Geo Experiments: Testing Advertising Across Markets

Choose a market question worth testing

The team wants to know whether an extra newspaper insertion or a larger digital budget will produce additional sales. Asking for a geo experiments plan is useful only when that question can be turned into an intervention across places. A region receives a defined change while a credible comparison provides evidence about what would otherwise have happened. The work begins with feasibility: can the media be varied geographically, are outcomes available by location, and can the expected effect be separated from ordinary sales variation? A map of two coloured regions is not yet an experiment. For a wider view of the options, explore our marketing frameworks and measurement guide.

Start with the intervention and the geography

Define whether the study introduces media, increases it or withdraws it. Specify the units you can actually control: cities, districts, service areas or other defensible regions. Avoid choosing boundaries simply because they look neat in the presentation. People commute, newspapers travel and ecommerce customers can order from another location. Google’s GeoX guidance includes design, historical data preparation and analysis rather than treating geography as an automatic solution. ¹ Explain how orders will be assigned to regions and whether that rule fits the media exposure. A customer’s billing address may not describe where they saw an advertisement. For the wider picture, see our guide to newspaper advertising effectiveness.

Use pre test behaviour to assess the design

Historical regional outcomes help reveal volatility, seasonality and structural differences. A city with large weekly spikes from one wholesale customer may be unsuitable for a small retail media test. Assess the amount of pre-test data needed for the chosen method and investigate missing dates or inconsistent geography codes. If regions differ in size, compare appropriate normalised outcomes or use a justified modelling approach. Simple equal-sized boxes on a map can hide very different commercial populations. Where feasible, randomisation improves the design, but operational constraints still need to be documented before results are visible.

Protect the comparison during execution

The intervention should be implemented as planned. Keep promotions, pricing and distribution changes as comparable as practical, and record any exceptions. Media leakage, a dealer opening or a competitor disruption can compromise interpretation. A geo experiments brief should name who monitors delivery and who can approve changes. Watch spend and geographic targeting, but avoid ending the study because an early chart looks favourable. Prespecified stopping rules reduce that temptation. If execution fails, report the failed implementation rather than treating the test as a clean estimate of the intended campaign. If that is your next question, read our guide to marketing incrementality.

Translate the estimate into a business decision

The analysis compares observed outcomes with an estimated counterfactual under the selected design. Report the incremental outcome, observation period and uncertainty interval. Relative lift uses the counterfactual outcome as its denominator. Incremental revenue divided by the added spend can produce an iROAS estimate when those definitions match. Profit assessment requires contribution economics and relevant costs. A test of an additional INR 100,000 does not establish returns at an additional INR 1,000,000; media response can change with scale. Keep the recommendation attached to the tested intervention and avoid extending a local result to every market without further evidence. Marketing mix modelling is explained in a separate guide.

Prepare the field teams as carefully as the model

Regional sales teams need to understand what can change and what must remain consistent. Brief them on campaign response handling, stock escalation and data recording. Do not ask them to suppress ordinary customer service to preserve a study. If an operational change is necessary, log it and ask the analyst how it affects inference. Quiamo can help write a media and communication brief that works with the research design, then coordinate the creative and response journey. Qualified experimental analysis should remain a defined responsibility rather than an assumption hidden inside a creative production fee.

Decide how the study will be reviewed

A useful review includes marketing, finance and operations. Begin with whether the intervention was delivered, then examine data quality, the estimate and its uncertainty. Only then discuss budget changes. Keep planned analyses distinct from exploratory regional findings. A surprisingly strong city may suggest a future test, but selecting that city after seeing the result does not turn the subgroup into the original hypothesis. Preserve the design document, data version and analysis decisions so the team can revisit the result when conditions change. The report should also name what was learned about response handling and distribution, even if the effect estimate remained inconclusive.

Put the measurement plan into practice

Specify how an order changes region when a customer relocates, collects from a store or returns an item elsewhere. The rule should fit the business and remain consistent through the experiment. Build checks for missing geography, delayed order feeds and accidental targeting changes. A daily operational log helps the analyst distinguish genuine market movement from data failures. If a local team launches an unplanned promotion, record the dates and scope rather than removing inconvenient observations after the fact. The study can then be assessed on its actual execution. Geographic boundaries are useful only when the commercial data and campaign delivery respect the assumptions behind them.

Before the campaign goes live

Use this checklist to organise evidence about test feasibility, spillover and the detectable commercial effect. Assign owners to the tasks and keep the supporting evidence with the campaign record.

• Specify the media intervention and the decision it will inform.
• Audit geographic outcomes and historical delivery data.
• Assess contamination, operational constraints and detectable effect.
• Prespecify assignment, duration, stopping and analysis rules.
• Monitor execution and record pricing, stock or distribution exceptions.
• Report lift, uncertainty and contribution economics together.

Questions people ask

Bring the measurement question into the brief

A geographic test is credible when the market boundaries, outcome data and execution support the comparison. A geo experiments brief should make those assumptions readable before any region changes spend. Bring Quiamo the intervention you want to test and the regional data available so the media and communication work can fit a realistic research design. The result should inform the specific budget decision, with its uncertainty intact, rather than become a permanent promise about every future campaign. Explore our campaign strategy and measurement support to see how the work can be structured. You can discuss your campaign measurement brief with Quiamo.

Sources

1. Google Meridian introduction to GeoXGeographic experiments compare marketing interventions and require historical geographic data and counterfactual analysis.
2. Google Meridian introductionMMM uses aggregated outcome, media and relevant control data; geo-level analysis is supported.

Worked examples on this page are fictional and labelled. Search phrases and figures from external sources are cited; nothing here describes a Quiamo client result.