Case Study: Advanced machine learning use cases for major UK-based insurer
One of the biggest issues companies face today is data fragmentation. Businesses store both online and offline data, but these may not always be connected. More and more companies now realise the need for personalisation and optimisation of the process in order to keep up with their competitors. Data is becoming the key factor in achieving the most critical corporate key performance targets – higher sales, conversion and retention rates, lower churn rates, and greater customer satisfaction.
Achieving the personalization KPIs resulted from more than ten months of our ongoing cooperation with a major UK-based insurance company. We joined their team with 10 data engineers and consultants and successfully deployed the solution for all their data and use cases by setting up the secure Databrics data platform on Amazon web services.
Our client has more than 3 million customers and collects both online and offline data. Online data is presented by activity on the website, third-party data, conversion funnel data, and the data related to the ads shown to the customer. Offline data include customer data (age, gender, etc.), insurance contract data, marketing communication (e-mail), call centre data, financial crime data (fraud), data from internal models (customer value, optimal price, etc.), claims, and operational data.
They have been storing data on different platforms such as Excel, Oracle, Salesforce, Adobe Analytics, or SAS. Therefore, the data science analytics was not applicable to these data. Additionally, basic insufficient reporting and analysis were done in a low-performance SAS system.
DataSentics, as the data machine learning partner, delivered the new, easy-to-use, and modern cloud platform Databricks. This cloud data architecture was built for the client base to support and enable using the data science cases.
The first step of this solution was to merge the primary or raw data and clean them up, and the second step led to stratifying all data into Bronze, Silver, and Gold categories.
Scheme: the Bronze layer contains raw data, the Silver layer contains filtered data, and the Gold layer contains business data ready for reporting.
All raw data was imported to Amazon Web Services cloud. Data stored in the Gold layer is now ready to report customer retention along with the conversion funnel passage and prepare data to calculate the customer's optimal discount.
Using the A/B testing reusable framework by DataSentics
A/B testing of the call centre script helps exclude other influences and shows the exact test results. This is automatic, reusable, and intended for business use. A/B testing fulfils the client's need to determine whether the new team has better results than the previous one or the new script is more efficient than the old one.
- Replacing SAS (data analytics) and Oracle (data warehouse) with a single platform based on Databricks and AWS connects all the company’s data.
- Building datamarts for data models helps to optimise acquisition, improve retention, prevent churn, and analyse claims and frauds.
- Applying our AI A/B testing framework improves the effectiveness of the call centre agents.
- Using our development framework speeds up the project by 30%.
- Creating a transparent business-oriented data catalogue and documentation keeps the company up to increasing standards.