Calibrating Attribution with Incrementality Experiments for Optimal Budget Allocation

Key Takeaways

• Traditional click-based attribution often misses the true impact of marketing
• Calibrating attribution models with the results of rigorous experiments that measure incrementality can improve the reliability of your marketing measurement
• Creating a multiplier to adjust your attribution results is a practical example of something businesses can start doing today to ensure accuracy

Effective marketing measurement is more complex—and more critical—than ever. Traditional methods like click-based attribution often miss the mark, especially as consumers engage with formats like video that don’t generate clicks. Gen Z, for example, is twice as likely to buy without clicking.¹

Many businesses try to solve these challenges by using a variety of different measurement solutions. In fact, a recent Kantar study commissioned by Meta found that businesses use nearly four different measurement solutions on average.² This “suite of truth” approach enables marketers to use different measurement tools based on their needs—for example, attribution models for day-to-day decisions—and periodically adjust or refine results with rigorous experiments to ensure accuracy. This process, known as calibration, helps to improve the overall reliability of marketing measurement.

Why incrementality matters

Meta VP of Analytics and CMO Alex Schultz says it best in his new book, Click Here: The Art and Science of Digital Marketing and Advertising: “Incrementality is everything.” Incrementality reveals the true impact of your marketing, helping you invest in channels and tactics that actually drive results.

The gold standard for measuring incrementality is through an experiment, such as Conversion Lift. By comparing a test group (who can see your ads) to a control group (who cannot), you can isolate the real effect of your marketing. Businesses using these types of experiments see ROI increases of 20% or more.³

While you can’t always run experiments, doing so regularly and calibrating your other measurement models helps you to make informed decisions without needing to experiment at full scale all the time, saving on costs.

Calibrating click-based MTA with incrementality

For marketers who are interested in leveraging incrementality to improve the overall rigor and accuracy of their marketing measurement, there are a few different ways to proceed. Click-based multi-touch attribution (MTA) is a common tool for making daily decisions, thanks to its quick, always-on capabilities. But in today’s landscape, clicks don’t always reflect true impact—meaning that MTA results can and should be calibrated with incrementality.

To calibrate attribution across channels, you can:

1. Run incrementality experiments on all platforms (e.g., Conversion Lift for Meta).
2. Compare experimental results to your attribution model to see how they align.
3. Test different attribution models (last-click, even credit, data-driven) and see which best matches incremental impact.
4. Adjust your attribution outputs using multipliers based on experiment results.
   

How to use a multiplier

For a practical example of something businesses can start doing today, let's look at multipliers. If you use multiple marketing channels and have the ability to run incrementality experiments on each one (for Meta, you would run a Conversion Lift test), this can be a great strategy. Here is how it might work: Coats & More, a fictional company, runs ads on Meta and several other channels. They measure marketing impact using a click-based MTA model, which showed:

• Meta: 100 conversions
• Platform X: 200 conversions
• Platform Y: 80 conversions
• Organic: 120 conversions

But a Conversion Lift test reveals Meta actually drove 150 incremental conversions for Coats & More—a 50% increase over what the attribution model showed. To align the attribution model with true impact, they would calculate a multiplier:

150 (Conversion Lift) / 100 (Attribution) = 1.5x

This multiplier is then applied to Meta’s results from the attribution model going forward. In this case, to adjust Meta’s results they would use the following: 100 × 1.5 = 150 conversions.

However, if the total conversions after applying the multiplier exceed actual total conversions, it’s best to repeat this process for other platforms. Run an experiment on Platform X and Y to get their multipliers and adjust accordingly. Any remaining conversions can be attributed to organic.

Remember: Multipliers are just one method for calibrating attribution models, and they should be updated periodically to stay accurate. There are also more advanced techniques available beyond multipliers; for example, you can use lift experiments to fine tune how much credit each channel receives in an MTA model. In an upcoming post, we will also be covering the basics of calibrating marketing mix models (MMM).

Make calibration a habit

Calibration isn’t a one-time fix—it’s an ongoing process. Regularly run experiments, compare results, and update your models. As conditions change, continuous calibration drives better efficiency, effectiveness, and growth.