First Touch, Last Touch, or Neither: Picking an Attribution Model
Pick a marketing attribution model by matching it to the decision you're making, because none of them is fully correct and each distorts the picture differently. First touch overweights awareness, last touch overweights whatever closes the deal, and multi-touch adds weighting assumptions of its own.
What does first-touch attribution overweight, and when is that fine?
First-touch attribution gives all the credit to whatever channel first got someone's attention, which makes sense if your specific question is which channels are good at starting new relationships from zero. It systematically undervalues everything that happens afterward, though, so if you use first touch to decide which channel deserves your entire budget, you'll starve the channels doing the harder, less visible work of actually moving someone from aware to interested to ready to buy, since none of that later work shows up in a first-touch view at all.
Last Touch Overweights Whatever Closes the Deal, Not What Opened It
Last-touch attribution tends to favor sales-adjacent activities: a demo request, a pricing page visit, the final action right before someone converts. This makes outbound and sales-triggered activity look disproportionately effective, because it's often the last thing that happens before a deal closes, even when a long earlier sequence of content and awareness work is what actually made that final touch land. If you're deciding whether to invest in top-of-funnel content based on a last-touch report, you'll almost always conclude it's not working, because last touch was never built to credit that kind of work in the first place.
Does multi-touch attribution fix the problem or add assumptions?
Multi-touch models split credit across several touches, which fixes the all-or-nothing problem of first and last touch but replaces it with a different problem: the specific weighting, equal credit across every touch, more credit to touches near the middle, more credit to touches near the end, is itself a modeling choice that shapes the answer you get. Two teams using different multi-touch weighting schemes on the same underlying data can reach different conclusions about which channel matters most, so it's worth being explicit about which weighting you're using and why, rather than treating multi-touch as one settled, objective answer.
Match the Model to the Specific Decision You're Making
Rather than picking one model for everything, match the model to the question. Deciding where to invest in top-of-funnel awareness spend calls for something closer to first touch or an early-weighted multi-touch view. Deciding which channel to credit for closing already-engaged accounts calls for something closer to last touch. Using one model for every decision guarantees it will be wrong for at least some of them, since no single weighting scheme fairly answers both of those genuinely different questions at once.
Match the model to the question:
- Deciding where to invest in top-of-funnel awareness spend calls for first touch or an early-weighted multi-touch view.
- Deciding which channel to credit for closing already-engaged accounts calls for something closer to last touch.
- Understanding overall channel contribution usually calls for multi-touch, with its weighting choice stated openly.
- Sanity-check whichever model you pick against a handful of deals whose real story you know well.
Sanity-Check Any Model Against a Handful of Real Deals You Know Well
Whatever model you pick, periodically trace a handful of deals you personally remember the real story of and check whether the model's attribution matches what you know actually happened. If the model consistently gives credit to a channel you know, from direct knowledge of the deal, played only a minor role, that's a sign the model's weighting doesn't fit how your specific buyers actually move through their journey, and it's worth adjusting rather than trusting the model's output over your own direct knowledge of how the deal really went.
Write Down the Model's Known Blind Spots Where Everyone Can See Them
No attribution model comes with a label explaining what it systematically undercounts, so write that label yourself and post it next to whatever report the model feeds. A short note next to a last-touch channel report saying this view undercounts early-funnel content by design keeps a well-meaning executive from making a budget call the model was never built to support. This costs almost nothing to set up and prevents the most common misuse of attribution data, which is treating any single model's number as a complete, unqualified answer to a question it was never designed to fully answer.
What Good Looks Like
A team with a mature approach to attribution can say which model they're using for a given decision and why that model's specific distortion is acceptable for that particular question.
Building The Capability (5-Stage Skill Ladder)
How to Get Started
Frequently Asked Questions
Is multi-touch attribution always better than first or last touch?
Not always. Multi-touch is usually a fairer default for understanding overall channel contribution, but it's more complex to set up and maintain, and its weighting choices are themselves assumptions. For a specific, narrow question, a simpler model can sometimes give a clearer, more actionable answer than a multi-touch view that's still technically more accurate.
How often should we revisit which attribution model we use?
Revisit it whenever your buying journey changes meaningfully, for instance if you add a new channel or your sales cycle length shifts, since the model that fit your old journey may distort a longer or differently shaped one. Otherwise, an annual check alongside your broader planning cycle is usually enough.
Can we use different attribution models for different decisions at the same time?
Yes, and it's often the right approach. Just be explicit internally about which model is answering which question, so a first-touch number and a last-touch number for the same channel don't get compared as if they were meant to answer the same thing, which is a common and confusing mistake.
About the numbers
This guide doesn't quote a sourced benchmark. Figures in it are estimates or general guidance, so check them against your own numbers.
Related Guides
Estimating the Word-of-Mouth Traffic Attribution Misses
Why 'direct' traffic is often a placeholder for untracked sharing, the proxy signals that help estimate it, and why chasing perfect tracking usually backfires.
Where B2B Conversion Rates Actually Fall Short of Plan
Why comparing your funnel to a published conversion benchmark usually misleads you, and what to check on your own numbers instead.
Building a Marketing Budget Model That Survives a Board Question
How to build a marketing budget allocation model with enough detail behind it to answer any board question about where the money goes.
When Content Syndication Pays Off, and When It's Just Expensive Leads
How to judge whether a content syndication network will bring in usable pipeline or just a list of names that never respond to a follow-up call.
Owning the Search Results for Your Category's Highest-Intent Terms
A practical way to identify and win the small set of search terms that show up right before someone buys, instead of chasing broad traffic volume.
Forecasting Marketing ROI Without Fooling Yourself
A way to build a marketing ROI forecast that shows its own uncertainty instead of presenting one confident number to the board.