Optimize your results & drive more impact with Actable AI

Explore tens of videos, examples, and articles to help you learn Actable AI, Data Science, AI, and much more.
Estimate causal effect of personalized promotions

An e-commerce company wants to understand the effectiveness of their promotion program and is unsure whether they should continue it. In the past, the company gave out discounts to different customers. Historically each customer is assigned a purchase power index and more discount was given to customers with higher purchase power index as they tend to buy more. It has a data set of historical records for each customer which contains the purchase power index, discount amount and profit generated from the given discount.

Simply doing a regression of profit on discount amount doesn't give us the correct effect of the discounts because profit is impacted by both discount and customer power index. One can do an A/B testing experiment by creating a new promotion program and give discounts to only a set of random people while keeping others as a control group. However doing so is time consuming and if giving a discount doesn't help, it might harm the business.

As we know that customer purchase index is the only factor that impacts both how much discount is given and how much profit is generated, our Causal Inference analysis can help estimate the effect of discount on profit controlled by customer purchase power without conducting an expensive A/B testing. After uploading the data set to our platform, you can simply select Causal Inference and set Outcome as profit, Treatment as discount amount and Common Causes as customer purchase power index. Click Run to run the analysis and wait for the result. You can also check out an example of the analysis here.

Figure 1: Causal graph indicaetes purchase power index (sale_prediction) has causal effect on both given discounts and profit.

The below figure shows the average treatment effect value which indicates how much profit changes if discount is increased by 1. As you can see that the result is -0.039 (95% confidence interval of [-0.011, -0.0676]), on average giving discount doesn't help increase profit but decrease it.

Figure 2: Average treatment effect (ATE) of discounts on profit
The example analysis shows how our Causal Inference can help easily estimate the effect of an intervention over an outcome with historical non-random data. It potentially can save the business a significant amount of time and money compared to A/B testings.
Actable AI
Actable AI Technologies LTD
UK Company Number: 12669336