Geo-Lift Testing

Paid Media

Also: Geo Testing · Geo Experiments · Geographic Lift Testing

Lift = (Sales in test regions minus expected sales) ÷ Expected sales
What it testsWhether ads actually caused sales
HowTurn spend on in some regions, off in others
TimeframeUsually four to eight weeks
ReplacesAttribution guesswork for big-budget calls

Quick definition

Geo-lift testing measures whether advertising actually causes sales by comparing regions where the ads run to similar regions where they don't. Instead of trusting a click-based attribution model, you run a real-world experiment across geography and compare the results.

How it varies across Australia

Geo-lift testing is still rare among mid-market Australian advertisers, mostly used by larger national brands with enough regional spread to build clean test and control groups. Smaller Australian markets make it harder to find comparable regions, but the method is becoming more common as businesses lose confidence in click-based attribution.

See acquisition benchmarks across Australian industries

What it actually means

Geo-lift testing answers a question attribution can't. Not 'who gets credit for this sale' but 'would this sale have happened anyway'. That's the harder, more useful question.

The method splits your market into two groups of similar regions. In one group, you keep advertising running as normal. In the other, you turn it off, or hold it at a reduced level. Then you compare the sales difference between the two groups over the test period. If the ad-on regions grew meaningfully more than the ad-off regions, that's the incremental lift the advertising caused.

This is the same logic as incrementality testing, just applied at a regional level instead of a user level. It sidesteps the problems that plague click-based attribution and multi-touch attribution models entirely, because it doesn't rely on tracking pixels, cookies or UTM parameters at all. You're not trying to trace an individual customer's journey. You're comparing aggregate outcomes across geography.

The tradeoff is scale and patience. You need enough regions, enough sales volume, and enough weeks to get a statistically credible answer. It's not a tool for a small local business testing a single Google Ads campaign. It's a tool for national or multi-region advertisers trying to settle a real argument about whether a channel is working.

Attribution tells you a story about who gets credit. Geo-lift testing tells you what actually happened if you'd never spent the money.

How to calculate it

Lift = (Sales in test regions minus expected sales) ÷ Expected sales

Worked example. Test regions generated $220,000 in sales during the four-week test. Based on the control regions and historical trend, expected sales without advertising were $190,000. Lift = ($220,000 minus $190,000) ÷ $190,000 = 15.8% incremental lift attributable to the campaign.

The Australian context

Australia's smaller population and concentrated capital cities make clean geo-lift testing harder than in the United States. Fewer comparable regions exist, and media markets like Sydney and Melbourne behave differently enough from regional Queensland or Western Australia that matching test and control groups takes real statistical care. Businesses that only operate in one or two states often can't run a proper geo-lift test at all and need to fall back on time-based holdout tests instead.

Where people get this wrong

Running the test for too short a period.Regional sales are noisy week to week. A two-week test rarely produces a result you can trust. Most credible tests run for a minimum of four weeks, often longer for lower-frequency purchases.
Picking test and control regions that aren't actually comparable.If the control region has different seasonality, different competitor activity or a different customer mix, the comparison is meaningless. Matching regions on historical sales patterns before the test starts is not optional.
Treating one geo-lift result as permanent truth.Lift can change with seasonality, creative fatigue and competitive activity. A result from last year's test doesn't guarantee the same channel still works the same way today.

Geo-Lift Testing vs Attribution

Geo-Lift TestingAttribution
What it measuresActual causal impact of spend on salesWhich touchpoint gets credit for a sale
Relies on tracking pixels or cookiesNoUsually yes
SpeedSlow, weeks per testInstant, always-on
Best used forSettling big budget decisionsDay-to-day channel reporting

Related terms

Common questions

How long does a geo-lift test need to run?

Most credible tests run for four to eight weeks, depending on your sales cycle and how noisy your regional sales data is. Shorter tests risk mistaking normal week-to-week variance for real lift. Longer purchase cycles need longer tests.

Can a small business run a geo-lift test?

Only if it operates across enough distinct regions with enough sales volume to find a credible control group. A business trading in one city usually can't split itself geographically and should consider a time-based holdout test instead.

Is geo-lift testing better than attribution?

They answer different questions. Attribution tells you which touchpoint gets credit inside your existing tracking. Geo-lift testing tells you what would have happened without the spend at all. Businesses making large budget decisions often need both, but geo-lift settles the causal question attribution can't.

What data do I need to run a geo-lift test?

Historical sales data by region for at least a year, enough regions to split into comparable test and control groups, and a sales or revenue metric that can be measured cleanly at the regional level without relying on individual user tracking.

Debrief

Get the next one

No spam. No fluff. Just the next article, straight to your inbox.

Keep exploring

About New Rebellion

New Rebellion is a marketing intelligence consultancy. We build tools, score Australian businesses on how their marketing actually performs, and publish Debrief every day. This dictionary is part of how we work in the open.

How we think →