Most B2B SaaS deals involve five to ten touchpoints across two or more channels and weeks of consideration time. If your reporting credits only the last click, you systematically over-fund bottom-of-funnel channels and starve the awareness channels that opened the journey. Pick the wrong model and you can scale the wrong half of your funnel for an entire quarter without realizing it.
How each model behaves in practice
First-touch attribution gives 100% of the credit to the channel that brought the prospect into your world for the first time. It rewards awareness-building investments — content, organic, brand campaigns — and punishes the bottom-of-funnel work that actually closed the deal.
Last-touch is the inverse. It rewards whatever channel was the final nudge before conversion, which usually means direct traffic, branded search, and remarketing. Use it as a sanity check, not a primary model.
Linear and U-shaped are the practical middle ground. Linear distributes credit equally across paid touches; U-shaped gives 40% each to the first and last touch and 20% to everything in between. U-shaped is the most common choice for SLG demo motions because the first and the closer matter most.
When to use source-specific
Source-specific attribution is Cometly’s answer to the multi-touch problem in PLG. The model gives a channel credit any time it appears anywhere in the journey, not a fractional weighting. If Meta showed up at touch 1 and touch 6, Meta gets full credit on its own report; if Google showed up at touch 3, Google also gets full credit on its own report.
It sounds like double-counting, and at the aggregate level it is. But for individual channel optimization decisions — should I scale Meta? — it answers the question more honestly than any fractional model. Run linear at the executive level and source-specific at the channel-owner level.
What to watch for.
- Mixing models across reports
If your CFO sees first-touch revenue numbers and your media buyer sees last-touch numbers, you’ll spend a meeting reconciling them. Standardize one model per audience.
- Optimizing on last-touch by default
Last-touch makes branded search and direct look like miracles. Use it as a tiebreaker, not a budgeting tool.
- Ignoring direct entirely
Direct traffic is real intent — but it’s rarely the actual origin. Use last non-direct touch to surface the real source behind the direct visit.
- Switching windows and models in the same comparison
If you change two variables at once you can’t isolate the effect of either. Lock the window when comparing models, and lock the model when comparing windows.
Recap.
- First-touch credits the channel that started the journey – best for top-of-funnel awareness
- Last non-direct touch ignores the direct visit and gives credit to the prior real source
- Linear weights every paid touch in the journey – best for multi-channel B2B
- U-shaped gives 80% to the first and last touch, 20% to middle — useful for demo-driven motions
- Source-specific gives a channel credit any time it appears in the journey — built for self-serve SaaS