Attribution Models April 2026

The 7 Attribution Models Ranked by When They Work for B2B SaaS

Ayoub Kaddouri
By Ayoub Kaddouri
Growth Hacker for B2B SaaS · €1M+ revenue tracked · LinkedIn
Updated Apr 2026

Every attribution tool supports multiple models and every marketer picks whichever one the tool defaults to. That is wrong. Model choice matters more than tool choice. Here is every model, ranked by when it actually gives you useful budget decisions.

Get Attribution Help Read the full tools guide

In This Guide

  1. 1. First-Touch
  2. 2. Last-Touch
  3. 3. Linear
  4. 4. Time Decay
  5. 5. Position-Based (U-Shaped)
  6. 6. W-Shaped
  7. 7. Data-Driven
  8. 8. Decision Matrix

1. First-Touch

How it works: 100% of the credit goes to the first interaction.

When it wins: You are trying to measure demand-creation channels like cold LinkedIn Ads, content SEO for top-of-funnel queries, or podcast appearances. First-touch rewards the channel that created the user.

When it lies: You think first-touch reflects all the work it took to close the deal. It does not. It ignores everything after the first click. including sales, nurture, case studies, pricing page revisits. Use first-touch for demand-creation analysis, not for credit assignment of closed revenue.

2. Last-Touch

How it works: 100% of the credit goes to the final interaction before conversion.

When it wins: Short, simple funnels with one dominant conversion event. DR-heavy e-commerce below $500k revenue. Transactional info products with a single ad-to-purchase flow.

When it lies: Almost everywhere else. For B2B SaaS, last-touch over-credits branded search, retargeting, and "direct" traffic. systematically hiding the channels that originated demand. This is the single most damaging default in marketing analytics.

3. Linear

How it works: Equal credit to every touchpoint.

When it wins: Short funnels (2–5 touches) where touchpoints have comparable cost and no single one is dominant. Small info product funnels, early-stage SaaS.

When it lies: Long B2B cycles with dominant touchpoints. Retargeting-heavy funnels (cheap touches dilute expensive ones). Branded search after paid demand gen. See my deep dive on linear attribution.

4. Time Decay

How it works: More credit to touches closer to conversion. Typical half-life: 7 days.

When it wins: Sales-led motions where later touches (SDR calls, demo, pricing pages, case studies) genuinely do more work than earlier ones. Works for mid-market B2B SaaS with 30–90 day cycles.

When it lies: Top-of-funnel matters. Time decay under-credits demand creation by design. if the LinkedIn Ad that created the user was 90 days before conversion, it gets almost nothing. For brands that depend on cold demand, this is a slow bleed.

5. Position-Based (U-Shaped)

How it works: 40% to first touch, 40% to last touch, 20% split across middle touches.

When it wins: B2B SaaS with a cold-to-close flow where both demand creation AND closing touches matter. My default recommendation for most B2B SaaS operators under $10M ARR.

When it lies: Your funnel has more than 2 inflection points. For example, if there is a critical "SAL" milestone in the middle that represents sales qualification, U-shaped under-credits it.

6. W-Shaped

How it works: 30% to first touch, 30% to lead capture (form fill or MQL), 30% to SAL (sales qualified), 10% split across remaining touches.

When it wins: B2B SaaS with a clear funnel: inbound → MQL → SAL → closed. Matches the reality of a sales-led motion better than any simpler model. Best for $1M–$10M ARR teams with defined pipeline stages.

When it lies: PLG or self-serve motions where lead capture and SAL are the same event (a signup). W-shaped collapses into something simpler and loses its advantage.

7. Data-Driven

How it works: An algorithm (Markov chain, Shapley value, or proprietary ML) determines credit weights based on your conversion data. GA4 defaults to this. HockeyStack and Dreamdata offer versions.

When it wins: High conversion volume (500+ events per month per channel). Enterprise B2B or high-velocity DTC. The algorithm needs signal; at volume, it beats any rule-based model.

When it lies: Low-volume funnels. Under 100 conversions per month, the algorithm over-fits to noise. Numbers will shift dramatically between reporting periods without any real change in the business. If you are under $2M ARR, data-driven is a distraction.

Decision Matrix

Your Situation Use Skip
B2B SaaS, $1M–$10M ARR, mixed inbound U-shaped (default) or W-shaped Last-touch, data-driven (too little data)
B2B SaaS, $10M+ ARR Data-driven + W-shaped comparison First-touch alone
DTC, $500k–$5M U-shaped or time decay Linear (biased by retargeting)
DTC, $10M+ Data-driven + MMM Any single rule-based
Info product, 7–14 day funnel Linear or time decay First-touch
Heavy demand-creation measurement First-touch (as secondary model) Last-touch, time decay
Heavy DR / conversion measurement Last-touch (as secondary model) First-touch

The best operators look at 2–3 models simultaneously. If U-shaped and last-touch agree on channel rankings, you can trust the answer. When they disagree dramatically, that is the interesting case. investigate what is actually happening, then pick the model that matches your funnel shape.

More attribution reads

From the field

Across hundreds of B2B SaaS and DTC audits, I see teams default to whatever attribution model their tool defaults to. rather than picking the model that matches their funnel shape. The result is always the same: systematically wrong budget decisions that look right on paper. The single cheapest upgrade most teams can make is opening two models side-by-side in GA4 or HockeyStack and asking where they disagree.

Frequently Asked Questions

Which attribution model is best for B2B SaaS?

For B2B SaaS, W-shaped (position-based) is the best rule-based starting point: 30% to first-touch, 30% to lead conversion, 30% to opportunity creation, 10% to middle touches. It rewards milestones that move deals between stages. Once you exceed ~600 conversions/month per channel and have 90+ days of data, upgrade to data-driven attribution using Markov chains or Shapley values. Combine with CRM-sourced data and offline conversion tracking. pure GA4 attribution misses too much of a B2B journey.

Which attribution model is best for ecommerce?

For ecommerce with short cycles (under 7 days from first touch to purchase), data-driven attribution typically wins because conversion volume easily clears thresholds. If DDA isn't available, time-decay is a strong second choice: it weights recent touches more heavily, matching ecommerce's impulse-driven behavior. For high-AOV considered purchases (furniture, electronics), use position-based to credit both discovery and closing channels. Avoid pure last-click. it dramatically over-credits branded search and retargeting.

What's the difference between rule-based and data-driven attribution?

Rule-based attribution applies fixed credit-splitting rules you choose: first-touch gets 100%, linear splits evenly, W-shaped weights specific milestones. Simple and transparent, but reflects assumptions, not reality. Data-driven attribution uses ML. Markov chains, Shapley values, or Google's algorithm. to assign credit based on how each channel statistically lifts conversion probability. More accurate with volume (600+ conversions/channel/month) but a black box. Pragmatic answer: start rule-based, upgrade to DDA when data volume justifies it.

Should I use multiple attribution models simultaneously?

Yes. running multiple models side by side is one of the most underrated attribution practices. Compare first-touch and last-touch to see how much credit shifts; the gap reveals which channels are demand creators vs closers. Run rule-based and data-driven in parallel to validate the ML model's signal. Use one model for budget decisions and another for sanity-checking. The risk is over-rotating on whichever model tells the story you want; mitigate by defining the decision rule before looking at numbers.

How often should I re-evaluate my attribution model?

Quarterly at minimum, with a deeper review annually. Re-evaluate sooner if you change channel mix significantly (launch a new paid channel, kill an old one), pivot ICP, or have a major sales-cycle shift. Tracking degradation (cookie deprecation, ad blockers, iOS updates) also forces re-evaluation. your model could drift unnoticed. Actionable signal: when attribution-driven decisions stop matching incrementality test results, the model needs work. Don't change models more than quarterly without a strong reason; constant churn destroys trust in the data.

What's the simplest attribution model for a startup with low data volume?

Run first-touch and last-touch in parallel using UTM parameters and a CRM. First-touch tells you which channels introduce buyers; last-touch tells you which close them. The gap between the two rankings is itself the most useful insight a low-volume startup can get. Implement with: disciplined UTMs on every link, first-touch UTM stored as a CRM contact field, last-touch tracked via standard analytics. Setup: under a week, no specialized tooling. Upgrade to position-based or W-shaped once you exceed 30 deals/month.