In geographic experimentation (GeoX), you can leverage 3 primary study types: go-dark, holdback, and heavy-up. Each is aligned with distinct testing objectives. Selecting the optimal methodology depends on whether your strategic focus is validating an existing budget, testing a new marketing strategy, or justifying investment growth.
For all open-source geo experimentation, the study implementation occurs through campaign modification directly in the advertiser platforms (in this case, Google Ads).
For the separate Google Ads beta, learn how to Set up Conversion Lift based on geography.
On this page
- Types of geo-based studies
- Implementation paths
- Set up a single-cell experiment
- Set up a multi-cell experiment
- Set up a cross-publisher multi-cell experiment
- Optimize performance
- Best practices
Types of geo-based studies
- Holdback (net new / prove new value): Validate the incremental returns of entirely new ad strategies and unproven channels.
- Example: You launch a brand-new Demand Gen strategy in certain regions while withholding it from a control group to prove its effectiveness before scaling. If the exposed regions show a statistically significant lift in sales, you have causal truth for the effect of adding Demand Gen to your ad strategy.
- Go-dark (validate base investments): Completely shut off spend in test geos to definitively prove the value of existing, live campaigns.
- Example: To test if "Always on YouTube" drives incremental sales or just captures organic conversions, turn off spend in select treatment regions. If sales in these "go-dark" regions drop significantly against the baseline, YouTube's impact is proven.
- Heavy-up (justify scaling): Inject incremental budget to existing campaigns to forecast the revenue impact of scaling.
- Example: Increase Performance Max spend in test regions to confirm if additional dollars remain profitable and guide long-term budget growth.
Implementation paths
There are 3 main tooling options available for implementing these setups:
- Google Ads UI: This manual process is best for straightforward, single-campaign setups and requires no coding skills.
- Data requirements: City IDs or ZIP codes for international targets (non-USA) and DMA regions for USA-based targeting.
- Google Ads Editor: This manual process requires no advanced technical skills and allows for copying, pasting, and updating campaigns in bulk (such as bulk campaign migration and mass bid & budget updates).
- Google Ads API: This fully automated process is the best option for highly complex, continuous testing frameworks and requires developer resources.
- Data requirements: Control & test city IDs (non-USA) and DMA (USA). ZIP code level targeting is not supported.
Set up a single-cell experiment
Holdback
- Go to Campaigns within the Campaigns menu, create a new campaign, and pause it.
- Select the Settings icon
and expand the Locations section.
- Remove any country-level targeting, and select Enter another location.
- Select Advanced search and check the box next to Add locations in bulk.
- First specify the country, then paste test geo IDs for targeting and select Save.
- Expand the Location options section and select Presence to prevent location leakage.
- Assign a hard, individual campaign daily budget cap (ensuring it isn't linked to a shared budget), and unpause the campaign when the experiment starts.
Go-dark
Uncapped budget (targeting only)
This setup is for targeting-only go-dark experiments that aren't budget constrained.
- Go to Campaigns within the Campaigns menu, and select the Settings icon
- Expand the Locations section, remove any country-level targeting, and select Enter another location.
- Paste and target control geo IDs only.
- Expand the Location options section and select Presence to avoid leakage.
- Ensure this campaign is not part of any shared budgets.
Capped budget (duplication and budget cut)
This setup is for duplication and budget cut go-dark experiments that are budget constrained.
Targeting-only is the most effective method to suppress spend. In budget-constrained scenarios, removing test geos triggers immediate budget shifting, as the system redistributes unspent funds to control regions and inflates the baseline. To prevent this, manually reduce the daily budget to align with control-only run rates, and apply individual campaign budgets instead of shared or portfolio options.
- Go to Campaigns within the Campaigns menu, and duplicate the active campaign.
- Keep the duplicated go-dark campaign paused throughout the experiment to serve as a placeholder.
- In your active control campaign, expand the Locations section and remove any country targeting.
- Paste and target control geo IDs only.
- Expand the Location options section and select Presence to avoid leakage.
- In the "Budget and bidding optimization" section, expand the Budget section.
- Set your average daily budget by finding the historical percentage of spend that occurred in the control geos and multiplying your current daily budget by this percentage.
Ensure that this campaign doesn’t use a shared budget. Learn how to Manage a shared budget across campaigns.
Heavy-up
Regional modifier (manual bidding and uncapped budgets)
For heavy-up experiments, you can only use location bid adjustments for campaigns using manual bidding with uncapped budgets targeting single-networks. Location bid modifiers are ignored by Smart Bidding and cross-network formats.
- Go to Campaigns within the Campaigns menu, and select the Settings icon
- Expand the Locations section, select Enter another location, remove any country-level targeting, and ensure both control and test geos are targeted.
- Expand the Location options section and select Presence.
- Adjust your bid modifiers. Keep control geos at 0% adjustment and set test geos to a positive modifier (for example, +50%).
Duplication
For heavy-up experiments, campaign duplication is required for any budget-constrained campaigns, Smart Bidding, and cross-network formats.
- Go to Campaigns within the Campaigns menu, and duplicate the active campaign.
- Configure your control campaign: Select the Settings icon
- Configure your heavy-up campaign: Select the Settings icon
- Manage the heavy-up spend based on your budget setup:
- Capped budgets: Manually increase the duplicated test campaign's daily budget (for example, by 50% relative to its historical run rate).
- Uncapped budgets: Loosen the bidding targets on the duplicated campaign (such as Target CPA or Target ROAS) to force the system to spend more. You can’t run heavy-up using non-target Max Conversions or Maximize Conversion Value bidding with uncapped budgets.
Use a Shared Budget on Smart Bidding campaigns to guarantee that highest-serving campaigns within this arm consume the budget efficiently without hitting individual daily caps.
Set up a multi-cell experiment
Budget neutral (control vs. heavy-up vs. go-dark)
To get started, go to Campaigns within the Campaigns menu, and duplicate the active campaign. Label them to distinguish between your control and heavy-up campaigns (for example, Campaign_Control and Campaign_HeavyUp).
Configure your control campaign
- Select the Settings icon
- Positively target control geos only, and explicitly exclude go-dark and heavy-up geos.
- Expand the Location options section and select Presence.
- Set the control campaign to an individual campaign budget matching the control geos' historical run rate.
Configure your heavy-up campaign
- Select the Settings icon
- Positively target heavy-up geos only, and explicitly exclude go-dark and control geos.
- Expand the Location options section and select Presence.
- Manage spend injection in your heavy-up campaign based on your budget setting:
- Uncapped budgets: Set a positive bid modifier (for example, +50%) for manual bidding campaigns, or loosen tCPA/tROAS targets directly. If managing multiple Smart Bidding campaigns in this arm, you can use an intra-arm portfolio strategy, but never share with the control campaign. Note: You can’t run heavy-up using non-target Max Conversions or Maximize Conversion Value bidding with uncapped budgets.
- Capped budgets: Increase the daily budget cap by 50% relative to its historical run rate. If managing multiple capped campaigns in this arm, you can use an intra-arm shared budget, but never share with the control campaign. Optionally, loosen targets or increase bids slightly if budget scaling alone does not consume the full test volume.
By explicitly omitting and excluding go-dark geos from both active campaigns, spend for those regions naturally suppresses to $0 without the need to create a third active campaign.
Set up a cross-publisher multi-cell experiment
Important: Multi-cell cross-publisher tests require explicit cross-platform parity. Guarantee that all platforms are operating on identical conversion windows, utilizing the same audience exclusion logic, targeting the exact same geographic boundaries, and being measured against an independent, unattributed first party data source. Ensure campaign types and goals between platforms are comparable.
Go-dark (ablation)
Use a 3-cell routing logic to compare baseline incremental value. Cell 1 is the Google suppression arm where Google Ads goes dark. Cell 2 is the External suppression arm where the external publisher goes dark. Cell 3 is the Control arm where both operate at standard levels.
Targeting only (uncapped)
- Go to Campaigns within the Campaigns menu, and select the Settings icon
- Expand the Locations section, select Enter another location, and remove any country-level targeting.
- Positively target only geo IDs assigned to Cell 2 and Cell 3, and explicitly exclude the geo IDs for Cell 1.
- Expand the Location options section and select Presence to ensure zero residual traffic leaks.
- Maintain standard individual campaign budget and flat business-as-usual efficiency targets.
Google spend in Cell 1 (Google suppression) naturally suppresses to $0 with zero baseline distortion in Cells 2 and 3.
Capped (duplication and compression)
- Go to Campaigns within the Campaigns menu, and duplicate an active campaign to isolate spend. Keep the duplicated go-dark campaign paused throughout the experiment.
- In the active control campaign, select the Settings icon
- Positively target Cell 2 and Cell 3 geo IDs, and explicitly exclude Cell 1 geo IDs.
- Expand the Location options section and select Presence for both targeting and exclusions.
- Manually adjust your daily budgets downwards to match the natural run rate of the control-only geos (for example, if Cell 2 and Cell 3 account for 70% of spend, reduce the budget to 70%).
Ensure that this campaign doesn’t use a shared budget. Learn how to Manage a shared budget across campaigns.
Optimize performance
Navigate learning periods
- Learning periods: Campaign duplication triggers a learning period because the newly duplicated campaign lacks baseline historical data, while the original campaign retains it. Remove the learning period from your final analysis.
- Mitigation strategy: Duplicate campaigns as soon as possible after the design phase and wait a specified number of weeks to ensure the campaigns are more likely to converge.
- Heavy-up experiments: Factor in the typical 4-5 day campaign-dependent learning period, which must be fully included inside the test period.
- 20% budget shift rule: Any budget change greater than 20% necessitates a new learning period, whereas bid modifiers under 20% can be applied without a severe learning reset.
Advanced budget management
- Label campaigns: It’s highly recommended to label your test campaigns e.g ‘Control’ and treatment label ‘Heavy-up’ or ‘Go-dark’ (depending on the study) during campaign modification to keep track.
- Shared budgets & portfolio bidding: Should ONLY be applied uniquely at cell level and within a cell, never across campaigns in other cells. Ensure any campaigns are broken out from existing strategies that are not part of the test. Recommended method for campaigns with capped budgets, but never for go-dark tests.
- Looker Pacing Dashboard: Map campaign labels to your GeoX design file in Looker to monitor real-time spend pacing and control stability, including to verify $0 go-dark leakage.
Cooldown protocols
- Treatment arm reset: During the post-experiment cooldown phase, advertisers should immediately revert the treatment arm campaigns to their "business as usual" (BAU) budgets for both go-dark and heavy-up tactics.
- Holdback and flat budgets: If the tested tactic was a holdback, completely stop the campaign post-test. Don’t increase the campaign budget under any circumstances for go-dark tactics during cooldown.
Best practices
- Clean test: Monitor for any major regional launches occurring during the test and proactively exclude those regions to avoid skewed results. Ensure no other regional testing is occurring on the client-side (such as offline testing). Finally, avoid running parallel user-level measurements, such as Brand Lift, Search Lift, or Conversion Lift, on campaigns included in a GeoX study.
- Follow campaign best practices: Your test is as efficient as the campaigns you run. Make sure all campaigns adhere to best practices, such as applying Demand Gen guidelines to your Demand Gen campaigns and Performance Max guidelines to your Performance Max campaigns.
- Limit changes: Maintain a stable test environment by limiting changes during the study. Ideally, avoid all changes to prevent introducing noise.
- Excluding country targeting: Note that switching from country to city or zip code targeting (as required) will reduce overall traffic, as granular location data can’t be determined for all users.
- Data analysis and learning periods: Pre-test data must be at least 3 times the duration of the test period. If any outages (website/app) or similar data noise occur that could impact the primary KPI, consider removing that date from the final analysis.
- Budget excluded geos: It’s important to understand that excluded geos from your experiment will not receive BAU budget when duplicating campaigns. Usually the fraction is small, but use campaign labels and Looker to check for spending in these regions and consider starting a BAU campaign to stay on track with your business and other regional sales priorities. Note that exclusions can also occur organically with manual selection or with geo mismatch when performing geo-unit mapping.
Related links
- Target ads to geographic locations
- About bid adjustments
- Set an average daily budget for your campaign
- About Lift Studies
- Meridian GeoX developers site - Learn how to anchor your marketing strategy and calibrate your marketing mix models (MMM) with Google's global, open-source incrementality solution.