Attribution decides which marketing touchpoint gets credit for a conversion. Last click gives all of it to the final click, while data-driven attribution spreads it using patterns in past data. Neither is the truth. Each is a model, and each hides something, so use attribution to compare channels directionally and not to settle arguments about exact return.
Why attribution exists
Customers rarely convert on the first visit. A person might see a video, search your name a week later, click a paid ad and then buy. Attribution is the rule that splits credit across that path. Change the rule and the same sales produce different channel results.
The models you can actually use
Google has narrowed the list. In Google Analytics 4, the Attribution reports offer three models, and Google’s help page says first click, linear, time decay and position-based were deprecated in November 2023 (Google Analytics Help).
| Model | How credit works | Typical strength | Typical blind spot |
|---|---|---|---|
| Data-driven | Machine learning compares converting and non-converting paths to estimate each touchpoint’s contribution | Reflects more than one touchpoint | Opaque, needs enough data, specific to your account |
| Paid and organic last click | 100 percent to the last channel clicked, ignoring direct traffic | Simple and easy to explain | Ignores earlier discovery |
| Google paid channels last click | 100 percent to the last Google Ads channel clicked, otherwise falls back to paid and organic last click | Useful for Google Ads review | Narrow view of the journey |
The same GA4 page says all three exclude direct visits from credit unless the whole path is direct visits. It also says data-driven models consider factors such as time to conversion, device type, number of ad interactions, exposure order and creative type.
Google Ads has a similar position. Its help page describes last click, which gives all credit to the final clicked ad and its keyword, and data-driven, which splits credit using your account’s past data and is the default for most conversion actions. Google no longer supports first click, linear, time decay and position-based in Google Ads, and conversion actions using them were moved to data-driven (Google Ads Help).
What the model changes in practice
In Google Ads, the attribution model affects the Conversions columns and any bid strategies that use them, such as Target CPA or Target ROAS. Changes apply to future counting only. The “current model” columns show how history would have looked under another model, and a model comparison report shows two models side by side. That makes it useful to compare, not just to switch.
If you are choosing campaigns, see Performance Max versus Search. The interaction between bidding and attribution is also covered in Google Ads budget and bidding.
The limits you cannot model away
- Upper-funnel work is under-credited. Video, PR and organic content often start journeys that last click then credits to search or direct.
- Click-based models miss views. Someone who saw a video and later searched may not appear as influenced.
- Different tools disagree. Each platform counts its own conversions and often uses different windows, so totals rarely add up.
- Privacy and consent reduce visibility. Not every visitor can be tracked.
- Small volumes mislead. A handful of conversions cannot support a data-driven conclusion.
- Offline outcomes are invisible unless you import them.
A practical reading method
- Make sure conversions are tracked correctly first, see conversion tracking setup.
- Compare two models side by side for the same period, and note which channels move most.
- Treat large differences as a prompt to investigate, not a verdict.
- Add non-click signals: branded search trends, direct enquiries, a “how did you hear about us” field.
- Where budget allows, test with a pause or holdout in one region or period and watch overall results.
- Judge the whole system on business outcomes such as enquiries and revenue, as described in measuring creative campaign performance.
Choosing a default
If you have limited data, last click is simple and transparent, which has value when a team needs a shared language. If you have healthy conversion volume and run several touchpoints, data-driven is a reasonable default and is what Google Ads uses for most conversion actions. Whichever you choose, document it and avoid changing models mid-campaign without noting the break in the data.
Where Radiance Impact fits
Radiance Impact’s Growth & Marketing division works on search visibility, campaigns and strategy, including how results are measured. To talk through how to read your own numbers, see /growth or send an enquiry through /contact.
Sources
- Attribution models in GA4, Google Analytics Help
- Attribution models in Google Ads, Google Ads Help
