The data said no. The customer said yes.

I am a spreadsheet person.

Before I understood marketing, I understood numbers. I liked that a number felt harder to argue with than an opinion. You could dislike a headline, but you could not have a philosophical disagreement with whether conversion rate went up.

That instinct has been useful throughout my career.

It has also been one of the things I have had to unlearn.

Because people are not spreadsheets.

They are inconsistent, emotional, social, distracted, risk-averse and occasionally completely uninterested in behaving the way our reporting model says they should.

The data can be right and the conclusion can still be wrong

This is an important distinction.

When I say data can mislead, I do not mean the numbers are fake. I mean we can ask the numbers a question they are not capable of answering.

Analytics can tell you what happened inside the things you measured. It is much weaker at explaining the invisible parts of a decision.

The conversation with a colleague. The recommendation in a WhatsApp group. The fact somebody has seen your name five times and cannot remember where. The customer who read three articles but finally enquired because a testimonial sounded like them.

Those things influence behaviour. Our dashboards usually flatten them into a neat source, medium and conversion.

Version one: the channel that looked small

I have seen versions of this repeatedly. A channel looks unimpressive in a top-line report because it does not create a huge volume of conversions.

Then you speak to the sales team and discover the conversations from that channel are consistently better.

They are better informed. They understand the offer. They ask more useful questions. They are closer to being a good fit.

If you only optimise for lead count, the channel loses. If you optimise for commercial value, it may suddenly matter a lot more.

The data was not wrong. The metric was incomplete.

Version two: the creative everybody hates

Marketers have a dangerous habit of becoming the audience.

We like clever work. Polished work. Work that looks like marketing.

Customers can be much less sophisticated, which I mean as a compliment.

Sometimes the plain ad wins. Sometimes the awkward headline gets the better response. Sometimes the page we want to redesign is doing its job because it answers the customer's question faster than the prettier alternative.

The lesson is not that ugly is good. It is that taste is not evidence.

When the result surprises you, the useful response is not to explain it away. It is to ask what the customer might be responding to that you were undervaluing.

Version three: the buyer who ignored the funnel

This is my favourite because it happens all the time.

Somebody follows a path that makes absolutely no sense in the reporting.

They discover you in one place, disappear, return directly, read something unrelated to the service they eventually buy, listen to a podcast, search the brand name, click an ad and then send an enquiry from their phone.

Which channel gets the credit depends on the attribution model.

Which channel caused the decision may be impossible to know.

The customer does not care. They were not participating in our funnel. They were trying to become confident enough to make a decision.

The irrational bit is not a flaw in the customer

Marketers sometimes talk about irrational behaviour as if buyers are malfunctioning.

They are not.

Human beings use shortcuts because decisions are expensive. We look for proof. Familiarity. Authority. Social reassurance. Simplicity. A reason to act now. A way to reduce perceived risk.

We compare things that are not directly comparable. We prefer the option that feels easier to explain. We avoid loss more strongly than we chase an equivalent gain. We trust people who feel like us.

That is not noise around the decision. It is part of the decision.

Use the data. Then add the human back in.

My argument is not 'ignore analytics and trust your gut'. That would be an excellent way to waste money with confidence.

Use the data. Measure properly. Track the trend. Test hypotheses. Understand the economics.

Then talk to people.

• Ask customers what nearly stopped them buying.
• Ask sales which leads feel different and why.
• Read the actual language people use in reviews and enquiries.
• Look for repeated objections, not only repeated clicks.
• Treat surprising results as something to investigate, not something to tidy away.

A dashboard is evidence. It is not the customer.

The spreadsheet man has a caveat

I still like a number going up.

I still think marketing decisions get worse when there is no measurement underneath them.

But I no longer think the spreadsheet gets the final vote simply because it looks objective.

The best marketing sits in the gap between what the data says and what the human being is actually trying to do.

Sometimes the numbers tell you no.

Then the customer buys it anyway.

Your job is not to be annoyed with the customer.

Your job is to understand what the spreadsheet missed.