Pakistan
USA
Canada
UK
Dubai
Australia
Germany
- Attribution shows who saw or clicked before buying, not who was persuaded.
- Lift tests compare people who saw ads with a holdout who did not.
- Geo tests work when platform lift tests are not available.
- Use test results to set targets, not to replace daily reporting.
What is the difference between attributed and incremental conversions?
An attributed conversion is any sale Meta can link to an ad within your attribution setting. That includes people who were persuaded by the ad and people who would have bought anyway, such as loyal customers, people searching for your brand, or buyers already on their way to checkout. Attribution answers who touched an ad, not why they bought.
An incremental conversion is a sale that happened because of the ad. The only way to measure it is to compare a group that could see your ads with a similar group that could not. The difference between the two is the lift. A campaign can have strong reported ROAS and weak incrementality, especially retargeting and brand-heavy campaigns. Our guide on what a good ROAS is explains why reported ROAS needs context.
How do Conversion Lift studies work?
In a Conversion Lift study, Meta randomly splits your target audience into a test group that can see your ads and a control group that cannot. It then compares conversions between the two using your pixel, Conversions API or uploaded offline data. Because the split is random, the difference is a fair estimate of what the ads caused.
Lift studies are set up in the Experiments section of Ads Manager. Self-serve access, and the conversion volume needed for a reliable result, can vary by account and change over time, so check what your account shows. Some advertisers run lift studies with help from a Meta representative. Plan the study before the campaign starts.
Can I use A/B tests in Ads Manager?
The A/B test tool splits your audience so two versions of a campaign do not overlap. It is excellent for comparing strategies, such as broad targeting against interest targeting, or one creative concept against another. It tells you which version did better, not whether either one caused sales that would not have happened otherwise.
To get closer to incrementality with A/B testing, you need an arm that receives no ads or far fewer, which is what a holdout test does. Meta's Experiments area includes holdout options for some accounts. Where they are not available, a geo test is the practical alternative. Both approaches need patience and stable settings.
How do I run a geo holdout test?
A simple geo test can be run by any business with sales data by region.
- Pick comparable regionsChoose cities or regions with similar sales patterns, for example two groups of cities that trend alike.
- Record a baselineCollect several weeks of sales by region before changing anything.
- Pause ads in the holdout regionsExclude those regions from all Meta campaigns for the test period, and keep other marketing steady.
- Measure total sales by regionUse your own sales, CRM or shop data, not Ads Manager, so every sale counts.
- Compare the changeLook at how sales moved in test regions against holdout regions, relative to the baseline.
How much budget and time does a test need?
There is no single minimum. A test needs enough conversions in both groups to show a difference larger than normal week-to-week noise. Businesses with few sales per week need longer tests or broader campaigns. Run the test for at least one full buying cycle, so people who saw ads early have time to buy.
Avoid testing during unusual periods such as major sales, Ramadan, Eid or Black Friday, unless that is exactly what you want to measure. Keep budgets, creative and other channels stable during the test, so the only meaningful change is whether people could see Meta ads. A clean test is worth waiting for.
Common measurement mistakes
These lead to wrong conclusions about Meta's value.
Trusting ROAS alone
Retargeting often shows the best ROAS and the smallest lift.
Testing too briefly
A week rarely captures the full buying cycle, especially for considered purchases.
Changing other channels
Launching a Google campaign mid-test muddies the comparison.
Comparing unlike regions
Holdout regions must behave like test regions before the test starts.
Testing once, never again
Incrementality changes as creative, audiences and markets change.
How should I act on the results?
Use lift results to calibrate, not to replace daily reporting. If a study shows Meta's reported conversions overstate what ads cause, adjust your target cost per result or ROAS for that campaign type. If prospecting shows strong lift but retargeting little, shift budget accordingly. If lift is close to zero, look at the offer, creative and audience before concluding that Meta cannot work for the business.
Then return to normal optimisation using Ads Manager data, with the adjusted targets in mind. Daily decisions about ads and ad sets still need fast feedback, which only platform reporting gives. Repeat the test when strategy, markets or spend change significantly. Our guide on why Meta and GA4 numbers differ explains the day-to-day side of reporting.
How we handle this for clients
We first check that tracking, including the Conversions API and any offline sales data, is reliable enough to measure anything. Then we agree the business question with the client, choose the test that fits their data and markets, and write down the success measure before launch. Without that agreement, any result can be argued either way.
During the test we keep other changes to a minimum and document anything that could affect results, such as a price change, a stock problem or a competitor sale. Afterwards, we use the outcome to set targets and budget splits. This sits within our Meta Ads management service and is often where an audit begins.
Pakistan, Gulf and UK advertisers
In Pakistan, much of the sale happens offline: WhatsApp orders, phone confirmations and cash on delivery. A geo test using delivered orders by city is often more realistic than a platform lift study. Lahore, Karachi and Islamabad behave differently, so compare groups of similar cities rather than one city against another.
In the Gulf, a test can split emirates or regions, though Dubai's size makes balanced groups harder. In the UK, regional tests work well for businesses with store or postcode-level sales data. Whichever market, avoid testing across a major holiday unless that period is what you want to understand. Plan around the local calendar.
Junaid has more than 18 years in SEO, paid media, web development and custom software. He founded SEOnumber1, builds WordPress plugins, CRMs and web applications, has served 500+ clients in 48 countries, and holds Google Ads, Analytics and Google Blue and Green Belt certifications. He is a National Level Award Winner in Digital Marketing. Read his story or ask a question about this article.