A regional SaaS company we worked with was confident their gated whitepaper campaign was driving their best enterprise deals — the attribution dashboard said so. Then someone actually called the five biggest deals that quarter and asked how they'd heard about the company. Four of the five said a personal referral. The whitepaper download those four had done was real. It just happened after the referral conversation, not before it — because that's what people do when someone recommends a vendor: they go look them up.
The MQL That Never Should Have Counted
This is the quiet failure mode in most B2B attribution: the model doesn't lie about what happened, it lies about what mattered. A lead who downloads a whitepaper, attends a webinar, and then converts looks identical in most CRMs to a lead who was already sold by a referral and did those same actions out of due diligence, not persuasion. Both get the same attribution credit. Only one of them represents marketing actually creating the deal.
Why First-Touch and Last-Touch Attribution Both Lie
First-touch attribution gives all the credit to whatever brought someone into your CRM initially, which overweights top-of-funnel content and underweights everything that actually moved a stalled deal forward. Last-touch does the opposite — it credits whatever happened right before close, which is very often sales activity, making marketing's contribution look smaller than it was for anything that happened months earlier.
- First-touch attribution rewards volume — it will tell you your blog is your best channel even if most of that traffic never converts, because it's counting who showed up first, not who closed.
- Last-touch attribution rewards recency — it will tell you sales calls are your only effective channel, because sales activity is almost always the final recorded touchpoint before a deal closes.
- Neither model distinguishes between a touchpoint that changed a buyer's mind and one that simply happened to occur during their evaluation.
- Most attribution disputes between marketing and sales are actually disputes about which of these two flawed models to trust, not disputes about the underlying data.
What We Actually Recommend
Don't try to build a perfectly accurate attribution model — it doesn't exist, because buyer intent isn't fully observable. Build a model that's honest about its own limitations, and combine it with the low-tech method that still works: asking closed-won customers directly how they found you, and comparing that answer against what the dashboard says.
A Multi-Touch Model That Marketing and Sales Both Trust
The models that actually survive contact with a skeptical sales team weight touchpoints by buying stage, not just chronological order. A touchpoint during active evaluation — a demo request, a pricing page visit, a technical Q&A — carries more weight than an early-stage blog read, because it's closer to an actual buying signal.
- Awareness-stage touchpoints (blog, social, early webinar attendance): light weighting — they build the pipeline that later touchpoints convert, but rarely close deals on their own.
- Consideration-stage touchpoints (case study downloads, comparison content, second webinar attendance): moderate weighting — this is where a buyer is narrowing options.
- Decision-stage touchpoints (demo requests, pricing conversations, technical deep-dives): heavy weighting — the strongest available signal that marketing activity is influencing an active deal.
- Sales-sourced or referral-sourced deals: flagged separately, not folded into channel attribution at all — crediting marketing for a deal marketing didn't influence just corrupts the model for the next quarter's budget conversation.
The ROI Number That Actually Means Something to Finance
"4.2x ROI" is a real number we've measured with clients, but the number that actually earns marketing budget in a board conversation is more specific than that: cost per qualified pipeline generated, next to cost per customer actually acquired, tracked separately by channel. The gap between those two numbers, per channel, tells you where spend is generating activity versus where it's generating revenue — and it's usually not the same channel.
The uncomfortable truth for most B2B marketing teams is that a defensible attribution model will sometimes tell you your favorite channel isn't your best one. That's not a reason to avoid building it. It's the entire reason it's worth building — a model built to always agree with the existing budget allocation isn't measuring anything, it's just decorating a decision that was already made.