Smarter segmentation to improve the customer experience in retail
Context & Objectives
A growing Belgian food retailer was facing heavy competition in the organic retail sector. They turned to Agilytic to use segmentation to increase customers' average spending, grow customer loyalty, and acquire new clients.
Obtaining more accurate customer insight was crucial for the retailer to allocate acquisition and retention resources better.
Therefore, we set out to perform a behavioral segmentation of customers to understand their purchase habits better.
Then, we identified the differences in behavior between shops to adapt the marketing and sales approach locally.
Finally, we developed a dashboard so that the retailer can explore and dynamically benefit from the analysis.
Approach
First, we began with data and scope validation. This helped us build descriptive analysis and validate the scope of clients, data, and transactions with the team (e.g., behavior per shop, the volume of unknown customers, and products).
Then, we performed data consolidation, which comprises building variables for segmentation and description of segments. In this project, we used quantities per type of product for the segmentation.
Next came the segmentation. We used an algorithm (silhouette score) to find the optimal number of clusters and performed a segmentation based on the results. This method involved a mix of “business” and “machine learning” (ML) segmentation: ML-algorithm used for segmentation based on variables that were selected beforehand.
In this project, the number of optimal clusters was ten. As ten clusters might be challenging to manage in marketing actions, we suggested two segmentations—the first with ten segments and the second one with five clusters. We developed a report on the segments to validate the segments with the retailer.
After this segmentation phase, we moved on to visualization. The client planned to use and build their own dashboard in the long term. So, we developed a temporary solution as a dashboard in Power BI. We first created a mock-up to validate the dimensions, metrics, and type of visualization the client needed before developing the dashboard in Power BI.
Finally, we delivered documentation to transfer knowledge to the retailer and help them take ownership of our model. We wrote documentation to describe the approach and scripts used to build the segmentation. The documentation also included the procedure in applying the segmentation on new data.
Additional Post-Analysis
We identified critical opportunities for marketing campaigns following the COVID crisis based on previous analysis. We did this during the first lockdown and after the first lockdown and considered metrics such as the shop traffic, evolution in customer segments, customers who left, and new clients. This resulted in actionable, time-sensitive, effective marketing campaigns for the client.
Results
At the end of the project, we delivered:
Consolidated and enriched dataset with segments and calculated fields to build the segmentation
Segmentation report with the approach and description of segments
Dynamic dashboard showing the behaviors globally, per segment, and/or per shop
We developed scripts to build the segments, apply the segmentation to new data and build the dataset for the dashboard
Documentation of the project
A detailed report on the analysis of customers' behavior before, during and after the first lockdown
A list of recommendations to test, based on the types of products to promote to specific customers / segments
Our model allowed the client to use the segments to inform future marketing campaigns.
Thanks to the data analysis and segmentation, the client can now see improvements in customer loyalty due to making smarter decisions around coordinating commercialization efforts. After completing the project in two weeks, the client quickly understood the necessary subsequent actions to increase its customer retention and commercial potential.
