Direct answer: Attribution records that an observed action was associated with a campaign touchpoint. Incremental lift asks how many more actions occurred because the campaign ran. A tagged lead report can help operate the campaign, but a causal claim needs a credible comparison with what would have happened without it. When that comparison is not feasible, report observed results and uncertainty plainly.
Suppose a launch film and landing page produce 120 tagged enquiries. That number is useful: the team can inspect the enquiry quality, the page, and the sales follow-up. It does not mean all 120 enquiries were new. Some people might have contacted the business through another route anyway. This distinction is a measurement problem, not a criticism of the campaign or the analytics tool.
What controlled evidence adds
In Vaver and Koehler's research on geo experiments, geographic regions can be randomly assigned to test and control conditions, with ad exposure changed in the test regions. The comparison aims to estimate the effect of the advertising rather than merely count conversions after an ad interaction. Google Ads' current description of Conversion Lift similarly compares groups that see ads with groups that do not, and defines the difference in conversions as lift. Google says that feature is not available to every account.
The appeal of a control group does not make a test easy or automatically precise. Kerman, Wang, and Vaver's follow-on paper explains that earlier geo methods can be less applicable when few geographic units are available; the authors developed a time-based approach to address that setting and evaluated it with simulations. That is a specialist analysis, not an invitation to split two cities casually and call the difference causal.
Another reality check comes from Lewis and Rao's analysis of 25 large advertising field experiments: even at substantial scale, return-on-investment estimates could be very imprecise. Their median ROI confidence interval was more than 100 percentage points wide. Their study does not give a required budget for every campaign. It shows why a small campaign may be unable to distinguish a modest lift from ordinary variation.
Use a measurement ladder that fits the decision
| Evidence available | What it supports | What it does not establish |
|---|---|---|
| Tagged visits and enquiries | Which links and landing pages recorded actions | How many actions were new because of the campaign |
| Enquiry quality and sales follow-up | Whether the recorded actions were useful to the business | A causal return on ad spend |
| Pre-planned controlled comparison with sufficient data | An estimate of incremental actions for that test and period | A permanent effect in every future market |
Before launch, write down the business action, the tracking definition, the time window, and any planned comparison. If a platform lift study is available, check eligibility and the likely precision before reserving budget. If it is not available, do not invent a control group after the results arrive. You can still report the campaign's spend, exposure, tagged enquiries, qualified enquiries, and limitations.
For a creative agency and client, this helps keep responsibilities clear. ARCHV can build campaign work and landing experiences around a defined objective. The media, analytics, and sales owners need to agree how outcomes will be captured and interpreted. A product-page A/B test can answer a narrower page-design question; it should not be mistaken for a test of the whole campaign's incremental effect. The honest final report separates what the campaign recorded from what the evidence says the campaign caused.