Marketing Analytics · End to End
ServiceNow · Staff Product Manager · 2023–2024
CampaignAttribute IQ
A multi-touch attribution engine built to give every marketing touchpoint the credit it deserves
In B2B marketing, the path from first touchpoint to closed deal is rarely a straight line. A prospect might attend a webinar, download a guide three weeks later, respond to an email, and finally book a demo — all before an opportunity is ever created. Yet most marketing teams measure success by crediting only the last thing that happened. CampaignAttribute IQ was built to change that. Starting from nothing, we designed and delivered the company's first multi-touch attribution model — giving marketing teams, for the first time, a truthful picture of which campaigns were actually driving pipeline.
Multi-touch attribution Markov chain modeling Data modeling Data science collaboration Cross-functional leadership Stakeholder management Go-to-market

What problem were we solving?

B2B buyer journeys are long, multi-threaded, and involve multiple people from the same account interacting across multiple channels — paid search, email, events, content — over weeks or months. Every one of those interactions plays a role in moving a deal forward. But last-touch attribution erased all of it. Only the final touchpoint before opportunity creation received any credit.

The result was that marketing teams had no reliable way to answer the most important question they faced: which campaigns are actually driving pipeline, and which ones are just noise? Budget decisions were made on instinct. High-performing early-stage programs were invisible. And the relationship between marketing activity and revenue was largely a matter of faith.

Last-touch attribution
Webinar Email Content Demo 100% All credit to last touch only
Multi-touch attribution
Webinar 28% Email 22% Content 19% Demo 31% Credit distributed across all touchpoints
What made it difficult?

This was a first-of-its-kind initiative for the marketing team. There was no prior model, no template, no playbook. The starting point was an abstract business pain and the task was to turn that into something measurable, trustworthy, and usable.

The data challenge was significant. B2B journeys involve multiple people from the same account interacting at different stages across different channels over months. Cleaning, validating, and modeling that data required constant collaboration across Marketing Ops, Digital Tech, Sales, Customer Success, and Data Science teams who each had different definitions of the same things.

The harder challenge was human, not technical. Keeping six cross-functional teams aligned on scope, decisions, and tradeoffs — while resolving data quality conflicts in real time — required as much product leadership as the model itself.

How did I approach it?

I started by partnering with Marketing Ops leadership to translate the abstract pain into one specific, answerable business question: which campaigns contribute most to pipeline creation? Everything else flowed from there. With the scope locked to one geography, one region, one business unit, and a 5-quarter data window, the project became deliverable.

Phase 1
Discovery and data exploration
Defined business questions, aligned stakeholders on scope, and audited data quality across marketing touchpoint records.
Phase 2
Model evaluation and selection
Evaluated Markov Chain and Structural Equation Modeling approaches. Selected Markov Chain for its probabilistic, many-to-many attribution framework best suited to B2B journey complexity.
Phase 3
Development and validation
Worked closely with data scientists to ensure modeling logic stayed true to stakeholder intent — facilitating interpretation conversations about real campaign behavior, not just the math.
Phase 4
Visualization design
Proposed the Sankey chart as the primary output — allowing the flow of influence across touchpoints to be read clearly in a single view. Iterated through several designs before landing on the final format.
Phase 5
MVP launch and adoption
Ran demo sessions across marketing portfolios, presented at executive quarterly reviews, and hosted weekly office hours to drive adoption across the organization.
Evaluated
Structural Equation Modeling
Strong for linear, defined journeys
Requires fixed path assumptions
Less suited for multi-threaded B2B accounts
Higher interpretability risk for stakeholders
Selected
Markov Chain Model
Probabilistic, many-to-many attribution
Handles non-linear, multi-person journeys
Distributes credit across all meaningful touchpoints
Best fit for B2B sales cycle complexity
Output visualization — Sankey chart (illustrative)
Direct $18.4M · 31% Content Syndication $12.1M · 25% Paid Search $8.6M · 15% Email Ebook / Content $16.2M · 22% Webinar $12.8M · 13% Seminar $7.3M · 9% Demo Channel Content / Event Type
Numbers are representative and not from actual data.
When did it click?

The shift happened when stakeholders stopped asking "did this campaign work?" and started asking "which combinations worked, and how do we do more of that?" That change in question was the signal that the model had earned trust.

ROI visibility expanded from roughly 30% of campaign spend being tied to outcomes before the model, to over 70% after launch. For the first time, marketing could point to specific interactions and say with confidence: this is what moved the deal forward.

The go-to-market was as important as the model itself. Running demo sessions, presenting at executive reviews, and hosting weekly office hours meant CampaignAttribute IQ was not just built — it was adopted. Teams across Sales and Customer Success began exploring similar data-led approaches off the back of it.

What did it unlock?
100%
Attribution visibility for newly created opportunities within model scope
70%+
ROI visibility post-launch, up from roughly 30% pre-model
+45%
Dashboard MAUs after integrating the attribution Sankey visualization
5x
Increase in stakeholder usage within first quarter of launch

Beyond the numbers, what this project unlocked was organizational confidence. Marketing teams gained a credible voice in revenue conversations. Budget decisions became data-driven. And the model became a springboard for broader analytics initiatives across the organization.

Reflection

Building a great data product is not just about the math — it is about trust. Trust is built through clarity, collaboration, and relentless communication. My role was not to write the code or build the model myself, but to make sure the right problems were being solved, and that every team involved had the same picture of what success looked like. Leading this from zero to one was a defining moment — it sharpened my ability to think strategically, manage ambiguity, and guide a cross-functional team through uncertainty toward meaningful impact.