Interpretable Price Elasticity Estimation
Understanding how price changes influence customer demand
The AI pricing revolution
In recent years, the rise of AI has revolutionized price-sensitivity estimation, moving the focus from expert-knowledge and common sense to data-driven approaches.
However, most demand estimation methods rely heavily on black-box algorithms that make it impossible for pricing experts to understand why a certain discount leads to a change in sales.
Estimating discount effects on car sales
By combining data on historical sales and pricing with product characteristics, we used Optimal Decision Trees to predict how customers would react to different pricing strategies. The resulting model performed similarly to the previous black-box methods, while remaining completely transparent to users.
Illustrative Optimal Tree predicting customer sensitivity to discounts
Turning data into trusted actions
-
Increased adoption rate
Business users gained trust in the predictions once they could understand the logic behind the models. We integrated interpretability directly into their workflow, with an intuitive dashboard where they can easily see the decision path leading to the predicted price sensitivity.
-
Faster model improvements
Interpreting the feature combinations that were being used in the predictions allowed data scientists to better understand the structure of the problem and to work efficiently to improve the models further.
Unique Advantage
Why is the Interpretable AI solution unique?
-
Auditable, robust and trusted
This framework allows for complex and accurate modeling of demand elasticity while maintaining interpretablility
-
Unlocks a culture of collaboration
Understanding the logic behind prediction enables collaboration between data scientists and domain experts