Incremental sales = sales with the marketing − sales without it
310 orders − 298 orders = 12 orders
Which sales would have happened anyway?
More than the ad accounts suggest. A customer who searches your business by name already knew you, and many would have found you without the ad. A regular who clicks a retargeting ad was going to reorder this week. The ad accounts count both as theirs, because an ad was clicked. Incrementality strips those sales out and leaves the ones the marketing caused.
The gap is widest on ads that sit close to the sale, such as brand search and retargeting, and narrowest on ads that reach people who had never heard of you.
How does a lift test work?
You need a comparison group that did not see the marketing. There are three common ways to get one. A holdout splits an audience at random, and only half is shown the ads. An on and off test pauses the ads for a period and compares sales with the period before, or the same weeks last year. A geographic test runs the ads in one area and not in a similar one.
Meta and Google Ads both run formal lift studies (Meta calls its version a Conversion Lift test), and both have eligibility rules on spend and conversion volume that many small accounts do not meet. Check the Experiments section of each account. For most owner-led businesses, a carefully run pause is the practical version.
Did an Albury florist's brand ads earn their keep? (example figures)
A florist in Albury spends $600 a month on Google Ads for searches of its own name. Google Ads credits those ads with 80 online orders a month, $7.50 an order. The owner pauses them for four quiet weeks, well clear of Valentine's Day and Mother's Day, and counts every online order.
| Four weeks with the ads | Four weeks paused | |
|---|---|---|
| Total online orders | 310 | 298 |
| Orders credited to the brand ads | 80 | 0 |
| Incremental orders | 310 − 298 = 12 | |
| Cost per incremental order | $600 ÷ 12 = $50 |
Of the 80 orders the ads were credited with, about 12 depended on them. The rest came through the free listing underneath when the ad was not there. At an average order of $95 and a 50 percent margin, each order leaves $47.50, so the brand ads were roughly breaking even at $50 an order, not earning a 12 times return as the account implied.
A drop of 12 orders is also within the florist's normal week-to-week swing, so the honest reading is that the ads added little, and possibly nothing. The owner kept a small brand budget for a reason the order count could not show: a competitor had started bidding on the florist's name, and without an ad of its own, the florist's listing sat below the competitor's ad.
Can CoraVision run a lift test for you?
CoraVision does not run lift tests or pause anything; you make every change yourself. It reads your Google Ads and Meta campaigns and your Google Analytics 4 purchases every night, so the before and after of a pause you choose to run is already on record. A test like this can become a step in your action plan, and the weekly check-ins are where its result gets read against the one number you are chasing.
Questions
How is incrementality different from attribution?
Attribution divides up the credit for sales that happened. Incrementality asks which of those sales would have happened anyway. An ad can be credited with 80 orders by attribution and be responsible for 12 of them by incrementality.
How long should a pause test run?
Long enough that ordinary ups and downs do not drown the result: several weeks, and longer if your sales swing a lot from week to week. Keep public holidays and big events out of the test, and compare with the same weeks last year if you have them.
Can a small business run a test by area?
Only if it serves areas far enough apart that customers do not cross between them, such as two towns an hour apart. Run the ads in one and not the other, then compare each town's sales against its own usual level. A business in one town is better off with an on and off test.