Fuzzy attribution: a more useful way to measure B2B marketing

Marketing attribution starts with a sensible question: what contributes to revenue? If you can work out what marketing leads to worthwhile business, you can make informed decisions about where to spend your budget, and ultimately get a better result.

That’s the part we all agree on, but often the focus ends up being more on producing an answer, rather than necessarily the right answer. And when those answers are rolled up into reports, you start to get lots of apparently useful data which is built on less than solid foundations, giving you a false sense of precision where something ‘fuzzier’ would be more useful.

Answers that might be wrong

Many systems fail to capture attribution effectively because;

• The CRM only has one source field, or at best two (eg. first and last touch)
• The record focuses on digital engagements because they are more measurable, rather than because they are the most useful
• Even multi-touch models can assign percentages that look more certain than the underlying evidence warrants.

Don’t get me wrong, all of this data is useful. You absolutely should be capturing first and last touch. And it’s not that it’s not a contributing factor, but you should avoid giving more weight to this information than it deserves just because that’s the data you have available. There is a difference between the fact that someone first found you via a Google Ads campaign, and the inference that they became a customer because of a Google Ads campaign.

Work backwards instead

We often find a more useful approach starts at the other end; take a handful of worthwhile customers and reconstruct as much of each journey as you reasonably can. Not just when they first interacted, but all the ways they interacted along the way, and particularly what appears to have helped them progress. 

This means looking at sales conversations, referrals, events, content, website activity and also long gaps or places deals stall. The key is not just finding more touches, but identifying recurring influences and points of progression 

Once you do this with enough customers, some patterns will emerge. From there it’s important to also look at other prospects who followed a similar path, but didn’t end up becoming customers. Look for differences in the behaviour across the two, and places where the non-customers stalled or dropped out of the journey.

You should come away with more of an idea of the recurring influences as well as possible gaps.

What “fuzzy attribution” means

What you probably won’t have is a certain picture of exactly what was the most important interaction for every given customer. But that’s ok, because patterns across multiple journeys can be more useful than assigning credit for each individual journey.

That gives you a “fuzzy” rather than mathematically exact view of attribution, but this is often more commercially useful.

Going back to the purpose of marketing attribution, it’s so you can make informed decisions about marketing budget and drive more sales. Having a fuzzy attribution approach of what usually works gives you enough evidence to make better commercial decisions. Having the false precision of being able to say this customer came from “source x”, whereas “source y” contributes to 32% of sales etc, is not as useful. A complex B2B journey may not permit mathematically exact credit.

This doesn’t mean we’re advocating giving up on measuring your marketing. Track and record as much as you reasonably can, but keep that information in the right context, and particularly don’t let the easy to measure digital interactions overwhelm the trade show conversations, personal relationships and other offline interactions that often play such a role in B2B sales.

And there are of course some cases where more precise data is worthwhile, including high-volume repeatable journeys and controlled experiments.

The useful standard

A better question than “Can we attribute every sale precisely?” is “Can we reconstruct enough of our important customer journeys to decide what to keep, change or investigate?”

If the answer is no, the first job probably isn’t to build another attribution platform, it’s to improve the visibility of the journeys that matter, and to accept you won’t be able to see everything.

Good attribution means you have enough information to make a better decision.

‍