Elizabeth Moorcroft, a senior data scientist at Aviva, is working in the customer service team as part of the insurance company’s global data science practice, Quantum. Her daily tasks depend on the stage of the project she is working on at the time. Currently she is at the start of a new project and so is spending a lot of time collecting the data she needs, as well as investigating and testing different modelling techniques.
She finds that the beginning of a project is always fun because there are a lot of different aspects to explore. “The most important part is to understand the system I am modelling, so I’ll be going around all the relevant areas of the business, talking to people and collecting their knowledge,” she said. The next stage of the process will involve joining up the data from different areas of Aviva, cleaning that data and then another “fun” bit, which is diving into the modelling.
Moorcroft really values the opportunities she has been given to experiment at Aviva. Through Aviva’s data science training programme, Quantum University, the insurer is advancing the capabilities of its data scientists and encouraging them to investigate new techniques and processes.
A challenge that Moorcroft has had to overcome is not having the exact data that she wants; often it will be missing a variable or there is not enough volume and she will have to make small changes to compensate. This happened when she was doing her PhD on the movement of big cats, officially called ‘modelling biological complexity’. The location data she had been “kindly” given by other researchers had intervals of five hours instead of the couple of seconds that she needed.
“It was collected for a completely different purpose which meant the structure was wrong for my initial way of thinking,” said Moorcroft. It was not possible to collect her own data so she made some tweaks and managed her expectations.
"Sometimes good enough is an acceptable outcome."
“I ended up changing the methodology entirely, and made a lot of assumptions that I would have rather not have made. In the end, the research wasn’t as good as it could have been with a different data source, but sometimes you have to accept that not everything can be perfect and that good enough is an acceptable outcome,” she said.
Incidentally, Moorcroft was drawn to study analytical subjects because she thought that a job that required those skills wouldn’t involve a lot of writing. Her dyslexia, which wasn’t diagnosed until later on, meant she wasn’t confident in her writing abilities. The upside was she was very competitive and independent at school and became an expert at solving problems. She would memorise texts to avoid making mistakes while reading aloud, and rewrite entire paragraphs to avoid misspelling a single word.
"Being great at problem solving makes me good with data."
“Because I spent so much time thinking about how to avoid my difficulties, I became great at problem solving. And being great at problem solving makes me good with data, and putting hacks in code so that things work even if it’s not the most straight forward thing to do.”
In spite of her independent nature, Moorcroft has benefited from encouragement offered by others in the data community, both virtually and in person. All her managers have pushed her to pursue any opportunities that have arisen and she has been able to draw on support from online data hubs. “The online community for data is massive. There are forums for any problem, and everyone wants to help.” She cited the blog on R programming by Dr Kasia Kulma as a particularly helpful online resource.
“It’s people like Kasia who are the real heroes of data science. Even after a long day at work, she continues to write helpful articles and teach other people.”