase studies follow the same pattern. They are not wrong. They answer the question they ask. But the question they ask is not necessarily the one you need answered.

Liking is not necessarily buying. And the assumption that one explains the other is not always wrong. It is just not always right.

Take Ferrari.

Do I know it? Yes. Do I like it? Yes. Do I want it? Yes. Can I afford it?

That is the question nobody asked.

And the answer changes.

Not today. Maybe in two years. Not with this mortgage. Perhaps with the next job.

The preference holds. Until it doesn't.

The desire is real. The moment is not right. Not now.

The research will record the love. It will not record the moment. It will not record the not yet.

Now take the bar at the airport.

You pay five euros for a water that costs seventy five cents anywhere else. You have no love for the bar. You probably do not know its name. You are not there by preference. You are there because you are thirsty and it is the only option before the gate.

The cash register goes kaching. The data records a transaction.

Nothing about that transaction reflects loyalty, preference, or desire.

And then there is the platform where price does most of the deciding.

No geographical monopoly. No single moment of captivity. Just a number low enough that brand becomes secondary, under current conditions, with the budget you have today.

You are not there because you love it. You are there because you cannot afford not to be.

The moment a credible alternative appears, or the moment your circumstances change, the behaviour changes with it.

There was no relationship. There was only a condition.

Three very different situations. Three very different reasons to buy. In the data, they can look identical.

This is not an exception.

People want things they may never buy. They buy things they do not want. They return to places they have never chosen.

Not because the feeling is false or the loyalty is real. Because the conditions made it inevitable.

And the systems we build to measure behaviour can only work with what we give them. Complete or not.

Preference and purchase sometimes move together. Not always. Not automatically. And not always for the reason assumed.

Do people buy because they like something? Or do they say they like it because they already buy it?

Both happen. Neither is universal.

From the outside, they look the same. Which is where interpretation can mislead.

Because we measure outputs.

Metrics are the reflection of variables that are not fully observed in real time.

Context. Friction. Habit. Attention. Constraint.

When these converge, behaviour tends to follow one direction. The path of least resistance. Not the most considered. Not the most aligned with desire. The easiest, under current conditions.

This is where efficiency and effectiveness part ways.

Efficiency follows the path as it exists now. It rewards what converts quickly, what is accessible, what captures demand already in motion.

Effectiveness operates differently. It works on the conditions under which choices are made. It increases the likelihood of being chosen before the decision becomes a race for convenience.

Not guaranteed. Not always measurable in the short term. But real.

Metrics sit between the two. They record behaviour. They do not explain it in isolation.

Optimising purely for efficiency moves systems closer to what is easy now. Pursuing effectiveness attempts to shape what becomes easy later.

Confusing the two is not a measurement problem. It is a human one.

We record what happened and call it a reason.

And we encode that human into the systems we build to measure it.

That is the world in which we live.

Which is where dashboards and Marketing Mix Modelling enter the picture.

They are a snapshot, not a map.

They show current behaviour under current conditions. They tell you where people are now. Not where they are going. Not what they would do if the conditions changed. Not what they will do when they can.

MMM quantifies patterns between inputs and outcomes. Spend and sales. Reach and response. Impressions and conversion.

But these are outputs. Proxies of activity. Not the full set of conditions behind behaviour.

Reach can move you closer to exposure. Spend can move you closer to investment. Neither captures buyer intent, constraint, or timing.

Each input metric carries its own limitations.

Spend is influenced by factors beyond contact. Placement, format, seasonality.

Impressions reflect contact, not attention or retention.

The question is not which one is right. It is how well it describes what you are actually trying to measure.

Which means the quality of what MMM tells you depends on the quality of what you give it.

We can only include current buyers in our models. We can only calculate current demand.

The person who admires the brand, intends to buy, and is simply not yet in the position to act leaves no trace in the data today.

We cannot know when they will be ready.

But they are real. And they are reachable.

The risk is not in the model. It is in forgetting that the model has edges.

Planning strictly within what converts now risks losing the audiences who will convert later.

We also cannot feed it the airport bar buyer's motivation.

The person who transacted without choosing. Whose purchase reflects geography and thirst, not preference.

That revenue looks the same as loyalty in the numbers. It is not.

If a channel has historically captured demand, MMM will confirm it performs well at that.

If another channel has been underfunded or used inconsistently, it may appear weak. Not because it cannot drive growth. Because it has never been given the conditions to do so.

The model reflects the path that has been taken.

The paths not explored leave no trace in the data.

Which means budgets tend to reinforce what already converts, while future demand remains harder to justify within the system.

So the question inside MMM is not just: what drove results?

It is: what behaviour did the system make easy? What behaviour did it make unlikely? And what desire exists right now that the system cannot see at all?

Because models built on past behaviour will tend to reinforce that behaviour.

Budgets flow to what converts. Conversion increases where budgets flow. The system stabilises around what is easy to measure now.

And the buyer who is not yet ready falls outside the frame entirely.

Effectiveness becomes harder to see. Not because it is absent. Because it operates upstream, shaping choices before they show up cleanly in the data.

Including desire that has not yet found its moment.

The risk is not that MMM is wrong.

It is that we ask it the same question we asked before.

We ask: what worked? And use the answer to decide what will.

Not necessarily the same thing.

The same mismatch.

A correct answer to a different question.

We do this in measurement, we do this in strategy, we do this in life.

We ask what is easy to ask. We measure what is easy to measure. We choose what is easy to choose. Right now. Under current conditions. With the constraints we have today.

And then we tell ourselves a story about why.

Do they love you, and do they buy you.

Not the same question. Not the same answer. Not the same moment. Not necessarily. Not yet. Not always.

Who you want is not necessarily who you take home. Who you take home is not necessarily who you love.