Adtriba – future-proof makreting measurement & optimization

How to Measure: Vouchers

Vouchers have long been a staple in the world of marketing and promotion. They are a go-to way for many businesses to attract first-time buyers and influence the purchase decisions of price-sensitive customers. In fact, the first “attribution” analysis in history was carried out by placing different discount codes in different newspaper advertisements about 100 years ago.

However, in modern times, marketers have to be careful while evaluating and deploying vouchers to not be fooled by a seemingly strong voucher campaign where all it did was deliver high redemption rates and reduced margins on customers who would have bought anyway, which is also known as "cannibalization.”

Measuring the “true” incremental impact of vouchers can be a challenge. We have to be especially careful since often, data related to vouchers is only available for customers who redeemed a voucher, which means, they already converted, or in other words, bought something. This may introduce bias in the analysis that can lead to an overestimation of the voucher impact. 

Multi-Touch Attribution (MTA)

Voucher redemptions as a touchpoint can not be included in machine-learning-based MTA models that rely on predicting conversion. This is because they only exist, by definition, in converting journeys. Due to this fact, it is almost impossible to include vouchers in MTA. We might imagine a situation, where the customer clicks on a site that displays vouchers, but there is no purchase involved yet, but generally, would not advise using MTA as a primary method of evaluating voucher performance. 

It doesn’t hurt, though, to include this touchpoint for reporting and exploratory analysis in the customer journey after the attribution modeling has been performed. 

Marketing Mix Modeling (MMM)

To properly measure the impact of vouchers, we can integrate them into MMM. Marketing Mix Modeling is a statistical approach that quantifies the impact of various marketing inputs on sales and other performance indicators. To include vouchers in MMM, we can consider either voucher distribution or the size of the discount as an input variable.

Voucher distribution refers to the number of vouchers, or unique voucher codes, distributed across different channels, such as email, social media, or print. Including this information in the MMM can help determine the effectiveness of different voucher distribution channels and their relative impact on sales.

The size of the discount is another input variable that can be used to measure the effect of vouchers on sales. By analyzing the correlation between the size of the discount and sales, marketers can identify the optimal discount rate to maximize revenue. This essentially makes the model a price elasticity model. 

In both cases, the inclusion of these input variables in MMM will enable marketers to understand the overall effectiveness of their voucher campaigns and optimize their marketing strategies accordingly.

What should, however, not be included, is the number of redeemed vouchers or the summed cost of the discount given - as an input variable in MMM. This is because the number of redeemed vouchers is not a direct indicator of their impact on sales, as customers may have made the purchase even without the voucher, cannibalization as mentioned above. What it is though, is part of the overall number of purchases, and thus, indirectly, also revenue, and conversions, one of which likely be the target variable in the MMM. We would thus include part of the response as a predictor, which is a cardinal sin in modeling and will lead to a severe overestimation of the impact. To summarize, the following voucher-related KPIs can be used as input variables to MMM:

  • Number of vouchers distributed across different channels
  • Size of the discount offered in the voucher campaign

Redemption data can, however be used to calibrate/provide priors to the MMM by using the redemptions as the upper bound for the true incremental impact of the vouchers. 

Incrementality Testing for Voucher Impact

Another way to measure the impact of vouchers is through incrementality testing. Incrementality testing evaluates the incremental value generated by a specific marketing activity, such as voucher campaigns, by comparing the results of a treatment group and a control group.

To conduct an incrementality test, start by randomly dividing your customer base into two groups: the treatment group, which will receive the voucher, and the control group, which will not. Ensure that the groups are similar in terms of demographics, purchase history, and other relevant attributes to minimize any potential biases.

After running the voucher campaign for a set period, analyze the difference in sales/conversions or other KPIs between the two groups. The difference in performance between the treatment and control groups is the incremental impact of the vouchers.

By using incrementality testing, marketers can obtain a clear understanding of the true value of their voucher campaigns, separate from other marketing efforts. This information can then be used to optimize future voucher strategies and maximize return on investment.

Often, it is however not possible to divide the customer base randomly into test and control segments, due to a lack of user identification because of privacy restrictions. In this case, a quasi-experimental method, like a geo-test, is the weapon of choice. Here, vouchers are given only to a certain geographical subset, while similar regions do not receive vouchers. The conversions from the control regions are then used to forecast the conversions in the test region. The difference is the voucher impact. 


While measuring the impact of vouchers can be challenging, integrating them into MMM and using incrementality testing can provide valuable insights into their effectiveness. By understanding the true value of voucher campaigns, marketers can make data-driven decisions to optimize their marketing strategies and drive business growth.

Leverage unbiased marketing measurement with Adtriba.