Last month, a customer looked at our influence attribution model and said: 'Wait. This campaign is getting 100% credit. And so is this one. And this one. The math doesn't add up.'
She wasn't wrong to be confused. She was wrong to think attribution should add up to 100%.
But here's what's really happening: She was using the wrong model to answer the wrong question. And that confusion is why most marketing teams kill their demand gen, defund the channels that actually build pipeline, and then blame "market conditions" when everything dries up six months later.
The problem is that most people think about attribution as a zero-sum game, which it is not.
The reality:
You orchestrate campaigns with a hypothesis about what works at each stage.
Different models test different parts of that hypothesis.
Rules-based MTA shows if channels play their intended role (awareness, education, conversion)
AI-based MTA finds patterns you didn't plan for &
Marketing Mix Modeling + incrementality shows what actually causes results
Most teams have attribution in place but never test whether their orchestration hypothesis is actually true. They just assume LinkedIn drives awareness, webinars educate, and case studies close, without ever validating it.
Your attribution model is probably fine. Your hypothesis about how channels work together? That might be the problem.
How Influence Attribution Actually Works
Here's what threw my customer off:
When our Solutions Engineer (who also happens to be an ex-marketer) explained: "This is an influence model. So it's giving a hundred percent credit for every single touch point."
The scenario: A $180K deal was influenced by 10 campaigns. In an influence attribution model, each campaign receives 100% credit for its contribution.
If you're thinking in linear attribution terms, this sounds insane:
Linear would give each campaign 10% credit
First-touch would give one campaign 100%, the rest 0%
Last-touch would give the final campaign 100%, the rest 0%
But influence says: Every one of those 10 campaigns gets 100% credit.
The math doesn't "add up" because it's not supposed to.
Influence isn't measuring "how much credit" each campaign deserves. It's measuring "how many deals did this campaign touch?"
That's a completely different question. And when you conflate the two, when you try to use influence data to make budget allocation decisions - you end up with nonsense.
The Three Layers of Measurement (And Why You Need All of Them)
Stop looking for the "one true model." You need three different tools:
Rules-Based Attribution: "Is each channel doing its job?"
Tests whether your orchestration hypothesis is working.
You thought LinkedIn drives awareness. Rules-based MTA shows it's appearing at the bottom of the funnel, at conversion.
Discovery: Your ICP uses LinkedIn differently than you assumed.
AI-Based Attribution: "What am I missing?"
Finds patterns you never designed for.
For example, You built webinar → nurture → demo. AI discovers buyers who went webinar → competitor comparison → demo converted 3x faster.
You never orchestrated that path, but buyers preferred it.
Marketing Mix Modeling + Incrementality: "What actually causes results?"
Shows causal impact, not just correlation.
For example, Influence says LinkedIn touches 80% of deals. MMX reveals you're spending $800K when optimal is $600K. You're past diminishing returns - spending more is actually hurting ROI.
Bonus: Journey Analysis shows where deals stall. Your "education" content slows deals by 14 days instead of moving them forward.
A scenario is where most teams torch their entire demand gen budget.
Here's the pattern I see, The 8-Step Death Spiral:
Marketing ops sets up multiple attribution models.
CFO interrupts: "So which one is right?"
Marketing tries to explain they're different "perspectives"
CFO loses patience: "For the love of God, just tell me which campaigns are working"
Team defaults to last-touch (because it's simple)
Brand awareness and demand gen get defunded
Pipeline dries up six months later
Everyone blames "market conditions" The mistake: Treating different attribution models as competing versions of truth.
The reality: There are different instruments in your measurement stack, designed to test different parts of your GTM hypothesis.
It's like asking: "Should I use a thermometer or a barometer?" Depends. Are you measuring temperature or pressure?
Putting the Pieces Together: A Customer Story
One of our customers ran two campaigns. Here is what different models revealed:
Campaign A: Webinar Series
Campaign B: Targeted LinkedIn ABM Campaign
Through the Rules-Based MTA/ Linear attribution Lens
Campaign A : Top and middle funnel, driving awareness and education, $400 K spend, $2.5M attributed pipeline credit
Campaign B : Bottom funnel, driving conversion, $380K spend, $2.4M attributed pipeline credit
Insight: Both playing their intended roles, they both look roughly equal.
Through the Influence Lens:
Campaign A : $8M total influenced pipeline (touching 200 opportunities)
Campaign B : $3M total influenced pipeline (touching 80 opportunities)
Insight: Webinar has 2.5x the penetration across pipeline
Through the MMX/Incrementality Lens:
Webinar series: $400K spend, Incrementality test showed it generated $3.2M in truly incremental pipeline Incremental multiplier: 8x (strong demand creator)
LinkedIn ABM: $380K spend, Incrementality test showed it generated $608K in truly incremental pipelineIncremental multiplier: 1.6x (demand capturer)
Insight: Webinar creates new demand (8x). LinkedIn ABM captures existing demand (1.6x), mostly harvesting what the webinar and other channels already generated.
Through the Journey Analysis Lens:
Accounts that attended webinar moved through stages 18 days faster
Campaign B accounts showed no velocity improvement
Insight: Webinar removes friction; Campaign B just captures ready-tobuy demand
If you only looked at linear attribution, these campaigns looked roughly equal.
If you only looked at last-touch, Campaign B would win.
But the full picture reveals: Campaign A is building demand and accelerating pipeline. Campaign B is harvesting demand that Campaign A created.
Cut Campaign A, and Campaign B's results will fall six months later.
What Each Model Actually Answers
Let's be very clear about when to use what:
The question isn't: "Which model is right?"
The question is: "What am I trying to learn about my GTM motion?"
What That CMO Did Next (And Why It Worked)
Remember that customer who thought our model was broken?
Here's what she changed:
She stopped fighting about which model was "right."
Started asking better questions:
"Is my hypothesis about LinkedIn playing an awareness role accurate?" (Rules-based MTA)
"Which campaigns are present in our best deals?" (Influence)
"What's the optimal spend before I hit diminishing returns?" (MMX + Incrementality)
"Where are buyers deviating from my intended journey?" (AI-based MTA + Journey Analysis)
"What closes the fastest?" (Last-touch + Velocity)
She reframed her entire board deck.
Before: "Marketing generated $11M pipeline" (last-touch attribution)
After: Our orchestration hypothesis tested across 4 measurement layers:
Top 3 campaigns were present in 80% of deals worth $500K+ (Influence)
Webinar series drove 8x incremental lift (MMX)
AI detected unexpected pattern: demo → case study → close converts 3x faster than intended path Webinar → Email drip campaign → White paper → Case study → Demo → Close (AI-based MTA)
Journey analysis revealed 14-day delay in Stage 2 from white paper (friction point identified)
The CFO stopped questioning the marketing budget.
Because she wasn't defending activities anymore. She was showing:
Strategic reach
Hypothesis testing
Causal impact
Continuous learning
She moved from "trust me, marketing works" to "here's how we're systematically testing and optimizing our GTM motion”.
Every week, I sit down with CMOs and marketing leaders who are reshaping their organisations with new tactics and AI. We unpack what’s actually working, where attribution breaks down, and how to turn messy marketing data into confident decisions.
If you care about the future of marketing, subscribe and join us.
Know a CMO who's struggling with attribution measurement? Forward them this issue.