Optimizing marketers' decisions with Causal Inference
Project created with Worthy.com in order to improve their revenue cycle
Worthy.com is an online marketplace for pre-owned jewelry, connecting sellers with potential buyers. After the sellers (AKA leads) register their items online, they are asked to ship their valuables to Worthy offices for grading and evaluation.
What is Causality?
A causal approach refers to a method used to understand and explain the relationships between causes and effects, focusing on identifying the factors that directly influence or cause specific events.
1. Improving AE's decision-making
A common sales strategy involves having account executives (AE) actively reach out and contact potential customers. However, not all contact attempts have a positive effect: some attempts don’t change customer decisions, while others might even interfere with the desired outcome.
2. Understanding Marketing personas
The Uplift Modeling divides the different types of customers into four categories:
Sure things, Persuadables, Lost causes, and Sleeping Dogs
Once we understand better our customers’ potential for purchase, it is easier for AEs to make decisions which leads to approach. Together with VIANAI's causality-based approach, we intend to optimize their revenue by building a policy to generate recommendations for their decisions.
Solution: personalized policy
We generated a personalized policy and recommended reaching out only to customers for whom it would be beneficial.
We created a platform for Worthy, containing all the data related to VIANAI’s recommended policy, as well as a monitoring tool in which they can track the performance of the model and take action if needed. The two main screens are the REVIEW and OVERVIEW.
A successful outcome of the AE’s communication policy is the delivery of the item to the company’s offices within fourteen days of the last call.
Once the process of generating the policy is done, the user can consume all the data regarding how to optimize their KPI, with data visualization of comparisons between the current policy (based on past data), and our policy. We demonstrate how using our policy will affect the business cycle, and offer more details about how our model is working behind the scenes.
This is a daily monitoring page that keeps track of the model runs. The user can get updates about the KPI’s performances in the last 6 months, including the average compliance rate with the policy by AEs.
This work demonstrates the value of a causality-based framework for optimizing business decisions. We validated the results in an A/B test over a 3-month period, in which 50% of the AE’s used the Worthy policy, and the other half used the VIANAI policy. Results revealed a 22% increase in item delivery rates for the targeted population.
The success of the experiment, led to its adoption as the standard practice today, meaning that Worthy has replaced its current policy with the VIANAI’s policy.
Article by Naama Parush, Ohad Levinkron-Fisch, Hanan Shteingart, Amir Bar Sela, Amir Zilberman, Jake Klein. (https://arxiv.org/abs/2207.01722)
Designed in collaboration with Shirily Bar-or