MandM - Leveraging Sophisticated Machine Learning Models to Support Business Growth in International
Team Members:
- Ben Powis, Head of Data Science
- Joel Lenden, Data Scientist
- David Hool, Analytics Lead
- Ben Nichols, Data Scientist
Country: United Kingdom
Organization: MandM
MandM is an online discount fashion retailer. It is one of the largest online fashion retailers in the UK. They sell highly discounted products from many of the world's biggest brands direct to customers throughout Europe.
Awards Categories:
- Best MLOps Use Case
Business Challenge:
MandM is a British fashion retailer that exclusively sells products online at discounted prices. While we are headquartered in the UK, we operate seven localized versions of the MandM website throughout Europe.
Often our models are developed initially in a single market, and if successful, we would look to roll them out across Europe. However, we understand that a model built for a UK customer base may not be suitable, without changes, to deploy in a mainland European market.
We wanted to leverage the power of Dataiku and support the growth goals of the business by utilizing the project and deployment architecture of DSS. This would allow us to quickly roll out our many versions of models for international markets at a speed and scale that was not possible before.
Business Solution:
The project structure of Dataiku gives us multiple ways to package up our machine learning projects for success. We found the solution that worked best for us was to have all international versions of a model within a single project, separated into different flow zones and with their own separate scenarios.
The first steps of implementation involved copying the existing project flow, then taking each copy in turn and optimizing it for a particular market. Some markets have less or more data points, so thresholds and features may need to be adjusted.
Secondly, the labs feature of Dataiku allows us, within a single project, to train multiple models. Once we had built and optimized a model for each market, we were able to then easily drop these into the appropriate flows.
We were aware that adding many more models to our workflow could add additional data engineering dependency, so we wanted the models to be as self-sufficient as possible. To that end, once all flows were set up, we also rolled out a sophisticated drift monitoring solution across each model. This allows every model to test itself every time it runs and retrain on fresh data should this be required. This keeps operational costs down and human intervention at a minimum while maintaining highly accurate models.
Day-to-day Change:
Being able to support all of MandM's international markets with the same sophisticated models gives our colleagues a rich toolset of machine learning outputs which they can leverage for marketing personalization and customer experience.
The speed at which we are able to do this, thanks to the tools within Dataiku, is a huge benefit for us. To rebuild and deploy a model outside of Dataiku could take weeks, but with our current solution, this has been reduced to days, and working on multiple versions of the same model simultaneously reduced this even further.
Finally, having MLops around all these deployed models (alerting, data-driven training and predictions, drift monitoring) allows our data science team to focus on building new projects in the confidence that these models can look after themselves.
Business Areas Enhanced:
- Analytics
- Marketing/Sales/Customer Relationship Management
- Product & Service Development
Use Case Stage: In Production
Value Generated:
The value generated for us with MLops is in the time saved in data science hours to get a product into a production state, but also in the incremental gains from getting a solution into market quickly. This varies depending on the project, but having a model out in the world adding value to the business is a clear win for the team.
Value Brought by Dataiku:
Dataiku makes deployment quick and easy - we can put a project live in just a few clicks using our automation node. Once live, we can also use the built-in drift monitoring plugin to set out models up for success by being dependable and available whenever they are required.
Value Type:
- Improve customer/employee satisfaction
- Increase revenue
- Reduce cost
- Save time