MandM: Lightning Fast Value From Data Science With Dataiku

The accelerated growth of MandM in 2020 meant more customers and, in turn, more data. MandM’s rapid growth resulted in two big challenges:

1. Getting all the available data out of silos and into a unified, analytics-ready environment
2. Scaling out AI deployment in a traceable, transparent, and collaborative manner.

MandM’s first machine learning (ML) models were written in Python and run on data scientists’ local machines, and they needed a way to prevent interruptions or failure of the ML deployments.

In an attempt to tackle the second challenge, the team moved these Python files to Google Cloud Platform (GCP). However, once the number of models in production went from one to three and more, the team quickly realized the burden involved in maintaining models. There were too many disconnected datasets and Python files running on the virtual machine, and the team had no way to check or stop the ML pipeline. They needed another solution.

Exploring Data Science at Scale With Dataiku + GCP 

MandM turned to the powerful combination of Dataiku and GCP to answer their two critical challenges. With Google BigQuery’s fully-managed, serverless data warehouse, MandM could break the data silos and democratize data access across teams. At the same time, thanks to Dataiku’s visual and collaborative interface for data pipelining, data preparation, model training, and MLOps, MandM could also easily scale out their models in production without failure or interruptions in a transparent and traceable way.

MandM now has hundreds of live models, all with visibility into model performance metrics, clear separation of design and production environments, and many more MLOps capabilities built into the Dataiku platform.

Teams can now easily push-down and offload computations for both data preparation and ML to GCP. Using Dataiku means this capability is accessible to all user profiles across MandM, without knowing the underlying technologies or complexity.

The team is also particularly proud of the work they’ve done to build out a feature library with Dataiku that contains more than 400 features specific to MandM’s business. Now, the feature library is the first place people go, sort of like a shop window for ML projects  — it takes away the monotony and repetition of their work.

Having a platform like Dataiku allows our data scientists to focus on building cool things, not spending hours and hours on maintenance and making sure things are running. With workflows deployed in Dataiku, we save literally days of work every month. Ben Powis Head of Data Science at MandM

Business Use Cases at MandM

The benefits MandM have seen by using Dataiku and GCP aren’t limited to time saved from tedious maintenance work — they are also having more impact across the business. The data team is now able to deliver a variety of business solutions on problems from adtech to customer lifetime value, whether that’s a dashboard, a more detailed piece of analysis, or an ML project deployed in production.

Using Customer Lifetime Value to Deliver Personalized Experiences

MandM started their journey into lifetime value by developing their own custom Python model, which the team set up in Dataiku. Shortly after MandM got this up and running, they came across the Dataiku Solution for Customer Lifetime Value Forecasting. MandM, in their words, “jumped in and installed this in minutes.”

The data scientists at MandM were impressed with many elements of the Dataiku Solution and, having access to all the code in Dataiku, decided to create a hybrid model, combining the best practices and Dataiku-specific implementations of the pre-built project with some of their own specific changes that fit MandM as a business.

The resulting project was far stronger having been through this process, and was immediately ready to package up and deploy to their automation node, where it now scores millions of MandM customers every single day.

Broadly, we love Dataiku. We do have a mix of people that go more toward AutoML and visual tools as well as one data scientist who loves to work in code. But that’s the beauty of Dataiku and why we chose it — we didn’t want a low-code tool where we could get lazy and just click a few buttons. Now the team has the flexibility: If they want to nerd out and go under the hood, they can do that. If they need a quick model, they can do that too. Ben Powis Head of Data Science at MandM

Leveraging ML to Support Business Growth in International Markets

While headquartered in the U.K., MandM operates seven localized versions of the MandM website throughout Europe. Often the team develops models in a single market and, if successful, they investigate rolling them out across Europe. However, a model built for a U.K. customer base may not be suitable, without changes, to deploy in a mainland European market.

MandM wanted to leverage the power of Dataiku and support the growth goals of the business by using the platform’s project and deployment architecture to quickly roll out many versions of models for international markets at unprecedented speed and scale.

The project structure of Dataiku gave the team multiple ways to package up their ML projects for success. Ultimately, they decided to have all international versions of a model within a single Dataiku project, separated into different Flow zones with their own separate scenarios.

In addition, Dataiku allowed the team to train multiple models within a single project. So once they had built and optimized a model for each market, the MandM team could easily drop these into the appropriate flows.

The team was aware that adding many more models to their workflow could add additional data engineering dependency, so they also wanted the models to be as self-sufficient as possible. To that end, once all Flows were set up, they 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 if needed, ultimately keeping operational costs down and human intervention at a minimum while maintaining highly accurate models.

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.