Data Discussions meets LBAG: Moneysupermarket’s Head of Data Strategy explains Machine Learning

Cem Baris 30.11.2016

Data Discussions is excited to announce a new partnership with the London Business Analytics Group.

Some of you may have seen a recent blog of mine about the launch of Data Discussions and our events. We are now very excited to announce the partnership between Data Discussions and The London Business Analytics Group. Having been fortunate enough to co-found LBAG alongside Mark Wilcock, I jumped at the chance to be involved in this again, as soon as the right opportunity came up. 

Data Discussions meets LBAG: Moneysupermarket.com

To see a group develop from what initially was just an idea to a community of over 4,100 members with a strong presence at each of our events, has been an absolute pleasure! Those of you who have attended some of our events, I’m sure have heard Mark and I joke about the few initial conferences and we make no secret of the fact that developing the group has taken a lot of commitment and hard work. We celebrated each milestone, first our 100 members (which now feels so far away), then our 1,000 members and once we reached 4,000 members, I think even we were surprised by the great interest in our events! 

I get asked all the time: What it is that makes this group so special? What makes it stand out from the 40 odd other data groups available on meetup.com? The answer is simple; our speakers, as well as the community we created! 

On Monday, we welcomed Harvinder Atwal, Head of Data Strategy and Advanced Analytics at Moneysupermarket.com, who spoke to us about Machine Learning, challenges, solutions and experiences. I don’t think Moneysupermarket needs much of an introduction, but for those who are a little confused: as Harvinder said, they’re "not the meerkat guys, so don’t call asking whether Oleg or Alexander are on their way!" You will, however, recognise Moneysupermarket.com from their memorable adverts, such as the ones below;

Moneysupermarket adverts

Apart from their clearly brilliant advertising team, Moneysupermarket.com are actually Google’s biggest paid search customer, well most year round, apart from the last few weeks of the year when Amazon take over for some reason. Having had a van full of Prime (Christmas) deliveries yesterday, I may be part of the issue (sorry Harvinder!)

Moneysupermarket is known to save its customers £1,000s whether on insurance, utilities or money products. Most, if not all now reach out to a price comparison site to make sure they’re getting the best deal, so you can only imagine the amount of traffic on these websites, let alone the amount of data that they obtain. In fact 19 million adults chose to share their data with Moneysupermarket.com and it is here that Machine Learning comes into play.

Andrew Ng explains Machine Learning to be “The Science of getting computers to act without being explicitly programmed”, or otherwise known as “Regular programming flipped on its head” (as described by Harvinder). The speaker introduced us to different classes of Machine Learning algorithms, including unsupervised clustering, supervised regression, supervised classification and supervised learning. He then gave some examples, such as Moneysupermarket’s use of supervised regression in their “home bill checker”, where they reduced the amount of information needed from their consumer from 70 to just one field in order to produce very similar and accurate data, simply from a post code.

Harvinder Atwal


Harvinder gave some real life examples, sharing his experiences, and it immediately became transparent that whilst some of it  may sound easy, the journey to implement a functional machine learning project is far from that! Issues such as messy data (often also meaning messy code), silos and lack of real understanding from the business are just some of the challenges needed to be overcome. 

Below are some of the hot tips Harvinder would give to someone beginning the Machine Learning journey within their organisation: 

  • Be prepared to manage change - your model may need continuous adaptations
  • Beware of feedback loops
  • Monitor and measure the model’s performance
  • Plan - clear and aggregate your data in advance
  • Develop a data lake
  • Educate the business
  • Do not assume anything...ever! 
  • Document and log


For further tips and information, click here for the full conference, or if you've any questions for Harvinder, click here to email him.

Finally, I wanted to say thank you for CodeNode for allowing the use of their venue, Mark Wilcock, for his continuous involvement in making these events happen, Harvinder, for kindly sharing his views and of course the Data community for your support and interest. 

I hope to see you at our next event!

Cem Baris's picture
Director
cbaris@morganmckinley.com

LATEST JOB VACANCIES

Application architect, Java, C++
London21.02.2020
Analyst architect, Workshops, Architect, Developer
London21.02.2020
Regulatory reporting, Architecture, Java, C++
London21.02.2020