The web retailer had a large email list built up through strategic media partnerships. A recent advertising campaign using the Facebook ‘Lookalike’ feature provided very minimal returns. The retailer leveraged SherlockML to understand their customer base and transform their intuitive marketing strategies to a data-driven approach.


Secure

Secure

Specific roles and granular access controls

The retailer created a project and uploaded their data securely onto SherlockML. They could then easily collaborate by inviting colleagues to their team, each with their own specific role and granular access settings. Even reporting to other stakeholders was done securely through SherlockML, avoiding the need for sensitive data being distributed widely. Conducting the entire project on SherlockML gave the retailer a full audit and a centralised place to manage permissions and security.

Scalable

Scalable

Comes with the latest open data released by the UK government

After the data was uploaded it was run through Lens, which automatically explores the quality of datasets. The retailer discovered that their customer dataset was relatively clean, but sparse.

SherlockML comes pre-installed with the latest open source data released by the UK government. The merged open data and customer postcodes were used, together with a custom algorithm, to build a model to predict the customer demographics. This was used for optimising acquisition channels and segmenting customers.

Powerful

Powerful

Data too large to be handled by a laptop

The machine learning models required different levels of computational power to run. Training a model required more computational power than exploring the data. The dataset was also too large to be handled comfortably by a laptop. SherlockML allowed the team to scale computational resources depending on the needs of the project, and to iterate quickly.

Model of the UK for one of the retailer's products. We see that marketing should be shifted from central London to the Thames estuary for greater impact (as shown by the darker shades).

Impact

3x more effective targeting

  • Increased email open rates
  • Better marketing campaigns

The new Facebook ‘Lookalike’ targeting was on average three times more effective than previous results. The modelling was also very effective for email marketing. In a test of the segmentation, the customers placed in the high likelihood category were ten times more likely to purchase than the low likelihood category. By switching to a more personalised message for each segment, email open rates increased by 50%.

Power your analysis with SherlockML

Subscribe to our Newsletter

Receive our curated publication of data science content.
Subscribe below and never miss an issue.