How to Train Your Data?
In order to thrive in today's competitive, technology-driven landscape, insurance companies must learn how to effectively harness the vast amounts of data they are generating. Doing so, however, requires a completely new approach to data analysis and the way insurers do business.
Computers Don't Think Like Humans
Underlying all of the above reasons is one, shared and fundamental point: many insurers unknowingly shoot themselves in the foot with the assumption that their computer systems process data in the same way a human would.
Humans have an easier time understanding structured data, such as a table or a list. Excel, for example, is the most widely used program for statistical analysis in the world because it allows users to clearly see, “feel,” and manipulate the data. The neat columns and easy lookup up and retrieval functions can give users the false impression that Excel experts are also data experts.
Many large insurers also employ BigData departments with data scientists who use more advanced tools for analyzing data. But, they too are either working with structured data or turning information into structured data in order to analyze it.
The real problems in a company's analytics system begin when humans try to organize data in order to "help" the system focus on the most important and influential information. The result ends up being completely opposite to what was intended. Patterns and other vital trends are misunderstood or overlooked outright because a computer intelligence system does not process data this way.
Consider the fact that Google doesn't employ human engineers to index the world's sites. Instead, the search engine giant develops algorithms that can organize the information themselves in a way that makes it easy to discover.
Letting An Algorithm Do The Heavy Lifting
The solution, of course, for insurance companies when it comes to their data analytics is to let their systems do their intended job. They just need to define what business data needs to be collected, what reports they want to see, what actions do they want to take and why. Then, they need to let their big data analytics systems take over. Just like they can trust Google's algorithm to give them the most appropriate search results, they can learn to trust their internal algorithm to arrange the data in the most relevant and convenient and useful way.
This shift in approach is a big game-changer that can free up precious time and other resources. It can also help to significantly focus the insurer's operational activity because it forces the company to figure out the most effective parameters of their data analytics and continually monitor the results.
Why Data Scientists Fails? (Spoiler: Because They Don’t Think Like Computers)
Traditionally, insurance companies relied on statistical analysis and other actuarial methods to assess risk. Now, most forward-thinking insurers are focusing their resources on automated predictive analytics that help them to offer personalized products and positive buying experiences. But more than that, instead of merely adding these technologies and strategies to their current systems, they are strategically embedding them into their workflow and decision-making processes. In some cases, they are completely rebuilding their legacy systems from the ground up.
Since the invention of the computer, information processing methods have improved in leaps and bounds. We've come a long way from SPSS and SAS. Data science, Artificial Intelligence, and other advanced machine learning technologies have extended the breadth and scope of data analysis, leading to capabilities that were simply unimaginable a few years ago.
Like the banking sector, which is old and heavily regulated, the insurance industry has been slower to embrace technological change as compared to other industries. Nevertheless, there has been a widespread push in recent years among insurance companies as well as actuarial departments to adopt some big data capabilities.
At the beginning of this movement, the goals were to fill in customer information gaps and adjust risk and premium profiles in accordance with mortality tables and additional statistical information. Lately, however, insurers have been moving on to more advanced abandonment models and cutting edge marketing strategies, such as Next Best Action. Many of these companies have hired in-house data scientists and invested heavily in sophisticated analytics technologies.
Yet, for all of the time, money and effort invested in these tools, many insurers fail to achieve a meaningful ROI. They may have invested enthusiastically in their new analytics system but are then disappointed to discover that they’re not really being utilized.
Why is this happening?
There is usually some combination of four main reasons:
1. Old legacy systems are unable to implement complex data models and effectively render most of the analytics unusable.
2. The information gathered in the organization is insufficient and arranged in tabular arrays, such as spreadsheets. It allows for some Business Intelligence activity and reporting, but little more. Insurers end up lacking an adequate, big-picture understanding of both customer and business realities.
3. Insurers must allocate costly resources in order to hire data experts who can understand and "talk" to these sophisticated models. But, this leaves out non-technical business users who might not have the analytics skills of the data science team. Access to their valuable business insight is thus limited.
4. The analytics tools are not properly integrated with the daily workflows and operational decision making processes.
Bottom line: The data that an insurance company collects and generates is supposed to support both the insurer and the clients it serves. But, this can only happen when it is effectively harnessed and employed. Sometimes that means insurance companies need to get out of the way and let a self-learning algorithm give them the kind of results and ROI they're looking for.