Blog / Only about 20% of your marketing can be attributed, so stop selling the rest.

Nice data, but  where is my money?

You spent half a million on the tech stack, plus whatever went into the media spend on top of that, and six months later the CFO wants to know what came of it. Where is my money? It’s a fair question, and it’s one of the least enjoyable conversations a marketing leader ever has to sit through when it can’t be answered easily.

I’ve been doing this for over 20 years, and one thing I’ve changed my mind on is the need to attribute every single dollar of revenue to every individual marketing activity. That always used to be the holy grail of what we used to promise clients, to know where EVERY dollar is going.

I sold that promise for years, and we worked towards it using all the latest attribution tools.

As I’ve gone on, I’ve realized it’s simply unrealistic, especially in 2026 where there are so many different ways for potential customers to engage. Dark social, which is the unattributable stuff, things like sending Slack messages, WhatsApp messages, forwarding links and emails, is simply untrackable. Some activities just need doing, but you can’t put every cent against them.

That isn’t a new problem either, which is the part I find funny. A journalist at The Atlantic called Alexis Madrigal named dark social back in 2012. He’d been going through his own traffic data and noticed a big share of visits to articles with long, ugly URLs were being logged as direct, meaning someone had supposedly typed the whole thing out by hand. Obviously nobody had. Those people had been sent the link privately, over email or messenger, and the referrer was stripped on the way. That was 14 years ago, and we still can’t see it.

What attribution software is good at is attributing what’s attributable, and this is what needs communicating through the organisation and especially up to the board and investors.

I don’t think there’s an exact number here that can be attributed, and I’d be wary of anyone who hands you a firmer one than that. My experience says somewhere around 20% of what’s going on is fully attributable. A lot of the activity you’re simply never going to know about, but you know that it’s going to be helping.

A good example of that would be a podcast. We make one (so take that with a pinch of salt). You just can’t know everybody that’s listening to it, how they’re processing the information, or where they’re sending it or how they’re sharing it. And it can be incredibly important for the business, for people to learn about it, and for the derivative content you get off the back of it. Nobody has ever proved the value of one, and nobody sensible has stopped making one either. But at the same time, nobody can tell you that they heard me on a podcast, then Googled my company name, and clicked our paid media ad. There is nothing attributable on a platform.

Measuring B2B Marketing ROI: What I Measure Instead

There is still a return that has to go on the budget, though. Mine is a coarse one. Total spend against what the business actually generated, measured at the end of each quarter and again across the year, with the media budget and the cost of marketing itself, agencies and staff included, both counted on the spend side. The individual pieces, the specific blog posts, the link clicks, I’ve come to terms with not knowing. It’s a blunter standard than the one I used to sell, and it’s one I can actually stand behind.

Dark Social and Marketing Mix Modeling: The Argument Against Me

There are people (the LinkedIn mafia) who’ll tell you I’ve given up rather than changed my mind. Incrementality testing holds spend back from part of your market and measures the difference. Marketing mix modelling works at the level of the whole budget instead of the individual click. Both do things attribution never could, and there’s no argument here from me with the “theory” surrounding all of this.

Then you go back to your desk and reality hits. Clients typically say something like, “We’ve still got a massive target, our budget’s been cut in half and I don’t have the people.”

Your data is a mess, you don’t trust it, and you’re embarrassed when you put it in front of the CEO. Those data modeling methods want scale and patience and years of clean history to work from, and they’re out of reach for almost every team we’ve worked with.

I worked with a cloud infrastructure company that spent the best part of a year cutting the volume of signups coming through the door. When we looked under the hood, most were never going to become customers.

With changes, website leads fell by around 85%, and on every dashboard in that building it looked like a disaster. Their marketing lead went into his board meeting and said the website leads are down 85% and the customers are up 800%.

Nobody could argue with him.

He could only say that because he’d agreed the standard with them a year earlier, before any of the numbers got ugly. That’s the part too many people skip, and then they’re arguing about the method at exactly the moment they should be talking about the business. So agree while things are calm. Say out loud what you’ll be judged on, and what you won’t be able to see underneath it. I’ve never seen anyone lose the board’s confidence for saying honestly that a channel is hard to measure, and I’ve watched plenty of people lose it (me included) by claiming a precision that everybody sitting there could tell had been invented on the way in.

Put what you spent and what the business generated together for the last 4 quarters and look at what they say. If the return is there you’ve got your answer. And if it isn’t, you’ve discovered something new.

I’d rather learn this for myself than have a CFO discover it for me.

The harder conversation is the one after this, when the number you’ve been reporting for years has to change and somebody in the room decides you’re moving the goalposts. I’ve written up how to do that without losing the room.