Client Hub →
Theme
Glossary Attribution

Holdout Group

A control group excluded from a marketing campaign to measure its true incremental impact by comparing their behaviour against exposed audiences.

Also known as: control group test control group incrementality test group holdout panel

What is a Holdout Group?

A holdout group is a segment of your audience deliberately excluded from a marketing campaign or channel activation. This group serves as a control, allowing you to measure the genuine uplift caused by your marketing efforts rather than attributing all conversions to your campaign.

Why Holdout Groups Matter

Traditional attribution models assume that everyone who converts after seeing an ad was influenced by it. This creates inflated ROI claims. A holdout group reveals how many conversions would have happened anyway – what marketers call "organic" or "baseline" conversion rates.

Without holdout groups, you risk:

  • Overestimating campaign impact by crediting natural customer behaviour to paid media
  • Misallocating budget to underperforming channels
  • Making poor strategic decisions based on false attribution data

For UK agencies managing multi-channel campaigns across competitive sectors (FMCG, retail, financial services), holdout groups provide the ground truth needed for confident optimisation.

How They Work

You randomly select a statistically significant portion of your audience – typically 5-10% – and exclude them from the campaign entirely. The remaining 90-95% receives your ads across chosen channels (search, display, social, TV, etc.).

After the campaign runs, you compare:

  • Holdout group conversion rate (no exposure)
  • Exposed group conversion rate (with exposure)

The difference is your true incremental lift. If your holdout group converts at 2% and exposed group at 3%, your incremental uplift is 1 percentage point.

When to Use Holdout Groups

Best for: - Large-scale campaigns with substantial budgets (the forgone revenue from 5-10% of audience is acceptable) - Brand awareness or consideration campaigns where measurement is ambiguous - Testing new channels or strategies - Mature campaigns where baseline conversion patterns are established

Less practical for: - Performance campaigns with limited budget (the cost of excluding potential converters is high) - Short-duration campaigns (insufficient time to gather statistical significance) - Niche audiences where 10% exclusion is prohibitively expensive

UK Media Context

Holdout groups are increasingly standard practice among sophisticated UK agencies and brands, particularly post-iOS privacy changes. As third-party cookies deprecate, holdout experiments offer a privacy-compliant alternative to cross-device tracking for measuring incrementality.

The Financial Conduct Authority and ASA expect higher evidentiary standards for marketing claims, making holdout-backed ROI figures valuable for compliance and stakeholder confidence.

Best Practices

  • Ensure holdout groups are randomly selected to avoid bias
  • Maintain sufficient sample size for statistical validity
  • Isolate variables – only exclude from one channel, not all marketing
  • Run for long enough to capture full customer journey
  • Document assumptions clearly for stakeholder transparency

Frequently Asked Questions

How big should a holdout group be?
Typically 5-10% of your audience, depending on campaign scale and budget. Larger holdout groups improve statistical confidence but cost more in forgone conversions. Use statistical power calculators to determine the minimum size needed for your expected effect size.
Won't excluding people from a campaign lose revenue?
Yes, in the short term. However, the incremental lift data you gain justifies the cost by preventing bigger budget misallocation. For a £100k campaign, sacrificing £5-10k in holdout revenue to accurately measure true ROI across future spending is strategically sound.
Can I use a holdout group for every campaign?
No. Holdout groups work best for large-scale, brand-building, or exploratory campaigns. For performance-focused campaigns with tight margins, the forgone revenue may outweigh measurement benefits. Consider using holdout tests selectively to calibrate your attribution models.
How long should I run a holdout test?
Minimum 2-4 weeks, ideally the full campaign duration. Shorter tests risk capturing anomalies rather than true patterns. Longer windows also account for variable purchase cycles across different customer segments.

Learn How to Apply This

Need expert help?

Our team can put this knowledge to work for your brand.

Request Callback