Marketing Attribution Model Accuracy & Deal Volume in B2B

Someone presents the numbers in a marketing review.

Search 34 percent. Social 22. Events 9. Precise, confident, professionally produced, and within twenty minutes money has been moved on the strength of it.

The question that determines whether any of it means anything almost never gets asked.

How many closed deals is this built on?

Not visits. Not enquiries. Not names sitting in a system. Deals that turned into revenue and can be found in the accounts.

Ask it, and the room usually goes quiet. Not because anyone is hiding the answer, but because nobody has ever needed to know it.

Two choices were made before it existed, and neither one is written down anywhere in what you were handed.

First: how many closed deals went into it. A pattern found across forty sales and a pattern found across four thousand look identical once printed as a percentage, though one is evidence and the other is an accident nobody has checked.

Second: how credit gets divided when several things touched the same deal. Somebody has to decide that. It is usually inherited rather than chosen, and rarely revisited once it is running.

Get either one wrong and the number still arrives, on time, looking exactly as it should.

Think about an election forecast.

Ask twenty people and you learn nothing, because two unusual answers move the entire result, while twenty thousand produces a real pattern because the oddities cancel each other out.

Attribution works on that principle. It searches your completed deals for combinations that tend to come before a sale, and combinations that come before nothing at all.

Which makes it a poll.

The respondents are your customers, and the sample is however many deals the business closed. Nobody would move money on a poll of twenty people. Here, the sample size is the one thing the answer never shows you.

The platforms that run these models publish what they need.

The figure sits at a couple of hundred enquiries a month. Every month, without a gap.

That word conversion is doing a great deal of work. It means an enquiry. A form completed, a call booked, a document downloaded. It does not mean a sale.

Now apply a rate most considered purchases would recognise. Somewhere around 3%.

200 enquiries a month, at 3%, is 6 closed deals.

So the volume the method needs in order to find a pattern is enquiries. The thing the business actually runs on arrives six at a time.

Here is what makes this so difficult to catch.

Nothing about the answer changes when the evidence behind it is thin. Same shape, same point in the month, same wording, whether it was worked out from forty sales or four thousand. And the volume it rests on is not printed anywhere on it.

So the question never gets asked.

The team presenting the number is also the team it assesses. And the person with the authority to demand the sample size is usually the one least equipped to know it is missing.

So it goes through. The next plan gets built on top of it.

One test, no cost and no expertise required.

Find the channel ranking from the last marketing review, the list showing which channel produced the most and which produced the least. Put it beside the ranking in front of you today, and read the two in order.

If the order has changed, and nobody can point to what moved in the market, the spending or the selling to explain why, you have your answer.

Buying behaviour does not reorder itself in a few months. Reputation, relationships and habit all shift slowly.

A ranking that moves without a cause is describing its own sample, not your market.

Now the second choice, the one flagged earlier.

It stays untouched for years.

Credit the last thing that happened before a deal, and the teams closest to the sale look like the engine of the business. Credit the first, and the work that built the relationship triples. Same year, same customers, same money. Only the rule moved.

So two teams can each be called your strongest performer, honestly, from the same deals.

That rule is rarely set by anyone neutral, and once a number is winning somebody their budget, nobody in the building has a reason to question how it was built.

If your company closes fewer deals than the method needs, the answer is not a better system.

It is fewer moving parts.

Run fewer channels, so any real effect is large enough to see unaided.

Ask buyers, in their own words, at the point they commit. They will tell you things no record contains, and they are rarely shy about it.

And settle the credit rule before the results arrive, not after, because a rule chosen afterwards is not a rule. It is a preference with arithmetic attached.

None of that requires a purchase. It requires deciding what you are willing to know.