2. Index
• About
• The Simply Business data journey
• Key concepts
• Learnings
• Examples
3. About Simply Business
• Largest UK business insurance provider
• Over 450,000 policy holders
• Using BML, tech and data to disrupt the business insurance market
• Acquired in 2016 (£120M) and again by Travellers in 2017 (£402M)
• #1 best company to work for in 2015 and 2016 among other awards
• Certified B Corporation since 2017
6. Beginnings
• The data team was separated from the tech team
• We were thrown data and asked to extract insight from it
• Had little saying about how that data was stored or collected
• Because of that, data had to be “fixed” and remodelled heavily
• Over time complexity accrued, making change difficult and slow
8. The awakening
• We recognised that:
• Data is a vital asset for digital businesses
• Being data-driven is an imperative to succeed
• All systems needed to be rethought with data in mind
9. First steps
• New people was hired with e-commerce background
• We started working on web and product analytics, but:
• We kept the data
• We used open source tools (Snowplow)
• We made the data available to everyone (Looker)
• We allowed to join those new data points with our entities
11. Scaling up
• More data engineers, analysts and the first data scientist joined
• More events were added to better understand customer behaviour
• Dashboards were created for every functional area of the company
• A/B tests were made easy and data driven decisions encouraged
• The first machine learning products were launched
13. Present and future
• The data infrastructure is the circulatory system of the company
• Events are the central channel of communication between apps
• Storing them for analysis is just one of their functions
• All decisions are data driven and people can self-serve their needs
• Streaming ML applications are being developed
16. Entities
• They are the “things” you care about in
your company:
• Policies
• Claims
• …
• They change over time
• They’re OK to understand the present,
but not how you got here
17. Domain Events
• A domain event is something that
happened that is business relevant:
• A customer has seen a page
• A customer has bought a policy
• A consultant has called a customer
• They happen at an exact moment and
they are immutable
• Can be used to decouple applications
18. Event Log
• Collection of all your domain events
• Reflects how customers interact with us
over time
• You can use it to:
• Improve your channels
• Improve your products
• Create new products
• Drive the direction of your company
20. Own your data
• Data is one of your most important assets you have
• Think hard before sending it to third party companies:
• You will be limited to what you can do
• You will be at their hands if they don’t evolve fast enough
• You won’t be developing data skills in-house
21. Data quality is a must
• Data should be captured right, put thought in it
• Schema-less and loose approaches will make it difficult to analyse
and reuse data
• You can’t get good insights nor use ML with bad data
22. Invest in data skills
• The data field is moving very fast
• Outsourcing your data needs will leave you without the skills you
need in order to evolve your data platform
• If you can’t evolve your platform it will become obsolete in a very
short amount of time
23. Make data simple
• To capture: anything worth building is worth analysing, build tools to
make it easy to do so
• To transform and aggregate: a single data atom doesn’t tell the
whole story
• To analyse: put all data in one place, make it as open as possible
• To discover: if no-one can find it, it doesn’t exist
24. Use domain events
• They have a lot of analytical power
• You can use them to:
• Make systems communicate with each other
• Auditability
• Understand how your customers interact with you
• Deliver a better service
26. The Lab
• Fortnightly meeting to discuss live and potential experiments
• Grounded in bayesian statistics and lean methodologies
• Run by a senior analyst
• PMs regularly update the whole company about the learnings
• Experiment results guide business and product decisions
27. Call centre improvements
• We built our own contact centre telephony software around events
• Every call generates more than 20 events
• Allows us to optimise:
• Routing and scheduling of calls
• Staffing and training
28. Lead scoring
• We can use the event log to understand which customers are
interested in buying and which are just looking around
• Allows us to understand what they want in real time
• We optimise our marketing and contact strategies based on that