Strategy & Tracking
Marketing Attribution Models, Explained Without the Jargon
An attribution model decides which channel gets credit when a homeowner books a remodel. Pick the wrong one and you defund the campaigns that actually fill your calendar. Here is each model in plain English, and how to choose.
An attribution model is the rule that decides which marketing touch gets credit when a homeowner signs a remodel. That rule controls your budget. If your model hands all the credit to the last click, you will pour money into branded search and retargeting while starving the ads that introduced the homeowner in the first place. Below, every common model in one sentence, when each one fits, and how to set up tracking that survives the loss of cookies.
Why the model you pick decides your budget
Most homeowners touch several channels before they convert. They see a paid social ad for a kitchen remodel, read a blog post weeks later, click a branded search result, then request an estimate. Every one of those touches helped. Your attribution model decides how much credit each gets, and that split becomes the report your budget follows.
Here is the trap. The default last-click view credits the final touch and ignores everything before it. So branded search and retargeting look like heroes, while the awareness channel that started the journey looks like a money pit. Cut it, and conversions fall for reasons your report can't explain. The model didn't just measure the result. It shaped the decision.
The models, defined in one line each
Two families exist. Single-touch models give 100% of the credit to one touch. Multi-touch models split credit across several. Data-driven sits on its own, because an algorithm sets the weights instead of a person.
Every common model, one sentence apiece
- First-touch: all credit to the first interaction. Use it to see which channels open the door, not which close.
- Last-touch: all credit to the final interaction. Easy to compute, easy to defend, and quietly punishes every upper-funnel channel.
- Last non-direct: all credit to the last touch that wasn't a direct visit, so a marketing channel gets the credit instead of someone typing your URL.
- Linear: equal credit to every touch. Good for seeing the full journey, bad for setting budget, because no touch is really equal.
- Time-decay: more credit to touches closer to the sale. Fits short, transactional cycles where the last few days do the work.
- Position-based (U-shaped): 40% to the first touch, 40% to the last, 20% split across the middle. A sensible default when discovery and the close both matter.
- W-shaped: 30% each to the first touch, the lead-creation touch, and the opportunity-creation touch. Built for B2B journeys with clear sales milestones.
- Data-driven (DDA): an algorithm assigns credit by comparing paths that converted against paths that didn't. The most accurate, when you have enough data to feed it.
Which model fits your business
Match the model to your sales cycle, not to what your dashboard shows by default. A short cycle and a long one need different rules.
If you sell something people buy in one or two sessions, time-decay or last non-direct is honest enough. If your homeowners take weeks of research before requesting an estimate and cross several channels, a single-touch model will lie to you. Reach for position-based or, better, data-driven.
B2B is the clearest case. Closing a deal can take dozens of touches across months, and the buying committee rarely follows a straight line. Position-based or W-shaped at least credits the milestones that move a deal forward. Last-click will tell you sales demos close revenue and content does nothing, which is both true and useless for planning.
The model didn't just measure the result. It shaped the decision.
GA4 changed the defaults, and last-click lost
Google forced the issue. In October 2023 it removed four rules-based models from Google Ads and GA4: first-click, linear, time-decay, and position-based. The reason was blunt. Fewer than 3% of Google Ads conversions still used them. GA4 and Google Ads now default to data-driven attribution, with last-click left as the main alternative.
So inside Google's tools, the menu is shorter than the list above. You choose between data-driven, which spreads credit using machine learning, and last-click, which doesn't. For most accounts with reasonable conversion volume, data-driven is the better default. It needs a steady stream of conversions to learn from, so very low-volume accounts may see noisy results.
The privacy reality: where MMM and incrementality come in
Multi-touch attribution rests on following one person across sites and sessions. That foundation cracked. After Apple's App Tracking Transparency and the decline of third-party cookies, multi-touch coverage now sits at roughly 30 to 60% of its 2020 level. Whole stretches of the journey are now invisible, and the gaps don't fill themselves.
Two older methods came back to cover the blind spots. Marketing mix modeling (MMM) ignores individual users entirely. It uses aggregate spend and sales data to estimate what each channel contributed, which makes it immune to cookie loss. Adoption reflects the shift: 49% of marketers worldwide now use it. Incrementality testing answers a sharper question by holding out a group from seeing your ads, then measuring the lift among those who did. That is the closest thing to proof that a channel caused a sale rather than rode along next to it.
The strongest setup triangulates. Use platform attribution for day-to-day tactical signal, MMM for the portfolio-level view, and incrementality tests to settle the arguments attribution alone can't.
How to actually set it up
Pick a model second. Fix your tracking first, because a perfect model on dirty data still lies.
Start with clean conversion tracking in GA4. Define the conversions that matter, a purchase, a qualified lead, a booked call, and make sure each fires once and only once. Deduplicate events so a single sale isn't counted across three tags. Add server-side tagging to hold onto data that browser-based tracking now drops, and route consent through a consent banner so you're collecting what you're allowed to.
Then connect your ad platforms and your CRM so the journey doesn't end at the form fill. For longer sales cycles, that CRM link is the difference between crediting a lead and crediting actual revenue. Once the data is trustworthy, choose your model deliberately, and revisit it when your channel mix changes. WellBuilt builds attribution and tracking systems that hold up after the cookies are gone, so the numbers you report match the money you make.
Key takeaways
- Your attribution model controls your budget. The wrong one defunds the channels that actually start sales.
- Match the model to your sales cycle: time-decay for short cycles, position-based or W-shaped for long B2B journeys.
- GA4 and Google Ads removed four rules-based models in 2023 and now default to data-driven, with last-click as the alternative.
- Cookie loss cut multi-touch coverage to 30-60% of 2020 levels, which is why marketing mix modeling and incrementality testing are back.
- Fix conversion tracking, deduplication, server-side tagging, and CRM connection before you obsess over the model.
SourcesSearch Engine Land, Google confirms sunset details for 4 attribution models in Ads and Analytics (2023) · Google via Search Engine Land, fewer than 3% of Google Ads conversions used rules-based models (2023) · Improvado, Multi-Touch Attribution Implementation Guide, MTA coverage at 30-60% of 2020 levels (2024) · Supermetrics, 49% of marketers worldwide use marketing mix modeling (September 2024) · IAB, 56% of US ad buyers to focus more on MMM in 2025 (December 2024)
Questions, answered straight.
What is the best attribution model?
There isn't one best model, only the best fit for your sales cycle and data volume. For most accounts with steady conversions, data-driven attribution is the strongest default. For long B2B cycles, position-based or W-shaped gives a fairer view. The honest answer for most businesses is to combine platform attribution with marketing mix modeling and the occasional incrementality test.
Why is last-click attribution a problem?
Last-click hands all credit to the final touch and zero to everything before it. That makes branded search and retargeting look great while the awareness channels that introduced the customer look worthless. Follow that report and you cut the top of your funnel, then watch conversions fall for reasons the report can't show you.
Did GA4 really get rid of attribution models?
Yes. In October 2023 Google removed first-click, linear, time-decay, and position-based attribution from Google Ads and GA4, citing adoption below 3% of conversions. GA4 now defaults to data-driven attribution and keeps last-click as the main alternative.
Does attribution still work without third-party cookies?
Partially. Multi-touch attribution now covers only 30 to 60% of what it tracked in 2020, so it can't see the full journey anymore. To cover the gaps, marketers pair it with marketing mix modeling, which uses aggregate data and needs no cookies, and incrementality tests, which prove a channel's lift directly.
Strategy & Tracking
Want this run for you, not just read about?
Clean tracking and honest attribution, so you know which dollars actually produce revenue.