Case note — paid media measurement
The four ads that never got credit
A lipstick brand almost doubled its budget on the wrong ad. Here’s what changed when we stopped asking which ad closed the sale, and started asking which ads made the sale possible.
Same customer, same five ads, same purchase — two very different budget conversations.
The brand sells premium lipstick, direct to consumer. Their Google Ads dashboard was clear about which ad worked: a plain brand-search ad, the kind that shows up when someone types the company’s name. It had the conversions. It had the credit. The obvious move was to pour more budget into it.
Before that happened, we turned on data-driven attribution and re-ran the same month of sales. The story didn’t just shift — it reversed. The brand-search ad hadn’t built the sale. It had simply been standing at the finish line when the customer arrived.
Last-click attribution gives the trophy to whoever scores. It has nothing to say about who passed the ball four times to get them there.
Why the last click lies
Last-click is the default because it’s easy to explain: whichever ad a customer clicked right before converting gets all the credit, full stop. Everything earlier in the journey — the ad that first caught their eye, the one that brought them back the next day, the one that pushed them from browsing to cart — gets nothing.
That’s not a rounding error. In a five-touchpoint path like the one above, four real, working ads report as zero-value line items. Anyone deciding budget from that report will fund the wrong thing — usually branded search, since customers tend to search a brand name right before they buy anyway, regardless of what convinced them to.
What data-driven attribution actually does
Data-driven attribution (DDA) doesn’t follow a fixed rule like “first touch wins” or “split it evenly.” It’s a model trained on your own account’s conversion history — the paths of customers who bought, compared against the paths of customers who didn’t. Where a rules-based model would insist every path looks the same, DDA looks at what device was used, how much time passed between ads, and what order those ads appeared in, then works out which touchpoints actually raised the odds of converting.
If people who see a particular video ad go on to convert at a meaningfully higher rate than people who don’t, that ad earns a share of the credit — even if the sale itself happened on a completely different keyword, weeks later. It’s a probability model, not a stopwatch.
Last-click vs. data-driven, side by side
| Last-click | Data-driven | |
|---|---|---|
| How credit is split | 100% to the final click | Weighted across every touchpoint, based on observed impact |
| Setup effort | None — it’s the default | Needs enough conversion history to model against |
| Data required | Any volume | Roughly 300 conversions in the trailing window works best |
| What it’s good at | A quick, transparent baseline; debugging a single path | Reflecting how awareness and consideration ads actually contribute |
| Where it misleads | Rewards whichever ad happens to sit last, usually branded search | Can behave like a black box if conversion volume is thin |
Why this matters more right now
This isn’t just a best-practice suggestion — Google has been actively narrowing the choice. In 2023, Google deprecated four rules-based models — first-click, linear, time-decay, and position-based — after adoption on them fell below 3% of conversions, and by mid-2026 those models stopped being selectable for any conversion action at all. Any conversion action still running on one of the retired models is now being force-migrated to data-driven attribution automatically, with no opt-out.
Today, the Google Ads Model Comparison report only lets you compare two models: data-driven and last-click. That’s the entire menu left. If your account hasn’t been touched in a while, it’s worth checking — the migration happens quietly at the conversion-action level, with no visible change to your campaigns, budgets, or dashboards, so an untouched account can look untouched even after its attribution logic has already shifted underneath it.
The simple guide — how to switch without breaking anything
Check what your conversion actions are running now
In Google Ads, open Goals → Conversions, and look at the attribution model listed on each conversion action. Some may already have migrated automatically.
Pull the Model Comparison report first
Under Measurement → Attribution, compare last-click and data-driven side by side before changing anything live. Look for campaigns and keywords where DDA credit is meaningfully higher than last-click — those are your under-funded assist players.
Confirm you have enough conversion volume
Aim for roughly 300 conversions in your lookback window per conversion action. Below that, expect the model to need more time to stabilise.
Switch the conversion action, not the whole account at once
Change one high-volume conversion action first, and let it run for at least two to three weeks before touching bids off the back of it.
Re-brief budget, not just bids
The real payoff isn’t a setting — it’s redirecting spend toward the awareness and consideration ads that DDA reveals are actually opening the paths that convert.
Don’t thrash it weekly
Let the model settle before comparing again. Attribution shifts are a quarterly decision, not a Monday-morning one.
When last-click still earns its place
- Short, one-or-two-step paths where there’s barely a journey to model.
- Low-volume accounts that haven’t reached the conversion threshold DDA needs to be reliable.
- Fast sanity-checks and debugging — last-click’s simplicity is a feature when you just need to see what happened, quickly.
Don’t let one click write the whole story.
The lipstick brand didn’t have a brand-search problem. It had four ads doing real work with nothing to show for it on a dashboard that only counts the last touch. Data-driven attribution didn’t invent that work — it just finally reported it.
