This case study focuses on the client company’s operations data and tells how shopper churn is contributing to the department’s strategic decisions. The grocery giant has a platform for on-demand grocery service. As a shopper, you’re the one that’s shopping for the customer’s order and then delivering it to the customer’s doorsteps. Fundamentally, shopper churn analysis provides a report to find an active and available workforce to deliver the goods. However, up until this point, we have only examined shopper churn via the prism of time analysis to obtain a figure that represents the proportion of once-active shoppers who are no longer active. In addition, no root cause analysis has been performed to determine why churn occurs (outside the churn survey), including,
- What categories of shoppers are most likely to churn? Is it more prevalent in part-time jobs, given Male/Female, Experienced/Rookie, etc. make up 70% of our staff.)
- Is churn seasonal?
- Is there a method to address the widespread churn-causing experiences that customers have in order to boost retention?
This case study intends to offer you an overview of the client ecosystem and how Factspan assisted in providing analysis for productive business results.
Business Challenge
A major grocery retailer sought to comprehend shopper attrition in order to control supply and demand imbalances, reduce costs related to shopper churn, and perhaps even improve retention by anticipating and preventing shopper churn. In the past, churn was indicated by a shopper’s failure to take up order after 45 days.
Solution
The team at Factspan handled the issue by defining churn and employing the XGBoost algorithm to identify potential churning shoppers. XG Boost algorithm aids in identifying Churn patterns and preventing them.
Utilizing cost of acquisition analysis allowed businesses to create more effective incentive programs and strategies for customer retention. The team used the conventional term of 30/45/60 days to define shopper churn. The data analyzed was from (June 2020 to July 2020), and the following two scenarios were taken into consideration by our team:
- Shopper’s who didn’t deliver in June: The active shoppers made deliveries in the Active Period prior to June 1 but did not do so again until June 30. And those customers will be regarded as churned if the gap between the final delivery made during the active period and the month’s end (June 30) is higher than 30/45/60 days.
- Shopper’s who delivered in June: The active shoppers who made a delivery in June but had a gap of more than 30/45/60 days between their last delivery during the Active Period and their first delivery in June would be deemed to be Churned.
So total churned shoppers would be the summation of all the Shoppers who fall under either of the conditions mentioned above.
Churn Rate = (Total Churned Shoppers / Total Active Shoppers)*100
As the activity period and churn period are established. Factspan’s team established all the KPIs (features) and generated them from the Snowflake database during the subsequent model-building phase.
Interested in reading: Conjoint Analysis: A Cheat Sheet To Know Customer Preferences
Results
After that, the team continued with the routine steps of cleaning and preparing the data, engineering features, exploring the data, modeling, setting up the model run cadence, and saving predictions. Furthermore, univariate analysis was performed on various features:
Education
Observation & Results –
- The majority of the shoppers do Alcohol and RX certifications.
- 27% of all the shoppers who completed the alcohol certification have churned.
- 30% of the shoppers who completed RX Certification have been churned.
Time Range to achieve 10th delivery from 1st delivery
Observation & Results –
- While churned customers took 28 days to make their first purchase, retained customers took 26 days.
- When opposed to retaining customers, who make their tenth purchase after their first purchase on average in 21 days, churned shoppers take longer.
- It has been observed that churned shoppers take longer to get going and that their idleness persists.
Weekly Order Delivery
Observation & Results –
- When compared to customers that remain loyal, churn shoppers deliver fewer orders.
- Their order delivery has decreased during the past six weeks, going from 10 orders per week to 5 orders per week (from Churn Start date). That is a sign that they are producing a lot.
- When we notice a drop in their constancy, we can take action about Churn shoppers.
Shopper Cohorts
Observation & Results –
- Usually Power, Core, and Casual Shoppers tend to stay with the company and Dormant Shoppers Churn out the most.
- 91% of the Dormant Shoppers have Churned out.
- 98% of Power Shoppers are Retained, and 96% of Core shoppers are Retained.
- Dormant behavior leads to Churn
- Shoppers Churn Across Order Delivery Buckets
Observation & Results –
- 77% of Retained Shoppers have done > 100 Shops
- 73% of Churned Shoppers have done < 200 Shops
- Shoppers with 200+ lifetime shops are observed to better retain; frequent touchpoints with the shopper until then would help preserve the momentum.
Shopper Churn Across Payment
Observation & Results –
- Avg Tip Received by Churn Shoppers is Roughly $1 less than the tip received by Retained Shoppers.
- Avg Pay Per Order is $12.82 for both Churned and Retained Shoppers.
- Largely there is no difference in the avg pay received by Churn and Retain Shoppers.
Late Delivered Orders/Rescheduled Orders
Observation & Results –
- Churn Shoppers have slightly more Late Delivered Orders
- Retained Shoppers delivered more orders on time.
- Both Churned and Retained Shoppers have 3% — 4% of Reschedule orders.
Shopper Churn Across Order Categories
Observation & Results –
- Churned shoppers have delivered slightly more Envoy orders (5%), 2% more Platform orders, and 2% fewer Marketplace orders than Retained shoppers.
- Out of all the orders fulfilled by Retained Shoppers, 14% are envoy orders, 42% are marketplace orders, and 42% are platform orders.