Just one in five marketers say they're confident their last-click attribution model actually reflects what's driving revenue, according to a July 2024 eMarketer survey of 282 senior US marketers. Despite that, the same survey found 78.4% of them still use it, and almost three-quarters say they're either moving away from it or wish they could.
That gap between what marketers use and what marketers trust is the real starting point for choosing a marketing attribution model. Most guides skip straight to a taxonomy lesson: first-touch, last-touch, linear, time-decay, position-based, data-driven. Fewer explain that the platform most of them run that taxonomy through already took half those options off the table.
TL;DR:
- Google removed first-click, linear, time-decay, and position-based attribution from Google Analytics 4 in November 2023. You can no longer manually pick those models inside GA4.
- GA4's data-driven model and Google Ads' data-driven model run on different data and will disagree, even though both call themselves "data-driven."
- The attribution model matters less than whether your conversion events, lookback windows, and CRM lead-source data are set up correctly feeding into it.
- Match the model to your sales cycle: short and single-session favors the default, long and multi-channel needs a CRM-level reconciliation on top of GA4.
What a marketing attribution model actually decides
A marketing attribution model is just a rule for splitting credit across touchpoints when a customer converts. Single-touch models (first-touch, last-touch) give 100% of the credit to one interaction. Multi-touch models spread credit across several touchpoints in the path, weighting the early, middle, and late interactions differently depending on the model.
That part of the story hasn't changed, and almost every roundup covers it well enough. What most of them skip is that the tool running underneath these decisions for most mid-market businesses, Google Analytics 4, no longer lets you pick most of them.
What actually changed inside GA4
As of November 2023, GA4 dropped first-click, linear, time-decay, and position-based attribution entirely, according to Google's own attribution documentation. The only options left are data-driven attribution, which is the default for every GA4 property, and a couple of rules-based last-click variants.
Data-driven attribution works by comparing converting and non-converting paths and distributing fractional credit based on which touchpoints actually correlated with a conversion, not just which one happened first or last. It runs on a lookback window that Google sets at 30 days for acquisition events like first visit, and 90 days by default for other key events, configurable down to 60 or 30 days.
Why your Google Ads numbers never match your GA4 numbers
Google Ads runs its own separate data-driven attribution model, and it is not the same calculation as GA4's. Per Google Ads' own documentation, the Ads model only evaluates ad clicks and video engagements across Search, Shopping, YouTube, Display, and Demand Gen, and it performs best with at least 200 conversions and 2,000 ad interactions within a rolling 30-day window.
GA4's model, by contrast, evaluates every channel it can see on your site, not just paid ads. Two models, two different input datasets, both labeled "data-driven," reporting two different numbers for the same campaign. That mismatch is exactly the kind of thing our analytics team gets asked to untangle: not "which model is right," but "why don't these two tools, which both claim the same methodology, agree."
The multi-touch models, briefly
Most listicles spend 1,500 words on this part, so here is the short version for context before the part that actually matters.
Linear attribution splits credit evenly across every touchpoint in the path. It's simple to explain to a client, but it treats a single branded-search click the same as the campaign that first introduced the brand, which flattens out the difference between awareness and closing work.
Time-decay attribution gives more credit to touchpoints closer to the conversion and less to earlier ones. It fits sales cycles where recency genuinely matters, like a retargeting-heavy ecommerce funnel, and fits poorly where a long nurture sequence did most of the real persuading early on.
Position-based (U-shaped) attribution weights the first and last touchpoints more heavily than the middle, on the logic that the discovery moment and the closing moment both matter more than whatever happened in between. It's the model most often requested by marketers who want to reward both top-of-funnel and bottom-of-funnel spend.
None of these three are selectable inside GA4 anymore, per the documentation cited above. They still matter if your CRM, your ad platform, or a dedicated attribution tool builds its own model on top of your CRM data, which is increasingly where this decision has actually moved.
The decision that matters more than the model name
Picking a model name is the easy 10% of this. The actual decision is whether the data feeding that model is trustworthy in the first place:
- Are your conversion events firing once per genuine conversion, not once per page reload?
- Is your lookback window long enough to cover your real sales cycle, not just GA4's default?
- Does a lead that closes three weeks after a form fill still get connected back to the original session, or does it just show up as a disconnected new contact in your CRM?
That last question is where attribution usually breaks for service businesses and agencies. If your leads flow through a CRM like the ones we build on GoHighLevel automation, the attribution that actually matters for sales conversations lives at the contact and pipeline-stage level in the CRM, not just the session level in GA4. A lead can touch five marketing channels over six weeks before ever filling out a form, and GA4 will never see most of that if it isn't tied back to the same contact record.
How to pick a model for your actual business
Start with your sales cycle length, not the model name:
- Sales cycle under a few days, mostly single-session conversions: GA4's default data-driven model is fine as-is. Don't fight the platform for a model picklist that no longer exists.
- Sales cycle spans weeks and multiple channels, B2B or high-ticket services: GA4 alone will undercount the early-funnel touchpoints. Reconcile GA4's session-level view against your CRM's lead-source field, set at first contact and never overwritten, so the full multi-week journey survives past the first form fill.
- Running paid ads on more than one platform: Expect each platform's own attribution to disagree with GA4 and with each other. Pick one system of record, usually the CRM, for the number that goes in front of leadership, and treat the ad platforms' self-reported numbers as directional, not final.
The rule
If your sales cycle closes inside a single session, stop shopping for a better attribution model and fix your conversion events instead. If it doesn't, the model name stops mattering the moment your CRM and your analytics platform stop agreeing on who the lead actually is.
Tell our analytics team how your leads actually move from first touch to closed deal, and we'll audit whether your current attribution setup, GA4, your ad platforms, your CRM, is actually built to answer that question before you spend another dollar debating which model name to pick.
