The Challenge
A national garden centre chain had built a robust loyalty database of over 800,000 members. However, communication relied on high-volume weekly emails built on broad assumptions such as season, region, or a recent promotion rather than on how members actually shopped.
Open rates had been falling for two years and opt-outs were climbing, and the CRM team suspected their most valuable members were developing inbox blindness because of the continuous stream of offers with no connection to their interests. While loyalty, transactions, and website data existed in good shape, none of it informed who received which email.
The loyalty base was segmented using real purchasing behaviour to rebuild the contact plan around each group’s specific preferences.
The Solution
Because transactions, loyalty records, and website activity were already linked via unique member IDs, the audit quickly confirmed three years of clean history to support behavioural clustering.
Members were grouped on purchase frequency, seasonal timing, basket composition, and category mix, with six distinct audience emerging: year-round hobbyists, spring project shoppers, gift buyers, plant specialists, outdoor living spenders, and a large group of lapsing regulars. Each showed distinct buying rhythms, with some concentrating almost all spend into key spring weekends while others shopped steadily through the year.
The sprawling campaign calendar was consolidated around those six audiences, so each segment received fewer, more relevant emails aligned with their natural trading patterns. The segments were loaded into the client’s campaign platform, supported by a quarterly refresh to track customer migration over time.
The Impact
Matching contact frequency and content to each segment’s natural rhythm lifted CRM-attributed revenue by 18% within the first 12 months of deployment, with the strongest commercial gains among year-round hobbyists whose inboxes had previously carried the same offers as members shopping twice a year.
The model also uncovered significant hidden value in the lapsing regular cohort. A dedicated win-back programme built around their historical category preferences reactivated 12% of the dormant group, a groups that months of standard discounting failed to move.
Crucially, reducing campaign volume for low-frequency shoppers protected long-term database health, driving a 25% drop in opt-outs and easing overall complaint volumes by 13%.
Key Takeaways
- Behavioural Clustering - Replacing broad seasonal assumptions with frequency and basket clustering revealed six distinct customer trading patterns.
- Optimised Contact Cadence - Consolidating campaign calendar around segment rhythms outperformed high-volume broadcasts, protected database health and improved active commercial response.
- Targeted Reactivation - Isolating a clear cohort of lapsing regulars allowed for a tailored win-back strategy that successfully recovered 12% of dormant members.
- Segment Migration Tracking - A quarterly refresh monitors how members move between groups, keeping the contact plan aligned with current behaviour.
- Overall Business Results - CRM revenue up 18%, opt-outs down by a quarter, overall complaints reduced by 13%.
Tools and Techniques
- Customer segmentation
- Exploratory data analysis
- Hierarchical clustering (Ward’s method)
- SQL - data extraction and transformation
- R - data handling and modelling
- Klaviyo - campaign activation
Related services
No matching items