Case Studies

Haymarket Automotive

Haymarket Automotive are the publishers of WhatCar and AutoCar magazines (amongst others).

The readers of these websites divide into two rough categories:

  1. People who just love cars and who look at the site for entertainment and to stay up to date with the industry.
  2. People who are in the process of buying a car.

How these two types of user browse the website and the functionality they need is very different.

Haymarket approached me to build a machine learning model that would classify people into one of these two categories so that this information would be available in Google Analytics for their team to use.


We ran a survey on the site and linked the responses with the user’s Google Analytics client ID. From the responses it was easy to classify a person into one of the two categories and, because of the link with the client ID, we could see all their other activity on the site. This was the training data and the challenge was to find a model that could use this data to correctly predict the user category at the end.


We found a model that was able to identify people in the “ready to buy” segment with over 80% accuracy. This model predictions were added to Google Analytics as a custom dimension using the client ID. Then other analysts in the Haymarket team could use it in their work.

This enabled the team to better identify which content performed best for the different audiences.

There were some A/B tests where the overall winner was not the same as the winner among the “ready to buy” segment. This was very important when the feature being tested was intended for this specific audience.

The Key

The Key is a small charity based in Newcastle. The Key targets their work at disadvantaged young people and aims to help them reach their potential by helping them to do something intrinsically motivating and meaningful for them. The things the young people do vary a lot; The Key provides a very flexible framework to help them with this.

As part of this process The Key collects some information on each young person and the projects they are involved with. They asked me to work with them as a part time Director of Impact to use this data to improve their operations.

The Key was one of my first clients and there is a lot we have done together over the years!

A big theme has been democratising the use of data within the organisation to enable everyone to make more data driven decisions. To start with I was the only person who knew SQL so I was the only person who could access anything more than the most basic figures from the database; this was a major blocker for the rest of the team because I was not working full time.

I built an internal website for the team where I can quickly build out new visualisations and reports. Anything that gets asked for more than a few times ends up as a page in this website where it becomes available to the whole organisations with daily updates so it always shows the latest figures.

There has also been a lot of more specialised analyses done to try to figure out what sort of young person The Key works best with, where they are located and what is the best route to reaching them.

The Key has been a very early (and rapid) adopter of data use within the charitable sector. This is a major competitive advantage for them when it comes to building relationships with funders; The Key are better able to forecast how well or how quickly they can distribute funds and, afterwards, they can offer more granular and accurate reports. These are both things that funders love!

A long the way we’ve also had a lot of fun; such as when I made this “noisy map” for the CEO to present at a conference. It shows the location of projects over the years with a different noise for different types of project.