Attribution modeling is the set of rules that decides which ads, channels or touchpoints get credit for a sale. Last click gives all the credit to the final touch, first click to the first one, and data-driven models share it based on observed paths. No model proves which ad caused the sale, so use attribution to compare and spot patterns, and check big budget moves with a test.
What is attribution modeling?
Before an order, a buyer often meets your brand several times: a Facebook video, a product page visit, an email, a search for your brand name. Each of these is a touchpoint, and the full list is the conversion path. An attribution model is the rule that splits the credit for the sale between those touchpoints.
This matters because credit drives budget. If your reports give every sale to the last touch, the channels that close (brand search, retargeting, email) look great and the ads that first bring people in look weak. Cut those ads and, a few weeks later, fewer people reach the closing channels at all.
Keep one idea in mind for the rest of this guide: attribution is not causation. Attribution shows who was in the path. Causation asks what changed the buyer's decision. A retargeting ad shown to people who were already about to buy will collect a lot of credit without creating many new sales.
The main attribution models compared
| Model | How it gives credit | It favours | Useful question |
|---|---|---|---|
| Last click | 100% to the final click | Brand search, email, retargeting | What closes demand that already exists? |
| First click | 100% to the first touch | Discovery ads, influencers | What brings new people in? |
| Linear | Equal share to every touch | Long paths | What is in the path at all? |
| Time decay | More credit to touches near the sale | Late touches | What matters in a short buying cycle? |
| Position-based (U-shaped) | Most to first and last, the rest to the middle | Openers and closers | Who opens and who closes? |
| Data-driven | Shares based on paths that did and didn't convert | Touches common in converting paths | What does the pattern in the data suggest? |
The rule-based models are easy to explain, which is why many tools still show them. But they apply the same fixed rule to every sale, and each one has a built-in bias.
What Google offers today
Google cut its list down. Google Ads now offers two models, data-driven and last click; first click, linear, time decay and position-based are no longer supported. Google announced the change on 6 April 2023 for Google Ads and Google Analytics 4, stopped new selections of those models in June 2023, and moved conversion actions that still used them to data-driven attribution in September 2023.
So if an older guide tells you to "switch to time decay in GA4", that option no longer exists there. You can still apply the rule-based models in a spreadsheet or in other analytics tools, as a way to look at the same paths from another angle.
How does Meta attribute sales?
Meta doesn't use a list of models like the one above. In Ads Manager, each ad set has an attribution setting: the time after someone clicks or views your ad during which a purchase is credited to that ad. Meta's help page lists options such as 1-day click, 7-day click and 7-day click or 1-day view, and the options available depend on the campaign. A setting that includes views credits more sales to the same ads than a click-only setting.
Two recent changes matter:
- Link clicks only. In March 2026 Meta announced that click-through attribution for website and in-store conversions would count only link clicks. Sales after a share, a save or another non-link click move to engaged-view attribution, which Meta is renaming "engage-through". Meta also lowered the engaged view for video ads from 10 seconds to 5.
- Incremental attribution. Meta offers an incremental attribution model in Ads Manager that aims to count only the conversions that would not have happened without the ad. In January 2026 Meta said its latest model update raised incremental conversions by 24% compared with its standard attribution model. That is Meta's own figure about its own product.
Remember that Meta only sees Meta. It credits sales to its own ads under its own rules, and Google, your email tool and your affiliate network each do the same. Add up what every platform claims and you will often get more sales than your store actually made.
Why one model is never enough
Every model answers one question well and others badly. Last click is good at showing what closes and bad at showing what creates demand. First click is the opposite. Data-driven models are more flexible, but they still learn from who was present in converting paths, so ads aimed at people who were already likely to buy can get credit they didn't earn.
Tracking gaps make it worse. People switch between phone and laptop, block cookies or refuse tracking, so part of every path is missing. The touches that are easiest to see, like a final click on a branded search, end up with more credit than the touches that are hard to see, like a video watched on a phone a week earlier.
That's why the useful approach is to compare views and then test, not to pick one "true" model.
A practical attribution setup for an online store, step by step
- Make your store the reference. Your store's order count and revenue are the totals that every report has to fit into. If Meta, Google and email together claim more sales than you made, the overlap is the double counting.
- Track a blended number. Each month, divide total revenue by total ad spend across all channels. It ignores attribution completely, so it shows whether more spending really brings more sales.
- Keep one fixed setting per platform. Choose a Meta attribution setting that fits how fast your customers buy, and keep it while you compare periods and ad sets. Clean tracking comes first; our guide to tracking Facebook ad performance covers the Pixel, the Conversions API and UTMs.
- Compare two views. Look at last click and one multi-touch or data-driven view of the same period. Channels that look strong in first-touch views and weak in last-click ones are probably bringing people in. Channels that win only on last click are probably closing.
- Ask your buyers. A one-question survey after checkout ("Where did you first hear about us?") catches discovery channels that tracking misses, like a video someone watched on another device.
- Test before big moves. Before you cut or double a channel, run a holdout or lift test: turn the ads off (or keep them off) for a random group or a set of regions and compare total sales. Our guide to incrementality testing explains how.
Give each creative a job
Attribution also helps creative planning. Ads that keep showing up as first touch are doing the introducing: they need to stop the scroll and explain the problem. Ads that win last click are doing the closing: proof, offer, a reason to buy now. Judge each creative against its own job, not against one ROAS number for the whole account, and test angles inside each role.
How to do this with SearchTheTrend
Attribution tells you which of your own ads get credit. SearchTheTrend can't see anyone's conversions, but it shows which creatives other stores keep paying for, which helps you plan tests for each job.
- In Ads, search your niche and filter by format: UGC, talking head, demo, unboxing, studio or meme. Sort by days running or Ad Rank to see what brands keep live, and note which ones introduce the product and which ones close with proof or an offer.
- Open a competitor's store to see its mix of formats and which ads it spends the most on in Europe.
- Save the angles you want to test to a board, one board per funnel stage.
More on this workflow for media buyers. The free plan needs no card.
FAQ
What is attribution modeling in simple words?
It is the rule that decides which ads or channels get credit for a sale when a buyer saw or clicked several of them first. Different rules give the same sale to different channels.
Which attribution models does Google still offer?
Google Ads offers data-driven and last click. Google announced in April 2023 that first click, linear, time decay and position-based models would leave Google Ads and Google Analytics 4, and conversion actions that still used them were moved to data-driven attribution.
How does Meta attribute sales to Facebook and Instagram ads?
Each ad set has an attribution setting, a window after a click or a view in which a sale is credited to the ad. Meta also offers an incremental attribution model that aims to count only the sales that would not have happened without the ad.
Is data-driven attribution the truth?
No. It is a better estimate of each touchpoint's share, but it still describes who was in the path, not who caused the sale. Before a big budget change, check the result with a holdout or lift test.
Why do Meta, Google and my store report different sales?
Each platform credits the sales it can see under its own rules, so one order can be claimed by Meta, Google and email at the same time. Your store counts each order once, so use its total as the reference.



