Attribution Models: Which One to Choose and Why
A prospect reads your blog post in March, clicks a LinkedIn ad in April, searches your brand name on Google in May, and books a demo in June. Three months, four touchpoints, one deal. So which channel gets the credit?
That question is what attribution models answer, and the answer changes where you spend money. Pick the wrong model and you will defund the channel that actually starts deals while overpaying for the one that happens to be last in line. In B2B, where buying cycles run weeks or months and five or more touchpoints are normal, this is not a rounding error. It decides budgets.
This guide walks through the main attribution models, what each one rewards, and how to choose based on your sales cycle and the data you can actually collect. No abstract theory. By the end you should know which model to set in GA4 and how to sanity-check what it tells you.
What an attribution model really does
An attribution model is a rule for splitting credit for a conversion across the touchpoints that led to it. Nothing more. It does not measure causation. It assigns credit according to a logic you choose, and different logics produce very different reports from the exact same data.
Two things make this hard in B2B. First, the path is long and multi-channel, so there is a lot of credit to split. Second, the conversion you care about (a closed deal) happens far downstream from the click, often in a CRM your ad platform never sees. Connecting those two ends is the real work, and the model sits on top of it.
Before the model matters, your tracking has to be solid. If your touchpoints are mislabeled or missing, every model lies with confidence. Get conversion tracking for B2B right first, then argue about attribution.
The single-touch models
These give 100% of the credit to one touchpoint. Simple to read, easy to misuse.
Last-click attribution
The final touchpoint before conversion takes all the credit. For years this was the default in most analytics tools, and it is still the model most people picture when they think "where did this lead come from."
Last-click flatters the bottom of the funnel. Brand searches, retargeting, direct visits: anything that closes the loop looks like a hero. The content, the cold ad, the webinar that introduced you months earlier gets nothing. If you optimize purely on last-click, you slowly starve the top of your funnel and then wonder why your pipeline dried up two quarters later.
It is not useless. For a short, transactional purchase with one or two touches, last-click is roughly honest. For B2B with a long cycle, it is a trap.
First-click attribution
The mirror image: the first touchpoint takes everything. This rewards discovery channels (organic content, top-of-funnel ads, that first webinar) and ignores everything that happened during consideration and decision.
First-click answers one useful question: what introduces people to us? It answers nothing about what closes them. Use it as a counterweight to last-click, not as your primary lens.
The multi-touch models
Multi-touch models spread credit across several touchpoints. This matches B2B reality far better, because most deals genuinely involve many.
Linear
Every touchpoint gets equal credit. Five touches, 20% each. Linear is the "everyone gets a trophy" model. Its strength is that it stops ignoring the middle of the funnel. Its weakness is that it pretends the throwaway third visit mattered as much as the demo request, which it almost never did.
Linear works as a gentle starting point when you are moving off single-touch and want to see the full path without imposing a strong opinion about which touches matter most.
Time-decay
Touchpoints closer to the conversion get more credit; earlier ones get less, on a decaying curve. The logic: recency correlates with intent, so the touch a day before the deal probably mattered more than one from six weeks ago.
Time-decay suits longer sales cycles where the late stages (demo, proposal, negotiation) are where channels earn their keep. The risk is the same as last-click in slow motion: it under-credits the awareness work that filled the top of the funnel. If your content engine is what generates demand, time-decay will quietly undervalue it.
Position-based (U-shaped and W-shaped)
Position-based models hand the biggest shares to the touchpoints that matter most structurally. The common version, U-shaped, gives 40% to the first touch, 40% to the last touch, and splits the remaining 20% across everything in the middle.
The thinking is that two moments do the heavy lifting: the one that found the prospect and the one that converted them. W-shaped adds a third anchor, the lead-creation or opportunity-creation point in the middle, splitting credit roughly 30/30/30 with 10% for the rest. W-shaped is popular in B2B precisely because it respects the moment a lead becomes a real opportunity, which other models ignore.
Here is how the same four-touchpoint journey looks under several models. Numbers are illustrative.
| Touchpoint | Last-click | Linear | Time-decay | U-shaped |
|---|---|---|---|---|
| 1. Organic blog post | 0% | 25% | 10% | 40% |
| 2. LinkedIn ad | 0% | 25% | 20% | 10% |
| 3. Email click | 0% | 25% | 30% | 10% |
| 4. Brand search → demo | 100% | 25% | 40% | 40% |
Notice how the blog post swings from 0% credit to 40% depending only on the rule. Same journey. The channel did not change. The accounting did.
Data-driven attribution
Data-driven attribution (DDA) drops the fixed rules. Instead it uses your actual conversion data to estimate how much each touchpoint contributed, comparing the paths of people who converted against those who did not. GA4 now uses DDA as its default, and Google Ads relies on it heavily for bidding.
When it works, DDA is the most honest model on this list, because the weights come from your data rather than from a guess about funnel shape. The catch is that it needs volume. With a handful of conversions a month, the algorithm has nothing to learn from, and you are better off with a rules-based model you understand. It is also a black box: you get weights, not a clear explanation, which makes it harder to defend a budget cut to a skeptical CFO.
The data-driven funnel, visualized
The shares above are illustrative. The point is that data-driven attribution does not assume the last touch wins or the middle is filler. It reads the patterns and assigns weight accordingly.
How to actually choose
Stop looking for the "correct" model. There isn't one. There is a model that fits your sales cycle, your data volume, and the decision you are trying to make. Work through these.
Match the model to your cycle length. Short cycle, one or two touches, transactional buy: last-click is fine and simple. Long cycle, many touches, considered purchase: you need a multi-touch model. W-shaped or data-driven for most B2B.
Match it to your data volume. DDA needs steady conversion volume to learn. A rough rule from Google's own guidance is that you want a healthy number of conversions per month before DDA produces stable results. Below that, use a rules-based model (W-shaped or time-decay) you can explain.
Match it to the decision. Allocating top-of-funnel budget? Look at first-touch and linear to see what starts journeys. Optimizing close rate and bidding? Lean on data-driven or time-decay. Run more than one model side by side; the gap between them is often more informative than any single number.
Pick one model as your source of truth, then keep a second for context. Switching models every month makes trends meaningless. Most B2B teams land on data-driven as the primary lens once volume allows, with last-click kept around because partners and platforms still speak it.
The deeper problem is that all of this lives in your ad platform until the deal closes in your CRM. To see which touchpoints produce revenue and not just form fills, you have to feed deal data back. That is what importing offline conversions into your ads does, and it is what turns attribution from a vanity report into a budgeting tool. Pair it with a clear method for attributing revenue to channels and you can finally argue about money instead of clicks.
Common mistakes that break attribution
Mislabeled traffic ruins everything upstream of the model. If half your paid clicks land in "direct" because the URLs were not tagged, no model can fix that. Consistent UTM tagging is the unglamorous foundation.
Judging brand search by last-click is the classic error. Brand search almost always wins last-click because people search your name when they are ready to buy. That does not mean brand search created the demand. It captured demand your other channels created.
Comparing across models without saying which one. A report that shows "Channel X drove 40 conversions" is meaningless without the model behind it. Always label it.
Chasing form fills instead of deals. A channel can look brilliant at generating leads and terrible at generating customers. Without connecting attribution to closed revenue, you optimize toward the wrong end. See how to measure PPC performance by revenue rather than by clicks.
Frequently asked questions
What is the best attribution model for B2B?
For most B2B companies with multi-month cycles, data-driven attribution is the strongest choice once you have enough conversion volume for it to learn. Below that threshold, a W-shaped or time-decay model gives you a defensible, multi-touch view without needing a large data set.
Is last-click attribution dead?
No, and it still has uses. It is honest for short, transactional purchases and it is the language most ad platforms and partners default to. The mistake is using it as your only model for a long B2B sales cycle, where it systematically under-credits everything except the final touch.
What attribution model does GA4 use?
GA4 uses data-driven attribution as its default for reporting and conversions. It can still show you other lenses, but rules-based models like first-click and last-click have been deprecated as choices in parts of the Google stack, so the platform is steering everyone toward DDA. Worth checking the current settings when you set up your property.
How many touchpoints does a typical B2B deal have?
It varies widely by deal size and industry, so treat any single figure with caution. The reliable point is directional: considered B2B purchases involve multiple people and multiple touches over weeks or months, which is exactly why single-touch attribution misleads. Look at your own CRM data for your real number rather than trusting a benchmark.
Do I need a CRM to do attribution properly?
To attribute leads, no; your analytics tool can do that. To attribute revenue, effectively yes. The deal closes in your CRM, often long after the click, so you need to connect CRM outcomes back to the original source. Without that link, you are optimizing toward form fills, not customers.
What is the difference between attribution and tracking?
Tracking is collecting the touchpoint data: which ad, which page, which click, properly tagged. Attribution is the rule that splits credit across those touchpoints. Tracking comes first. A perfect model on broken tracking still gives you garbage, which is why getting GA4 set up correctly matters before you fuss over models.
In short
Attribution is a budgeting decision dressed up as an analytics one. The model you pick determines which channels look worth funding, so choose deliberately:
- Short, transactional cycle: last-click is fine.
- Long B2B cycle, low data volume: W-shaped or time-decay.
- Long cycle, healthy conversion volume: data-driven.
- Always connect attribution to closed revenue, not just form fills.
- Label every report with the model behind it, and tag your traffic consistently.
The hardest part is rarely the model. It is wiring touchpoints to deals so you can see what actually pays back. If your reports stop at clicks and you cannot tell which channels produce revenue, that is the gap worth closing. We help B2B teams build closed-loop analytics that ties ad spend to pipeline and deals; if you want a second set of eyes on your setup, get a short audit of your attribution and tracking and we will tell you where the credit is leaking.