# Data-driven vs last-click attribution: what changes and which to use > Switching models does not create or destroy a single sale. It changes who gets the credit, and because automated bidding reads that credit, it changes where your money goes next. Source: https://www.nashmarketinglabs.com/en/blog/data-driven-vs-last-click-attribution Author: Nash Marketing Labs, a paid media agency (Google Ads, Meta Ads, TikTok Ads, LinkedIn Ads). Date: 2026-10-21 --- **Last-click attribution gives 100% of a conversion to the last ad clicked before it. Data-driven attribution splits the same conversion into fractions across every touchpoint, weighting each one by comparing the paths of people who converted with the paths of people who didn't.** Total conversions stay the same; the split changes. **Where things stand in 2026:** in Google Ads only two models are left, data-driven and last click. First click, linear, time decay and position-based were retired, and conversion actions using them were moved to data-driven ([Google Ads Help](https://support.google.com/google-ads/answer/6259715)). GA4 dropped the same four models in November 2023 and keeps data-driven plus two last-click variants ([Analytics Help](https://support.google.com/analytics/answer/10596866)). So "which attribution model should I use?" is now, in practice, this exact choice. It looks like a reporting detail. It is a budget decision: under last click, brand search and retargeting look brilliant because they sit at the end of every path, while whatever creates the demand (video, display, generic search, prospecting) looks expensive and gets cut. ## What each model actually does **Last click.** A fixed, deterministic rule: the whole conversion goes to the last ad click recorded before the purchase. It is transparent, it fits in one sentence and it matches the intuition of "what closed the sale". Its flaw is structural: an ad that opens every path but never closes one can never receive credit. **Data-driven.** No fixed rule. The platform builds a model from your own data, comparing converting paths with paths that did not convert. If paths that include a given campaign or format convert more often than similar paths without it, that touchpoint gets more weight. In Google Ads the touchpoints are ad clicks and video engagements across Search, Shopping, YouTube, Display and Demand Gen ([Google Ads Help](https://support.google.com/google-ads/answer/6394265)), which is why a YouTube campaign can earn credit without being clicked. GA4's data-driven model goes further and splits credit across channels, organic included. The split is specific to your account and to each conversion action, and it is recalculated as your data changes. Two practical consequences follow. First, **data-driven weights are not comparable between accounts**: there is no "standard share for display" you can copy from a benchmark. Second, data-driven needs multi-touch paths to have something to split. In an account where nearly everyone arrives through brand search and buys in the same session, the model has little to work with and its split will look a lot like last click. It is not broken; there is simply no journey to model. ## The two models side by side Last clickData-driven How credit is assigned100% to the final ad clickFractions across all ad touchpoints, weighted by your data Who winsBrand search, retargeting, bottom of funnelProspecting, video, display, generic search gain credit Conversion countsWhole numbers per campaignDecimals per campaign (the account total is unchanged) ExplainabilityOne sentenceA model you cannot fully inspect Status in Google AdsAvailable, chosen manuallyThe default for most conversion actions Best forSingle channel, short same-session paths, accounting reconciliationSeveral campaign types, multi-step paths, automated bidding ## A worked example: same 100 sales, two splits An online store gets 100 Google Ads conversions in a month from three campaigns, and spends €2,000 on YouTube / Demand Gen; average order value is €60. Nothing in the business changes; only the model does: CampaignLast clickData-driven Brand search6041.3 Generic search3036.9 YouTube / Demand Gen1021.8 Total100100 The figures are invented to show the mechanism; the size of the shift varies a lot from account to account, the direction usually does not. Under last click, the video campaign shows 10 sales, €600 of revenue for €2,000 of spend, and looks like the obvious cut. Under data-driven it gets 21.8 sales, about €1,300 of revenue: still below its cost, yet no longer a write-off, and only one of the two models gives it a fair hearing. ## What changes in your reports when you switch - **Brand and retargeting lose conversions.** They sit at the end of the path, exactly where last click gave them 100%. Their cost per conversion gets worse in the report while nothing has changed in reality. - **Upper-funnel campaigns gain.** Conversions appear where there used to be zeros, and campaigns that looked like pure cost get a defensible return. - **The total holds.** If total conversions really move up or down, it is not the model: look at the conversion window, a tagging change or a duplicated conversion action. (In GA4, conversions can be reattributed for up to 7 days, so the last few days can still shift a little.) - **Decimals appear.** "18.42 conversions" is not a rounding error, it is a fractional split. Reconciling that against a backend that counts whole orders will never match exactly, and that is not a tracking bug. - **History behaves differently in each tool.** In Google Ads, a model change only affects how conversions are counted from that point on; to see the past under the new model you add the "Conversions (current model)" columns. In GA4, changing the reporting attribution model applies to historical and future data ([Analytics Help](https://support.google.com/analytics/answer/10597962)). A GA4 export taken before the switch and one taken after will not match even for the same month. Before switching, export what you have and save it with the date, and annotate the change. Most "performance dropped in March" conversations we audit turn out to be a model or window change nobody wrote down. ## The part that matters most: the model steers bidding Attribution is not just a report. Smart Bidding strategies such as Maximize conversions, Target CPA and Target ROAS optimise against the conversions in the "Conversions" column, which are counted with the model you chose. Change the model and you change what the algorithm is chasing. With last click, the system learns that brand and retargeting produce sales and concentrates bids there. With data-driven, it starts seeing return higher up the funnel and bids for it. In our experience, expect a transition of a couple of weeks while spend redistributes on its own, and **never change the attribution model and the bid strategy in the same week**: if you move both, you cannot tell which one produced the result. Measure the effect with something the model cannot touch: total revenue against total ad spend, that is [MER](https://www.nashmarketinglabs.com/en/blog/mer-vs-roas), or new orders from your backend against spend. Compare the same number of weeks before and after, and rule out other changes in that period (prices, promotions, seasonality). If MER holds while the internal split reshuffles, the switch has not hurt the business, even if your brand campaign's [ROAS](https://www.nashmarketinglabs.com/en/blog/roas-vs-roi) has collapsed in the report. Before switching, the Model comparison report in Google Ads shows the two models side by side by campaign, so you can see who will gain and who will lose before you commit. ## What neither model solves Inside the ad platforms, both models are **within one platform**. Google Ads' data-driven model splits credit across Google's touchpoints; it does not know the user saw three Instagram ads first. GA4 sees more channels, but only through visits and clicks: an Instagram ad that was seen and never clicked does not exist for it. Meta has its own system and only sees Meta: it credits its own ads within the [attribution window](https://www.nashmarketinglabs.com/en/blog/meta-ads-attribution-window) you set (7-day click and 1-day view, plus 1-day engage-through since March 2026). Add up ROAS from both platforms and you count the same sale twice; no attribution model fixes that. And neither measures incrementality. A campaign receiving credit does not mean the sale would not have happened without it. That takes holdouts or lift tests, not a different way of splitting credit. ## Which model to use in your account The default recommendation is **data-driven**, and not because it is new: if you use automated bidding across several campaign types, last click is teaching the algorithm a false version of your funnel. The exceptions: - **One campaign type, short paths** (Search only, same-session purchases): the two models will give near-identical splits. Last click is fine and easier to explain and reconcile. - **Low conversion volume.** With few multi-touch paths, data-driven behaves much like last click. Do not expect a big shift. - **Lead generation with qualification later.** Here the problem is not the model but the event. If you optimise to form fills rather than qualified leads, a perfect split of the wrong goal is still the wrong goal. Fix [what you are counting](https://www.nashmarketinglabs.com/en/blog/google-ads-vs-ga4-conversions) before you touch attribution. - **Finance reconciliation.** For matching numbers with accounting, whole-number last click is easier. In Google Ads each conversion action has one model, so do not duplicate the action just to get a last-click view (it inflates All conversions): use the Model comparison report, or reconcile against the backend. The rule we apply: the attribution model decides the internal split; whether total spend is working is judged by MER and [CAC](https://www.nashmarketinglabs.com/en/blog/cac-vs-cpa) against the backend. When those two layers get mixed up, the channel that creates demand gets cut because it does not close the sale. ## Frequently asked questions ### What is data-driven attribution? A model that splits each conversion into fractions across all the ad touchpoints on the path, instead of giving it all to one. Each touchpoint's weight is not a fixed rule: it is calculated from your own data by comparing paths that converted with paths that did not. That is why the split is different in every account and changes over time. ### What is the difference between data-driven and last-click attribution? Last click applies a fixed rule: 100% of the credit to the final ad click before the conversion. Data-driven has no fixed rule and splits credit in fractions according to the weight your data gives each touchpoint. Total conversions are the same under both; what changes is which campaigns they are credited to, which usually means less credit for brand and retargeting and more for prospecting, video and generic search. ### Is data-driven attribution better than last click? For an account with several campaign types, multi-step paths and automated bidding, yes, because last click teaches the algorithm a funnel that does not exist. For a single-channel account with same-session purchases, or with very few conversions, the difference is small and last click wins on transparency and on how easily it reconciles with your backend. ### Which attribution models are still available in Google Ads? Data-driven and last click. First click, linear, time decay and position-based are no longer supported, and conversion actions that used them were moved to data-driven. Data-driven is the default for most conversion actions. GA4 removed the same four models in November 2023 and offers data-driven, paid and organic last click, and Google paid channels last click. ### Why do I see conversions with decimals? Because data-driven splits each conversion into fractions across the touchpoints on the path. A campaign that took part in several paths without always being the last click gets pieces of conversions, and the sum shows decimals. It is not an error: across all campaigns the total adds back up to whole conversions. ### Does changing the attribution model affect bidding? Yes, and that is the most important effect of the switch. Smart Bidding optimises against the conversions in the Conversions column, which are counted with the chosen model, so changing the model changes what the algorithm is chasing and spend redistributes over a week or two. Do not change the model and the bid strategy at the same time, or you will not know which change caused the result. ### Does changing the model rewrite my historical data? It depends on the tool. In Google Ads, the change only affects conversions counted from then on; you can see past data under the new model with the 'Conversions (current model)' columns. In GA4, changing the reporting attribution model applies to historical and future data, so exports taken before and after the switch will not match. Export and date your data before changing anything. ### Does Meta use data-driven attribution? Not in the Google Ads sense: there is no model to choose. Meta credits its own ads for conversions that happen within the attribution window you set, by default 7-day click and 1-day view, plus 1-day engage-through since March 2026, and the window is what decides how many conversions you see. Neither system sees the other: Google's model only splits credit among Google campaigns and Meta's among its own, so adding up both ROAS figures counts sales twice. ## More resources - [Meta Ads attribution window: 7-day click vs 1-day click](https://www.nashmarketinglabs.com/en/blog/meta-ads-attribution-window) - [MER vs ROAS: which metric to use and when](https://www.nashmarketinglabs.com/en/blog/mer-vs-roas) - [CAC vs CPA: the difference and which one to track](https://www.nashmarketinglabs.com/en/blog/cac-vs-cpa) - [Why Google Ads and GA4 conversions don't match](https://www.nashmarketinglabs.com/en/blog/google-ads-vs-ga4-conversions) - [ROAS vs ROI: what's the difference and when to use each](https://www.nashmarketinglabs.com/en/blog/roas-vs-roi) --- Contact: https://www.nashmarketinglabs.com/en/request-audit - Your performance plan: concrete actions to improve your account, no commitment.