Picture a store spending $60,000 a month across Google Ads, Meta, and email, split evenly. Under last-click, whichever channel a shopper touched right before checkout gets full credit for the sale, so a $20 Meta retargeting ad that closed a journey a Google Search ad and an email actually started looks like the hero. Under data-driven attribution, GA4 would likely spread that same sale's credit across all three touchpoints, weighted by how much each one measurably moved the shopper toward buying. Same store, same spend, same sales, two completely different answers about where next month's budget should go. That gap is not a rounding error. It is the entire reason an ecommerce attribution model choice matters more than most stores treat it.
What an attribution model is actually deciding
An attribution model is not a reporting preference. It is the input to every budget decision downstream of it: which channel gets scaled, which gets cut, and which creative gets called a winner. Get the model wrong for your situation and you are optimizing toward a number that does not reflect what is actually driving sales, which is a slower, quieter version of the same mistake as trusting an untested "winning" A/B variant.
Ecommerce stores feel this more acutely than most businesses because the typical path to purchase touches three to five channels: a paid ad for discovery, an email or SMS nudge, a retargeting ad, sometimes an influencer link, before the final click. Whichever model you run decides how much credit each of those gets, and that credit is what your bidding, budget splits, and "which channel is working" conversations are actually built on.
The models, in plain terms
Attribution theory describes six classic approaches, and it is worth knowing all six because you will still see them referenced in older guides, in third-party attribution tools, and in conversations with anyone who learned attribution before 2023:
- Last-click gives 100 percent of the credit to the final channel clicked before the sale. Simple, but it systematically overweights whatever channel tends to close, usually branded search or retargeting.
- First-click gives full credit to whatever introduced the shopper, useful for measuring top-of-funnel discovery but blind to everything that happened after.
- Linear splits credit equally across every touchpoint in the path. Fair in theory, but it treats a passing impression the same as a decisive click.
- Position-based (U-shaped) weights the first and last touch heavily and splits the rest, a compromise between first-click and last-click.
- Time-decay gives more credit to touchpoints closer to the sale, less to earlier ones.
- Data-driven attribution uses your own account's conversion and non-conversion paths to calculate actual contribution per touchpoint, rather than applying a fixed rule to every path. Google's own documentation describes this as the model that "distributes credit for the key event based on data for each key event," specific to your account and your conversion type, and it is what GA4 recommends by default.
That six-model menu no longer exists inside Google's own tools, though. GA4's attribution documentation confirms that first-click, linear, time-decay, and position-based were retired in November 2023, and Google Ads dropped the same four models from its own attribution settings around the same time. If you open GA4 today expecting to pick from all six, you will not find four of them. That consolidation is the real story: Google decided the rules-based middle ground was not worth the added complexity once data-driven attribution existed as an option.
What GA4 actually gives you by default
GA4 ships with data-driven attribution as the default reporting model, a meaningful change from the last-click default that Universal Analytics ran for over a decade. What is actually left to choose between, now that the rules-based models are gone, is narrower than most guides admit: GA4's attribution settings let you pick a cross-channel data-driven model, a cross-channel last-click model, or a Google-paid-channels-only last-click view, along with the lookback window each model uses to decide how far back to search for touchpoints.
One detail that catches teams off guard: changing the reporting attribution model in GA4 applies retroactively, rewriting historical conversion counts by channel, not just future data. If your dashboards or exported spreadsheets are built on the old model's numbers, switching means re-pulling everything, not just watching new numbers roll in differently.
Why the platforms will never agree, no matter which model you pick
Even a perfectly configured data-driven model will not make Google Ads, Meta, and GA4 report the same conversion count, because each platform only attributes what it can see, on its own lookback window, using its own model. Summing platform-reported conversions and comparing that total to actual store revenue is a common way this shows up: in a delivered tracking audit, Anlyto found summed platform-attributed revenue exceeding actual store revenue by roughly 40 percent in a single reconciliation window, with the same order IDs appearing in more than one platform's export. Full findings from audits like that one are in the sample tracking audit report.
Third-party attribution tools do not resolve this either. Two of the better-known ecommerce attribution platforms, Wicked Reports and Northbeam, have been observed reporting roughly a two-to-one gap in new-customer conversion volume for the same underlying traffic, because each applies its own modeling assumptions on top of the same raw data. If you are weighing a dedicated attribution tool against GA4's native model, our comparison of Triple Whale and GA4 covers what a third-party layer actually buys you and where it does not.
Before blaming any model for a discrepancy, check the plumbing underneath it. A recurring pattern in delivered audits is channel fragmentation at the tracking level, not the modeling level: one store had roughly 31 percent of its traffic landing as direct or none in GA4 while ad platforms recorded the matching click volume, caused by three different UTM spellings for the same channel (fb, facebook, meta-paid) plus untagged email and influencer links. No attribution model, however sophisticated, can correctly credit a touchpoint that never got tagged as a touchpoint in the first place. Our GA4 audit checklist walks through the tagging and event checks worth running before you touch attribution settings at all.
When last-click is still fine
Last-click is not automatically wrong. It is a reasonable default when the sales path is genuinely short, when one channel dominates the media mix, or when nobody on the team has the time to interpret a more nuanced credit split and act on it. A store running one primary paid channel with a same-session checkout does not have much of a multi-touch story to tell, and a simpler model is easier to explain to a founder or a finance team that needs to trust the number quickly.
When to graduate, and the number that decides it
The honest trigger for switching off last-click is conversion volume, not preference. Google's own guidance for data-driven attribution in Google Ads recommends at least 200 conversions and 2,000 ad interactions within a 30-day period before the model has enough signal to identify real patterns rather than noise. Below that volume, a data-driven model still runs, but the credit it assigns can shift meaningfully week to week for reasons that have nothing to do with actual channel performance, which is arguably worse than a simple, consistent last-click view.
If your store is clearing that volume monthly, has more than one paid channel in the mix, and someone is actually going to read the attribution report and act on it, that is the point where switching to data-driven attribution starts paying for itself. Below it, spend the effort on cleaning up tagging instead. It moves the needle further.
How to switch without wrecking your reporting
Validate the tracking underneath the model first. Confirm the core ecommerce events, view_item, add_to_cart, begin_checkout, purchase, are all firing once and only once, and that UTM tagging is consistent across every channel before you touch the attribution setting. An analytics audit is the fastest way to confirm that baseline if you have not had one recently.
Then change the reporting model deliberately, document the date you switched, and expect historical channel-level numbers in GA4 to shift retroactively once you do. Re-pull any dashboard or spreadsheet built on the prior model rather than assuming it updates itself, and give the new model at least one full 30-day cycle before making a budget call based on it. A model change is itself a change in what "working" means, and it deserves the same skepticism as any other test result until the new numbers have had time to settle. Our marketing analytics service sets up and documents GA4 data-driven attribution against the older single-touch models specifically so the switch does not quietly break every report built before it.
The 200-conversion, 2,000-interaction threshold is worth writing down somewhere your team will actually see it. It is the single number that separates "this model change will sharpen our budget decisions" from "this model change will just add noise we mistake for insight."
