Being a Data Science Intern at FreeAgent: The first month

Posted by on July 10, 2019

As the end of my second year at university was approaching I knew I needed to get some real-world experience to help me get a clearer vision of what career path I want to follow in the future. The best way of achieving this was getting a summer internship. After reading some blog posts of past Interns at FreeAgent (who knew I would be writing one myself a few months later!), the company stood out to me among the plethora of opportunities on offer (which Edinburgh has a lot of with its thriving tech scene). So, without hesitation I submitted my CV, put together a short cover letter and applied. The application process was very straightforward. All it took was solving a short practical problem and then having a video interview with Dave and Owen.

Working at FreeAgent

I was part of the largest ever batch of new joiners to join the company. 13 new people joined the company on that day! I was impressed with the friendly and laid-back atmosphere at FreeAgent, which immediately made me feel welcome and part of the company. After attending some inductions in the morning, I went upstairs to my desk and met the team I would be spending the summer with and was shown around FreeAgent’s stunning office which boasts magnificent views of the castle and Edinburgh’s city centre. Most importantly, Dave made sure I knew where the biscuit cupboard is!


FreeAgent’s office is located right next to the Union Canal

Getting Analytical

The first few days were spent getting familiar with all the different tools used by FreeAgent and the analytics team. A working day at the analytics team starts with a stand-up meeting – a brief gathering of the team where the plan for the day is set. The meetings are informal but productive and everyone shows genuine interest to your opinions, even as an intern. This is not your typical internship where you just spend your day completing mundane tasks and working on unimportant projects that everyone will forget shortly after their completion. The projects Lea (the other Data Science Intern) and I have been assigned are projects that will have a real impact in the company. I was surprised by how many people across the company were genuinely interested in the outcomes of our projects and by how much feedback they provided.

Predicting Subscriptions

My project is to setup machine learning to calculate the probability of subscription of a customer after the end of the 30-day free trial. Before starting the project I met with various stakeholders to get a clear idea of what they saw as a successful project outcome. The comms team run a range of different advertising campaigns (for example through Google Ads). Currently the ways they can assess the performance of a campaign are either through the number of sign-ups (however, not all sign-ups will lead to subscriptions) or wait 45 days to measure the number of companies which subscribed.

I’m using data from the first 3 days of the free trial to train a machine learning model to calculate the probability of subscription, which will then be used to estimate the conversion rate yielded by a certain advertising campaign – it will help us better understand our customers.

Among the different parameters that influence the probability of subscription there were two that stood out:

  • The greater the range of features a customer interacts with during the first 3 days, the greater the probability of subscription. This shows that customers who explore the FreeAgent application more, can see more of its benefits which increases the likelihood they will convert to paid subscribers. 
  • Customers who have any interaction with the support team are significantly more likely to subscribe compared to those who don’t, which shows the positive impact of providing good support to customers. The support team are the face (or maybe, the sound) of FreeAgent.

In the coming weeks I will work on building a pipeline to automate the training – predicting process. I will also experiment with Unsupervised Learning algorithms for customer segmentation. The results of the segmentation can then be used as predictor variables in the subscription prediction model.

FreeAgent holds weekly Town Halls where people from across the company enthusiastically present in front of everyone what they are working on. There are also weekly Engineering Forums which are similar to the Town Halls but are more focused on the technical side of things. By the end of my 3rd week I had already presented the work I had done in both the Town Hall and the Engineering Forum. It was a great experience to show people what I had been up to early on, answer questions and receive feedback.

Leave a reply

Your email address will not be published. Required fields are marked *