Skip to main content Skip to Footer


Advanced analytics for insurers: Building machine learning capabilities

Insurers have long been leading proponents of using data for pricing and risk management, and have invested significantly in technologies and people to support these functions.

However, the prolonged focus on these traditional capabilities has resulted in a partial failure to exploit the maturity of advanced analytics. Other industries far outpace insurance in terms of their analytics capabilities, and are re-engineering their end-to-end value chains around machine learning.

Insurers gather enormous quantities of data. Some of this data feeds existing management information (MI) and business intelligence (BI) applications, but most data is unused and unexplored as it may be unstructured, excluded from the traditional MI infrastructure, or its value simply not recognized.

Maturing technologies such as cloud-based NoSQL databases, open-source R language, data ingestion and visualization layers are transforming the pace at which data can deliver tangible and actionable value using analytics.

It is becoming indisputable, based on events in other industries, that companies which do not use advanced analytics to power their businesses will be outcompeted. Insurers are too early in the advanced analytics adoption curve to observe significant differences in business performance between insurance analytics leaders and laggards, but there is no doubt that this is coming.

Given this imperative for making advanced analytics a core business competency, insurers should seek to maximize the benefits of this technology by ensuring they have built the requisite data, technology and teams.

Data built for analytics

Insurers’ data is typically organized, managed and governed for MI before being adapted for analytics. This is the cause of increasing frustration for data scientists within organizations, who spend most of their time re-engineering data for analytics purposes and then trying to retrofit the outputs into existing systems.

Insurers need to rethink how they structure their data pipeline for MI and analytics, taking the following into account:

  • Machines learn from data in a different way to humans: Gather all the data, keep it as raw as possible, and use algorithms to mine data produced from controlled experiments.

  • Grow the organization’s insight account: Rather than centralizing these data sets from end-user devices and data centers into monolithic analytical records, maintain virtual analytical records that allow data scientists to recreate data sets across any time period and for any customer cohort.

  • Reduce the friction from insight to action: A centralized virtual analytical record of the initial predictive analytical model can be used in near-real time by any decision system through an API call. This removes the need to translate model outputs, reduces errors at run-time, gets benefits flowing quicker and makes everyone’s lives simpler.

Technology built for analytics

Advanced analytics requires different technology stacks from those used to deliver MI and BI. Points to consider when working out how to build and deploy technology for analytics include the following:

  • Every analytics business is a cloud business: Advanced analytics thrives on variable bursts of storage and compute—only the cloud can provide this.

  • Analytics needs the right tools: Data scientists need to experiment, using toolsets that enable rapid data manipulation, feature engineering and analysis. Data engineers need robust ETL toolsets that will not fall over every night. It is important to consider a balance of both these toolsets, and to build processes that deploy insights and models from one environment to the other.

  • Design analytics to enable rapid insight to action: Deliver packaged insights to teams across the organization by giving them restricted access to the virtual analytical records through a modern visualization tool. Also ensure data models deliver near-real time responses.

Teams built for analytics

Analytics teams need to be structured to deliver value at pace. The best-performing analytics delivery teams adopt and adapt many of the key principles of agile-lean delivery. This means focusing on value, championing speed, and taking an organizational approach that ensures a transparent product is created within weeks of starting projects.


Stay In The Know

Receive e-mails from Accenture featuring new content that matches your interests.

Visit the subscription center to make your selections and subscribe to New from Accenture.