Background and Challenge
A mid-size specialty pharmaceutical company with a flagship immunology brand, ImmuVex (a biologic for moderate-to-severe ulcerative colitis), faced a fundamental budget allocation problem. The brand was investing $50 million annually across six HCP-facing channels, but the marketing leadership team had limited confidence in how that budget was distributed. The existing allocation was based largely on historical precedent, internal politics, and last-touch attribution logic that overwhelmingly credited the field sales force for prescription outcomes.
The brand's $50 million annual marketing budget was allocated as follows at the start of the initiative:
| Channel | Budget | Share | Assumed Contribution |
|---|---|---|---|
| Field Force (Reps) | $22M | 44% | 60% |
| Approved Email | $6M | 12% | 8% |
| Digital (NPP, Programmatic) | $5M | 10% | 5% |
| Webinars & Virtual Events | $4M | 8% | 7% |
| Live Events & Congresses | $5M | 10% | 10% |
| eSampling | $8M | 16% | 10% |
The assumption that the field force drove 60% of prescribing outcomes was based on last-touch attribution: whichever channel touched the HCP immediately before a new prescription was credited with the full conversion. Because reps were the most frequent last-touch point (seeing high-value HCPs monthly), they accumulated the majority of attribution credit by default.
The brand team suspected this was inaccurate. Digital channels had expanded significantly, webinars were generating strong engagement, and eSampling was driving trial behavior. But without a proper multi-channel attribution model, they could not prove it. Meanwhile, the board was pressuring the team to justify the field force investment, which had grown 12% year-over-year despite flat prescribing growth.
"We were allocating $22 million a year to our field force based on an attribution model that was essentially 'last rep to knock on the door gets all the credit.' We knew that was wrong, but we could not articulate what the right answer was." — Director of Marketing Analytics
Approach
The team undertook an 18-month initiative to implement a proper multi-channel marketing attribution model. The project was structured in three phases.
Phase 1: Data Foundation (Months 1-6)
The first challenge was building a unified data layer that connected all six channels to individual HCP-level prescribing outcomes. This required integrating data from seven different systems:
- Veeva CRM: Rep call activity, call notes, samples distributed, and HCP access status
- Veeva Vault: Approved email sends, opens, clicks, and content-level engagement
- Demand-side platform (DSP): Digital ad impressions, clicks, and HCP-level verification through NPI matching
- Webinar platform: Registration, attendance, engagement duration, and Q&A participation
- Event management system: Congress booth interactions, symposia attendance, and peer-to-peer event participation
- eSampling platform: Sample requests, fulfillment, and HCP acknowledgment
- IQVIA/Xponent: Prescription data (TRx and NRx) at the individual HCP level, linked via NPI
The data integration was the most resource-intensive part of the project, requiring three months of engineering work to establish a common HCP identifier across all systems and build a time-stamped interaction log. The resulting dataset contained 4.2 million individual channel interactions across 18,600 target HCPs over a 24-month lookback period.
Phase 2: Attribution Modeling (Months 6-12)
The team implemented three attribution models in parallel to compare results and build organizational confidence in the findings:
- Rule-based models (Last-Touch, First-Touch, Linear, Time-Decay): Implemented as a baseline comparison. These models distributed attribution credit using fixed rules and required no statistical estimation.
- Markov Chain model: A probabilistic model that estimated each channel's contribution by measuring its "removal effect" - how much overall conversion probability would decrease if that channel were removed from the HCP journey entirely.
- Regression-based model: A logistic regression model that estimated the probability of a new prescription as a function of channel exposure counts, recency, and interaction quality scores, controlling for HCP baseline prescribing propensity and market factors.
The Markov Chain model was selected as the primary attribution model after validation against held-out data showed the strongest predictive accuracy (AUC of 0.78 versus 0.71 for the best rule-based model). The regression model served as a secondary validation check.
Phase 3: Budget Reallocation and Optimization (Months 12-18)
Armed with the attribution results, the team redesigned the budget allocation for the following fiscal year. This involved scenario modeling to estimate the TRx impact of different allocation strategies, building consensus with the field force leadership team (who were understandably concerned about budget reductions), and implementing a phased reallocation to minimize operational disruption.
Results
The Attribution Reality Check
The multi-touch attribution model revealed a dramatically different picture from the assumed last-touch attribution. The actual channel contributions to prescribing outcomes were:
| Channel | Assumed | Actual (Markov) | Difference |
|---|---|---|---|
| Field Force (Reps) | 60% | 40% | -20 pts |
| Approved Email | 8% | 12% | +4 pts |
| Digital (NPP, Programmatic) | 5% | 15% | +10 pts |
| Webinars & Virtual Events | 7% | 11% | +4 pts |
| Live Events & Congresses | 10% | 9% | -1 pt |
| eSampling | 10% | 13% | +3 pts |
The most striking finding was the field force attribution gap. The last-touch model credited reps with 60% of prescribing conversions, but the Markov Chain model estimated their true contribution at 40%. This did not mean the field force was unimportant - it remained the single largest contributor - but it was significantly over-attributed because reps were often the last channel to interact before a prescription was written, even though digital channels and webinars had done the heavy lifting of building awareness and clinical confidence earlier in the journey.
Conversely, digital non-personal promotion (NPP) was dramatically under-attributed. The last-touch model credited digital with only 5% of conversions, but the multi-touch model estimated 15%. Digital touchpoints were frequent first-touch and mid-funnel interactions that primed HCPs for later rep conversations. Without the digital exposures, many HCPs would never have progressed far enough in their familiarity with the product to warrant a rep visit.
Budget Reallocation
Based on the attribution findings, the team reallocated approximately $10 million (20% of the total budget) for the following fiscal year:
- Field force: Reduced from $22M to $18M (-$4M). This was achieved by reducing headcount by 12 reps (from 142 to 130) through attrition rather than layoffs, and by optimizing territory alignment to eliminate overlap.
- Digital NPP: Increased from $5M to $9M (+$4M). Investment redirected toward HCP-targeted programmatic, Doximity campaigns, and EHR point-of-care advertising.
- Webinars: Increased from $4M to $6M (+$2M). Expanded from 12 to 20 webinar programs annually, with more personalized content tracks.
- Approved email: Increased from $6M to $7M (+$1M). Added behavioral trigger sequences and dynamic content personalization.
- eSampling: Maintained at $8M but shifted from 70% physical to 55% electronic sampling.
- Live events: Reduced from $5M to $3M (-$2M), consolidating from 8 congress activations to 5 with higher quality engagements.
Outcomes After Reallocation
The reallocated budget produced a 14% increase in TRx volume over the subsequent 12 months, despite the total budget remaining at $50M. This translated to an estimated $12.4 million in incremental net revenue. The digital NPP investment delivered the highest marginal ROI, generating $3.80 in revenue per dollar spent versus $2.10 for the field force (though the field force still had the highest absolute contribution at its reduced budget level).
New prescriber acquisition improved by 18%, driven primarily by the expanded digital and webinar programs reaching HCPs who had not been accessible to the field force. The webinar program alone contributed 340 new prescribers, a 45% increase over the prior year, at a cost-per-new-prescriber of $17,647 compared to $31,428 for field-force-driven new prescriber acquisition.
Importantly, the field force maintained its prescription volume despite the budget reduction. The 12 fewer reps were offset by improved territory alignment (fewer HCPs per rep, but better-matched to prescribing potential) and the enhanced digital-to-field handoff enabled by the attribution data. Reps now had visibility into which HCPs were engaging with digital content and could prioritize follow-up accordingly.
Key Takeaways
Lessons Learned
- Last-touch attribution systematically over-credits the field force. In pharmaceutical marketing, reps are often the final touchpoint before a prescription, but they are rarely the only driver. Multi-touch models reveal that digital, email, and webinar channels contribute significantly more than last-touch models suggest.
- Invest in the data foundation first. The data integration work (unifying CRM, email, digital, events, and Rx data at the HCP level) was the most challenging and time-consuming phase, but it was essential for credible attribution results.
- Use multiple models for organizational buy-in. Running rule-based, Markov Chain, and regression models in parallel helped the team build confidence in the results and address skeptics. When three different approaches point in the same direction, the findings are harder to dismiss.
- Reallocate incrementally, not drastically. The team phased the $10M reallocation over two quarters and used attrition rather than layoffs for the field force reduction. This maintained organizational support and minimized disruption.
- Digital channels drive new prescriber acquisition efficiently. The cost-per-new-prescriber through digital and webinar channels was 44% lower than through the field force, making these channels particularly valuable for expanding the prescriber base.
The MCM attribution model is now a permanent part of the brand's annual planning process. The marketing team refreshes the attribution weights quarterly based on the latest 12 months of data and uses the model to simulate the TRx impact of proposed budget changes before committing. The approach is being rolled out to two additional brands in the immunology franchise in 2026.
Try the Tools Used in This Case Study
Build your own multi-channel attribution model and optimize your marketing budget with our free calculators.