Your best-looking campaigns are usually the ones adding the least. Here is how to find out which paid spend is actually creating pipeline.
Incrementality testing tells you how much of your paid pipeline would have shown up anyway. You turn a channel off, or hold it back from part of your market, and watch what happens to total leads and revenue. If the total barely moves, that spend was taking credit, not creating demand. For B2B teams with long sales cycles and real monthly budgets, this is the difference between scaling a channel that works and feeding one that only looks good in a report.
Most paid teams have never run this test. They optimize toward the dashboard instead, and the dashboard has a bias built into it.
The short version:
→ Last-click reporting rewards the channels sitting closest to the conversion, not the ones that caused it.
→ The cleanest test is to pause a channel and measure total pipeline, not that channel's own numbers.
→ When you cannot go dark, hold the channel back in some regions and compare against the rest.
→ None of this means anything until your conversion tracking is trustworthy.
→ Then move budget toward the spend that proves it is incremental, not the split you agreed to at kickoff.
What incrementality testing actually measures
Incrementality is the share of conversions that happened because of the ad, not merely alongside it. A branded search click from someone who already typed your company name into Google was probably going to reach you either way. A retargeting impression served to a buyer who already booked a demo did not change the outcome. Incrementality testing separates the demand you captured from the demand you created.
That distinction is the whole game in B2B. When a deal takes four months and touches eight people, almost every channel can claim it touched the buyer somewhere. The question is not who touched the deal. It is which spend, if you removed it tomorrow, would actually shrink your pipeline.
Ready for paid ads that pay off?
Book your free auditWhy last-click makes your worst channels look like your best
Last-click attribution hands credit to whatever the buyer clicked right before converting. Branded search and retargeting live at the finish line, so they collect the trophies. The channels doing the harder work of creating demand earlier, like non-branded search or paid social, get underpaid in the report.
Here is the pattern I see constantly. Say you spend $50,000 a month across four channels. Branded search shows a $28 cost per lead and a conversion rate three times everything else. It looks like your hero. So the instinct is to pour more budget into it.
Then you pause branded for two weeks and total leads drop by four percent. Not forty. Four. Those buyers just found you through the map pack, an organic result, or a direct visit instead. You were paying a premium to intercept demand you had already earned.
The prettiest line in the report is usually the least incremental. That is not a coincidence. It is the mechanics of last-click.
The channel-pause test
The most honest test in paid media is also the scariest one: turn a channel off and see if anything breaks. Most teams never do it because going dark feels like malpractice. That fear is exactly why the test is worth running.
Here is how to run it cleanly:
1.) Pick the channel you are most reluctant to pause. The reluctance is the signal that you have never actually verified its value.
2.) Define a window long enough for your sales cycle to register the change. Two weeks works for fast-moving lead gen. Longer cycles need four to six.
3.) Hold everything else steady. No new landing pages, no budget shifts elsewhere, no seasonal spike in the middle of the test.
4.) Measure total pipeline, not the paused channel's dashboard. You want the effect on the whole business, because that is the only number a channel cannot game.
5.) Give Smart Bidding time to recover afterward. Pausing and restarting resets the learning, so do not judge the channel's first few days back.
6.) Repeat before you act on it. One test in one month is a data point, not a verdict, especially in B2B where volume is thin.
If total pipeline falls hard when a channel goes dark, that spend is real. If it barely flinches, you just found budget.
When you cannot go dark, run a geo holdout
Some channels are too important to switch off entirely, and some executives will never sign off on it. That is fair. The alternative is a geo holdout: keep the channel running in most of your markets and turn it off in a representative slice, then compare conversion rates between the two groups.
This is the logic behind Google's built-in Conversion Lift tools, which run a controlled experiment by holding some users back from your ads and measuring the gap. For larger advertisers, running it inside the platform is far less painful than building a geo study by hand. For smaller accounts, a manual regional split still works if you have enough volume to see a real difference.
The principle is identical to the pause test. You are creating a group that does not see the spend, so the difference between the groups is the incremental effect.
Do not test on a broken foundation
An incrementality test is only as good as the conversion data underneath it. If your tracking is dropping conversions, double-counting, or firing on the wrong event, the test measures noise and you will make a confident decision on garbage.
So before you pause anything, confirm the basics: click identifiers are captured at lead creation, conversions are deduplicated across accounts, and the event you optimize toward reflects qualified pipeline rather than raw form fills. We wrote separately about why leads suddenly drop to zero when tracking breaks, and it is worth reading before you run any experiment. A clean signal is not a nice-to-have here. It is the prerequisite.
Move the budget, not just the slide
The point of testing is not a tidy chart for the quarterly review. It is a decision. Once you know which spend is incremental, budget should follow live conversion economics, not the media plan you locked in at kickoff.
That usually means uncomfortable moves. Trimming the channel everyone loves because it turned out to ride on existing demand. Funding the channel that looked mediocre on last-click but moved total pipeline when it ran. Most of what we found about branded versus non-branded search came from exactly this kind of test, and the answer is rarely the one the dashboard predicts.
Pick the one channel you are most afraid to turn off. Run a two-week pause, watch total pipeline, and let the result decide the budget instead of the report. If you want a second set of eyes on the design before you go dark, that is the kind of thing our B2B Google Ads team does before touching a dollar of spend.
