Attribution the CFO Can Actually Defend

The CFO has seen six attribution decks in the last two years. Each one claimed marketing "drove" 40 to 60 percent of pipeline. None of those numbers matched what actually booked. He knows. You know. Everyone in the room knows. And yet next Thursday at 2pm, you're scheduled to walk in with deck number seven.

This is the playbook for not being deck number seven.

The trick isn't a better model. The trick is dropping the pretense that any model is "right" and replacing it with a story the CFO can defend to the board without flinching. One number. One diagnosis. One acknowledged bias. That's it. If you can't fit your attribution narrative on a Post-it, you've already lost the room.

Let's get into it.

Why Every Attribution Model Is Wrong (and Which Way)

Pick a model. Any model. It's biased. The only question is whether you know which way.

Last-touch is the lazy default in most CRMs. It assigns 100 percent of the credit to whatever happened immediately before the opportunity opened, usually a demo request from a paid search ad or a "Contact Sales" form. Last-touch makes paid search look like a money printer and makes content marketing look like a charity. If your CMO is bragging about Google Ads ROI, ask which model. It's last-touch. It's always last-touch.

First-touch swings the opposite way. It rewards the random whitepaper download from 14 months ago, ignores everything that re-engaged the buyer through three job changes and a procurement cycle, and gives content a halo it hasn't earned. First-touch decks always make the SEO team look brilliant. They're not necessarily wrong, but they're not telling you what closed the deal. They're telling you what started the file.

Linear splits credit equally across every touch. Mathematically clean. Directionally useless. A homepage visit from an intern gets the same weight as the executive briefing that finally got procurement off the fence. If you present linear to a CFO, expect to be asked, with a straight face, whether the email footer link counts as a touchpoint. It does. That's the problem.

Time-decay weights recent touches more heavily. This sounds reasonable until you realize it functionally guarantees the SDR sequence wins every quarter and the content team loses every quarter, regardless of what actually happened. Anything that fired in the last 30 days eats most of the credit. Brand investments made 18 months ago (the kind that put you on the shortlist in the first place) get crumbs.

Data-driven (Shapley, Markov chains) is the best math available. It looks at the conditional probability that an opportunity converts given each touch. Done correctly, it's the closest thing to fair credit assignment that exists. Two problems. First, it's opaque. No CFO is going to nod along while you explain Shapley values. Second, it requires volume that B2B SaaS rarely has. If you close 200 opps a year, your Shapley calculations are statistical noise wearing a lab coat. Google Analytics 4 will happily compute data-driven attribution on 47 conversions. Don't let it.

Every model is biased. The job isn't to find the unbiased one. The job is to pick one, name the bias out loud, and stop pretending.

B2B SaaS Reality Check

Before we get into the math, let's calibrate on what we're actually measuring.

A typical B2B SaaS deal looks like this:

• Sales cycle: 90 to 180 days
• Touches per opportunity: 8 to 25
• Buying committee: 6 to 11 humans, each with their own touch history
• Channels involved: paid, organic, content, email nurture, SDR sequence, sales-led webinar, peer review site, two competitor comparisons, and a Slack community thread

Forrester's research on B2B buying committees consistently lands around six to eleven decision-makers, and Gartner's data shows the typical buyer spends only 17 percent of their evaluation time talking to vendors. The rest is independent research, peer conversations, and internal politicking you'll never see in your CRM.

So when a single-touch model assigns 100 percent of credit to one event in that storm, it's not "simplification." It's malpractice. And when a multi-touch model elegantly distributes credit across 8 events, it's still a model: a story you're telling about a process you can only partially observe. You're inferring causation from clickstream data while the actual deal got pushed over the line by a CIO who saw your CEO speak at a conference your CRM doesn't know about.

This is the level-set you owe the CFO before you show any number. Not because they need to be educated (they don't), but because saying it out loud changes the conversation. You stop selling a number. You start defending a methodology.

The Framing the CFO Will Actually Accept

Here's the move. Stop claiming "marketing sourced X percent of pipeline." Replace it with two clearly labeled, non-overlapping numbers:

Marketing Sourced: Pipeline where the first known touch in the buyer's journey is a marketing-owned channel. Paid, organic, content, webinars, events, partner referrals where the partner is marketing-managed.

Marketing Influenced: Pipeline where any marketing touch occurred in the trailing 90 days before the opp opened, regardless of source. This includes deals that originated from outbound SDR, customer referral, or direct sales motion, but where marketing nurtured, re-engaged, or supported.

These numbers don't add. They overlap. That's the point. You're not claiming additive credit. You're describing two different lenses on the same pipeline.

Why this works for the CFO: it stops the "marketing drove 60 percent of pipeline" theater that he knows is fiction. Sourced is a high-confidence, low-credit number. Influenced is a high-credit, low-confidence number. Together they describe reality without overclaiming. He'll respect it because you stopped pretending you have one true number. You have two honest ones.

A typical mid-market SaaS in the 5 to 20 million ARR range will see something like 25 to 40 percent sourced and 60 to 80 percent influenced. If your sourced number is over 50 percent, your sales team isn't doing outbound. If your influenced number is under 50 percent, your nurture program isn't working or your tracking is broken. Either way, it's a signal worth investigating before the QBR, not during.

A Worked Example: 11 Touches, 127 Days, $180K Opp

Let me walk you through a real-shaped opportunity. Names changed, numbers preserved.

Acme Corp, a 600-person logistics company, opened a $180,000 ARR opportunity with our hypothetical SaaS vendor. From first known touch to opp creation, 127 days elapsed. Eleven recorded touches across three contacts on the buying committee. Here's the journey:

Now, the same opp under five different models:

Same opportunity. Marketing's contribution swings from 0 percent to 100 percent depending on which model the CFO believes. That spread (100 percentage points on a $180K opp) is your honesty budget. Pretending it doesn't exist is what got the last six MOps managers fired or quietly demoted.

Sourced vs Influenced for this opp:

• Sourced: Marketing (first touch was organic whitepaper)
• Influenced: Marketing (multiple touches in trailing 90 days)

Both are true. Neither is the "real" number. Both are defensible.

What Dreamdata, HockeyStack, and Bizible Actually Do Well

The vendor landscape is loud. Let's cut through it.

Dreamdata does account-level stitching, not lead-level. This matters more than the marketing makes it sound. In B2B, the "lead" abstraction is broken. A deal isn't a person; it's an account with multiple humans, multiple emails, and multiple anonymous sessions that need to be reverse-engineered into one timeline. Dreamdata's strength is making that timeline coherent across HubSpot, Salesforce, and your web analytics without you running a SQL warehouse. If you're under 5 million ARR with a small RevOps team, this is the highest-ROI attribution tool on the market right now.

HockeyStack built its reputation on the self-reported attribution overlay: the "how did you hear about us?" question on the demo form, weighted against your machine-tracked attribution. The insight: when you compare what buyers say drove their decision against what your tracking says, the gap is huge and informative. Buyers consistently credit podcasts, peer recommendations, and "I just knew about you," channels that don't show up in any UTM. HockeyStack normalizes that as a layer on top of click-based attribution. It's not magic. It's a correction factor for the things your pixel can't see.

Bizible (now part of Adobe) and 6sense are stronger when you blend intent data into the attribution view. They don't just count your touches. They overlay third-party intent signals so you can see when an account was researching solutions in your category before they ever touched your domain. This is genuinely useful for explaining why a deal closed in 47 days when "your funnel" said it should have taken 120. The catch: enterprise pricing, enterprise complexity, and they shine for 50M+ ARR companies running ABM programs. Below that, you're paying for capability you won't use.

The honest take: any of these tools gives you a better story than spreadsheets and Salesforce campaign influence reports. None of them solves the underlying problem, which is that attribution isn't causation.

Attribution Is Not Causation (Say It Out Loud)

The single sentence that buys you credibility with a CFO: "This is correlation. Causation requires a hold-out test."

A hold-out test means turning off a marketing channel in a defined geography or segment for a defined period and measuring the lift. If you turn off paid search in Boston for 8 weeks and pipeline in Boston drops 30 percent while pipeline in comparable cities holds steady, you have a defensible causal claim about paid search. Without that experiment, all you have is a model, a sophisticated correlation engine.

The tier of marketing organizations that actually run incrementality tests is small. Geo experiments, conversion lift studies, and matched-market tests require enough volume to detect a signal, enough discipline to not panic-restart the channel mid-test, and enough CFO trust to "leave money on the table" for two months while you measure. Most B2B SaaS companies don't have any of those.

If you're not running incrementality tests, the move is to say so. In the deck. Out loud. "We don't have an incrementality test for this quarter, so the bias acknowledgment is that our sourced number could be over- or under-stated by an unknown factor. Next quarter we're proposing a 6-week paid search hold-out in tier-3 markets to calibrate." That's a sentence that gets you funding for the test, not a deck rejection.

The CFO doesn't need you to be right. He needs you to know what you don't know.

How to Present This in 60 Seconds

The QBR slot for marketing attribution is rarely longer than 5 minutes, and the CFO is going to remember 60 seconds of it. Plan for those 60 seconds. Everything else is supporting evidence for someone who already has their question loaded.

The script:

"Marketing sourced $4.2M of new pipeline this quarter — 31 percent of the total. Influenced pipeline, defined as any marketing touch in the trailing 90 days, was $9.8M, or 72 percent. These don't add. The sourced number is conservative because it credits only the first known touch; the influenced number is generous because it captures any touch, including ones that may not have moved the deal. The model is biased toward our paid search channel because it's the easiest to track and the hardest to misattribute. We don't have an incrementality test running this quarter, so I'd flag that as an open question. Next quarter we're proposing a 6-week geo hold-out in three tier-3 markets to calibrate. Questions?"

That's 90 seconds spoken. One number that matters. One diagnosis (paid search bias). One acknowledged limitation (no incrementality test). One forward action (the hold-out proposal).

The CFO will not push back. He's been waiting two years for someone to talk to him this way.

What the CFO Actually Wants

Here's the secret that no attribution vendor will tell you, because they're selling bigger numbers.

The CFO doesn't want a bigger marketing number. He wants a number that doesn't change story every quarter. He wants to be able to defend the marketing budget to the board using the same framework next quarter and the quarter after. He wants a number that, when it goes up, he believes it; and when it goes down, he believes that too.

What kills CFO trust isn't a low attribution number. It's a model change. The quarter you switch from last-touch to multi-touch and marketing's number jumps from 22 percent to 47 percent: that's the quarter you became unbelievable. The quarter the new vendor onboarded and "found" 30 percent more pipeline you'd been undercounting was the quarter you stopped being a strategic partner and started being a sales rep.

Pick a methodology. Document it. Stick with it for at least four quarters. Report sourced and influenced as separate numbers. Acknowledge the bias. Propose the experiment that would calibrate it. Repeat.

That's the deck the CFO defends. That's the seat at the table. Everything else is theater.

Learn More

• Marketing Operations Manager JD
• The MOps Stack That Doesn't Collapse Under Its Own Weight
• Lead Scoring That Sales Will Actually Use
• Pipeline Hygiene Your CRO Will Thank You For