A UK high-street fashion retailer specialising in men’s formalwear.
A UK men’s formalwear retailer was looking for innovative ways to drive revenue growth from its existing online customer base. Analysis of customer behaviour showed a large proportion of customers buying similar products year after year. However customer research suggested regular changes to the range assortment (SKU names and descriptions) was creating friction when a customer wanted to repurchase when reactivated by email.
While partially personalised, specific products included in emails were not specifically related to prior purchases. In a nutshell, if a customer had previous resulting in 5+ clicks needed for the customer to find their preferred product. The client needed a way to remove this obstacle and engage customers with fresh style of content to drive engagement and conversion.
Our customer behaviour analysis showed that more than 80% Subsequent Time Buyers (STBs) were interested in buying a similar shirt (size, colour, style) to their first purchase. However due to seasonal stock changes, most customers took more than 5 clicks (once on the website) to re-purchase.
With an aim to reduce re-purchase friction for STBs, we developed a recommendation algorithm to recommend SKUs to each unique customer based on their previous preferences e.g. size, colour, fit and style. Our solution was fully integrated with the client’s existing CRM, DotDigital.
Recommendations were further targeted at customers who our algorithm assessed to be most likely to buy habitually.
As a result, each email for a STB is now fully personalised with at least 50% of the creative directly related to the exact prior purchase.
We then automated the model by combining data feeds such as transaction log, supply levels and CRM data. This was embedded into the client’s existing CRM. We also supported in the design of marketing content to best leverage the recommendations, which were served to customers via email.
DotDigital / DotMailer CRM
Microsoft Dynamics CRM
Advanced SQL
Python (for initial data analysis and hypothesis generation)
Personalised product recommendations are now being regularly sent to >100k customers using fully automated data pipelines. These emails have open-rates close to double that of non-personalised emails, and drive a 25% higher conversion rate as a result of reduced friction.
There are plenty of off-the-shelf product recommendations however these are no substitute for recommendations workflows that are intuitively based on your customer behaviours.
What’s essential for effective personalisation is fully understanding how your customer behave, what frictions are holding back demand, where you can encourage repeat purchases. In our experience, there’s no shortcut to this despite many CRM solutions claiming they can automate personalisation.
In this project we successfully incorporated customer behaviour insight (customers prefer to buy a similar colour / fit / size as their initial purchase and they struggle to find this given range changes) into a straightforward analytical solution that fully integrated with existing IT and CRMs.