Multi-touch revenue attribution is the practice of distributing credit for closed revenue across the marketing and sales touches that contributed to it. In B2B, where 5 to 12 stakeholders interact with 20 to 60 touches over 6 months, single-touch models lie. The right model depends on what decision you are making — different attribution models answer different questions, and pretending one model rules them all is a mistake that costs marketing budget every year.
The major attribution models
| Model | How it credits | Best used for |
|---|---|---|
| First-touch | 100% to first interaction | Brand and demand-gen budget |
| Last-touch | 100% to final interaction before opp | Conversion-rate optimization |
| Linear | Equal share to every touch | Reporting parity, low signal |
| Time-decay | Weight rises closer to close | Mid-funnel investment decisions |
| U-shaped | 40% first, 40% last, 20% middle | Balanced view, default for most B2B |
| W-shaped | 30% first, 30% MQL, 30% opp, 10% other | When you have clean stage data |
| Data-driven (algorithmic) | ML-derived per-touch weights | Mature stacks with clean data |
No model is “true.” Attribution is a credit allocation policy, not a measurement of physical reality.
How to choose a model
Different decisions need different models:
- “Should I spend more on this channel?” → Use a multi-touch model (U-shaped or W-shaped) so you don’t starve top-of-funnel.
- “Is this campaign creating qualified pipeline?” → Use first-touch or sourced-pipeline.
- “Is this nurture sequence accelerating deals?” → Use time-decay or stage-based.
- “Should I cut this paid keyword?” → Use last-touch on opportunity creation.
The best practical setup is to track at least two models in parallel: a first-touch (for sourcing credit) and a U-shaped or W-shaped (for influence credit). Reporting both prevents single-model gaming.
How to instrument it
- One identity graph. People-level (not session-level) tracking, with company resolution. Without this, you are guessing.
- Touchpoints in one place. Web visits, content downloads, ad clicks, emails opened, demo requests, sales activities, all stitched to the lead and account.
- Closed-loop to revenue. CRM opportunities tagged back to the lead and account so closed-won amounts can flow to the touch history.
- Time bounding. Touches in the 12 months before opportunity create are typically considered; older touches are noise.
- Account-level rollup for ABM. Person-level attribution misses 70 percent of buying-committee activity in enterprise.
Common pitfalls
- Single-model dogma. Reporting only first-touch starves nurture; only last-touch starves brand. Always run two.
- Last-click web analytics on B2B. Google Ads or GA last-click tells you nothing about a 6-month, 8-stakeholder enterprise deal.
- No account-level view. A buying committee with 5 people each having 6 touches looks like 30 disconnected leads without account stitching.
- Comparing channel ROI across models. Each model produces different ROIs for the same channel. Pick a model per decision and stick with it.
Related
- Pipeline velocity — attribution feeds the volume input
- Lead scoring — touchpoint data feeds both