B2B Attribution Is Broken: Trace It, Don’t Model It

You are three slides into the quarterly review when the CFO asks the only question that matters: which of these programs actually made us money? Your attribution dashboard has an answer. Sales has a different one. Your customers, if anyone asked them, would give a third. And everyone in the room quietly suspects none of the numbers are real.

That suspicion is measurable. In Gartner’s 2024 Marketing Analytics Survey, only 52% of senior marketing leaders said they can prove marketing’s value and get credit for it - and the most skeptical audiences are exactly the people who set your budget: CFOs (40%) and CEOs (39%). Why B2B marketing attribution is broken is not an abstract question. It has a specific answer: for twenty years the industry has tried to model where revenue came from, when what a board actually trusts is a record.

There are three ways to answer “where did this revenue come from.” Two of them are familiar. The third is the one worth building on.

Marketing attribution models: where the software breaks

The first way is software attribution: first-touch, last-touch, position-based, or full multi-touch. Tracking code watches every click, and a formula splits credit for each deal across the touchpoints it saw. It sounds rigorous. The math is fine. The input is the problem, because a model can only distribute credit across the touches it can see - and in B2B, most of the journey is invisible.

Start with when buyers show up. 6sense’s Buyer Experience research (900+ buyers, 2023) found B2B buyers are roughly 70% of the way through their journey before they first contact a seller - and they initiate that contact 83% of the time. By the time your tracking code gets a named person to follow, most of the decision already happened somewhere you could not see.

Then look at how long and wide the journey runs. Dreamdata’s GTM Benchmarks (2022, 414 accounts) measured an average B2B journey of 192 days and 31 touches from first touch to close - 233 days and 34 touches for B2B SaaS. Those touches are spread across colleagues who never fill out a form, personal devices, Slack communities, podcasts on a commute, and group chats your pixel will never enter.

And the tracking itself has been collapsing. When Apple shipped App Tracking Transparency with iOS 14.5, Flurry Analytics measured that only about 4% of US iPhone users opted in to tracking at launch (12% worldwide); opt-in later stabilized around 20-25%, which still means most of the mobile journey is dark by choice. Meanwhile the CRM data the model leans on decays underneath it - industry estimates put B2B contact data decay at roughly 2-3% per month.

This is why last touch attribution produces the absurd reports every operator has seen: an 11-month, 31-touch journey, and the demo form gets 100% of the credit because it was the one touch the software could see clearly. Multi-touch spreads the error around instead of concentrating it, but it is still allocating credit over maybe a tenth of what actually happened.

Self-reported attribution: honest but lossy

The second way is to ask. Put “how did you hear about us?” on every form, and have sales ask it on every first call. Self-reported attribution became popular precisely because it surfaces what software cannot.

Self-reported data is honest about the invisible journey. But it is lossy in its own way. Memory is fuzzy, buyers compress a year of touches into one answer, and “I saw you around” credits a channel, not a message. You cannot tie a dollar to a specific campaign with it, you cannot run it weekly, and you cannot defend it line-by-line when a skeptical CFO leans in.

The third way: traced attribution

Here is the alternative, and it is worth naming: traced attribution. The rule is simple. You only claim revenue where you hold the receipt - an unbroken, recorded chain of events from a specific campaign send, to the reply it got, to the meeting that reply booked, to the deal that closed. No weighting formula. No probabilistic credit. If the chain exists, the revenue is yours to claim. If it does not, you do not claim it.

Traced attribution undercounts on purpose. It will never credit the podcast that warmed the buyer up. What it gives you instead is a floor - a number nobody in the room can argue with, because every dollar on it is backed by a record anyone can click into and read.

The reason traced attribution has become practical is that more of the funnel now runs through recordable conversations. A campaign email or text goes out to a known contact. The reply comes back on the same thread. The meeting gets booked from that thread. Each link in the chain is an event that happened, not a weight someone assigned. This is also why speed to lead and attribution are secretly the same project: the team that answers every reply fast is also the team whose revenue chain never breaks.

What about marketing mix modeling?

MMM deserves an honest mention, because it is the statistically serious way to estimate what tracing and asking both miss. But it has enterprise economics. Industry estimates put managed MMM at roughly EUR 30K-150K per year, generally worth it only above about EUR 300K-500K in annual ad spend, and it needs around two years of weekly data before the model has anything to chew on. If you are the first marketing leader at a Series A or B company, that is not your tool yet. It is a fine layer to add later, on top of a traced foundation - not a substitute for one.

How to run traced attribution this quarter

You do not need a data team. You need your outbound and your replies in one system, so the chain is recorded as it happens instead of reconstructed after the fact.

1. Record every send at the contact level. Not “the March campaign went out” - this message, to this person, at this time.
2. Land every reply in the same system. The moment a reply gets forwarded into a personal inbox, the chain breaks and the deal becomes unattributable.
3. Link meetings and closed deals back to the thread. The booking that came from the reply, the deal that came from the meeting.
4. Report the floor. Bring the traced number to the board as “revenue we can prove,” with self-reported data alongside as “demand we can see but not trace.” That two-line format is the honest version of the marketing report your board actually wants.

This is the part Marqeable was built for. Campaigns go out over email and text, replies land in one inbox, and revenue attribution links dollars to the exact message that started the chain - recorded events, not modeled attribution. When a deal closes, you can open the thread and read the receipt. It is also a useful filter for AI claims in this category: plenty of tools promise smarter models, and what CMOs get wrong about AI is assuming a smarter model fixes a visibility problem. It does not. A record does.

Where tracing stops

Traced attribution is not the whole answer, and pretending otherwise would repeat the industry’s mistake in a new costume.

• It covers conversation-centered funnels. If revenue moves through emails, texts, chat, and booked meetings, tracing works. A pure self-serve motion with no conversations will trace very little.
• Brand and dark-social demand still deserve self-reported data. The podcast, the community thread, the conference hallway - keep asking “how did you hear about us,” forever.
• Big ad spenders can layer MMM on top. Past the spend threshold above, a model estimating incrementality across channels earns its keep. Use all three, each for what it is actually good at.

Frequently asked questions

What is traced attribution?

Traced attribution means only claiming revenue you can back with a recorded chain of events - the campaign send, the reply, the meeting, the closed deal - instead of modeling credit across touchpoints. It reports a verifiable floor rather than a modeled estimate.

What is the difference between traced attribution and multi-touch attribution?

Multi-touch attribution splits credit across every touchpoint software can see, using a weighting formula. Traced attribution assigns credit only where an unbroken recorded chain connects a specific message to a reply, a meeting, and revenue - no weights, no guessing.

Should we stop asking customers how they heard about us?

No. Self-reported attribution is your only window into dark social, podcasts, and word of mouth. Keep the question on every form and in every sales call, and read the answers alongside your traced numbers.

Is marketing mix modeling worth it for a startup?

Usually not yet. Industry estimates put managed MMM at roughly EUR 30K-150K per year, and it generally pays off above about EUR 300K-500K in annual ad spend with around two years of weekly data. Most Series A and B teams are below that line.

The bottom line

Attribution software models a journey it mostly cannot see. Asking buyers is honest but cannot tie a dollar to a message. Traced attribution is the third way: claim only what you can prove with a recorded chain from send to reply to meeting to closed deal, report it as a floor, and keep self-reported data for everything the chain cannot reach. It is a smaller number than your dashboard shows today - and it is the first one your CFO will actually believe.

See it live: Marqeable’s revenue attribution links dollars to the exact message that started the chain, with campaigns and every reply in one shared inbox so the record never breaks.

Marqeable runs your campaigns, answers every visitor, text, and email in seconds, and turns them into booked jobs and meetings - even at 9pm on a Saturday. We’re in private beta with a small early cohort. Get early access