What happens to analytics when your data stinks?

October 25, 2017 / 3 min read

Are you deriving the maximum possible return from your analytics? Would you know if you weren’t? ISO and Earnix recently conducted a survey on analytic practices by property/casualty insurers in North America. Preliminary insights revealed key successes and challenges among insurers.

What happens when your data stinks

Predictive analytics here to stay

Overall, initial survey findings support the popular belief that predictive analytics are here to stay. In the survey of insurance professionals, 55 percent of respondents found predictive analytics to be highly valuable to their organizations, and another 40 percent found them to be at least somewhat valuable.

But what about the other 5 percent who aren’t achieving value? And what is keeping the middle 40 percent from achieving maximum value? Initial results reveal that achieving maximum return on analytics involves navigating several tricky obstacles.

The chart below illustrates what respondents indicated to be the most significant challenges to successful analytics.¹

data stinks graphic

The top responses may generally be categorized as:

Your data stinks! Poor data quality, perhaps as a result of legacy systems or myopic database design, can doom an analytic project before it starts. At the same time, reengineering data warehouses isn’t necessarily practical in every organization. For reasons not limited to these, data quality was the most frequent answer to the “challenges” question, noted by 57 percent of respondents.
There’s not enough of it! In the happy event that data quality issues are unsystematic and/or treatable, a key to separating signal from noise is having sufficiently large volumes of data so that predictive analytics can leverage the law of large numbers. But such volumes may be only a pipe dream for all but the largest insurers. In the survey, 48 percent of respondents identified “insufficient data volume” as a key challenge.
You don’t know what to do with it anyway! Even in the presence of large volumes of high-quality data, lack of vision can run an analytics project into the ground quickly (or worse, slowly and expensively). Analytic efforts can fall victim to a “cart drives horse” mentality, where the aforementioned data constraints dictate the types of problems solved, rather than business problems dictating the analytics applied to the data. Over a third of respondents identified focusing on the right problems, finding enough time, and evidencing cost-benefit as key challenges.

As wicked as the challenges are, the good news is that many of our respondents feel these issues are manageable and that their organizations derive great value from analytics in the areas of profitability, cost/risk reduction, operational efficiency, and revenue growth. Data tools such as ISO RiskElements^TM can provide insurers with “model-ready data” at the policy, risk, and term level that they can use to help short-circuit their volume and quality limitations.

Segmentation tools such as ISO Risk Analyzer^® can assist in solving the business applicability problem by helping insurers classify and price risks based on highly relevant and predictive insurance risk factors. By and large, success requires committing to analytics as a purposeful problem-solving exercise, rather than going through the motions. What’s holding you back from getting the most value from analytics, and what approaches are best suited to overcoming these challenges?

¹The chart displays the percentage of initial survey respondents who indicated each issue (on the left axis) was a significant barrier to analytics success in their organization. Shortfalls in data quality, volume, and strategic acumen dominate the responses.