Insurers of all sizes are investing in predictive analytics as a way to moreaccurately understand the risks their companies will undertake.However, not all of carriers have secured buy-in across theorganization. In fact, at the recent InsureTech Connect event, James (Jimi)Crawford, former NASA scientist and founder of Orbital Insight,theorized that the hardest part of selling analytics solutions toinsurance companies was finding a person on the innovation team whohad a keen understanding of the operations side of the business (orvice versa). 

Orbital Insight is a company that utilizes satellite imagery tohelp governments and businesses create a better understanding ofthe world. Like other technology solution providers, they struggleto find people in insurance who understand both the business andanalytic sides of their organizations. 

This article looks to bridge the gap between the numbercrunchers adjudicating policies and those responsible forprofitability, identifying the common verbiage across both sides toilluminate where common pitfalls may lie.

Related: Doyour data analytics team members speak the samelanguage

Accessing the right data

When insurers decide to invest in predictive analytics, they'reoften confronted with the same set of initial decisions. One of themost common is deciding whether to build a model exclusively within-house data, or to enlist a third-party data vendor to fill inthe gaps. Developing an underwriting analytics model with theappropriate information is essential for success, as no amount ofmodeling expertise can make up for a lack of data.

Related: The power of analytics for insurance: You ain'tseen nothin' yet

Internal versus third-party decisions apply for both large andsmall companies alike. While large insurers like Travelers, USAA,and AIG have enough data to build an effective model, theyregularly seek third-party data sources to augment their modelingefforts. Small and mid-sized carriers often lack the volume of dataneeded, and are more heavily reliant on third-party data.Sufficient data assets are the first hurdle to meet in order toachieve the business outcome goals of a predictive model. Thevolume, breadth and depth of the data are all vital. Lack of "theright" data in a model will create issues like blind spots orsample bias that render model conclusions invalid, or at the veryleast, untrustworthy.

This isn't a small issue. Those on the analytical side of aninsurer will typically consider "more data" as better, but thebusiness side may need more convincing in terms of the expense orthe need to participate in contributory databases where manyproprietary data assets reside. Interestingly, the very data thatdrives profitability and account longevity can be found in thesedatabases. In these instances, the information that data scientistslack access to is often the data that drives the numbers thebusiness side is most interested in.

Related: Insurance, data analytics and internal operations:untapped opportunities

Time management

Equally important to identifying whether an insurer has enoughdata, is data availability. Data scientists spend 50-80% of theirtime wrangling data instead of providing true modelingcapabilities, according to The New York Times. Instead of locatingdata they could be spending time connecting business strategy toanalytics project, overseeing management of the project ormore.

As such, here are the steps organizations must take to ensuretheir data is conducive to predictive analytics initiatives, andwill ultimately drive the highest ROI.

Technical Inventory

A technical inventory ensures the data is usable, properlyidentified and organized prior to building a predictive model.Profiling the data available, looking at information like aspecific population and basic summary statistics, will help datascientists make initial assessments of whether the insurer hasenough to work with. 

Insurers will also want to check for structural validation(matching the fields). This can be done different ways, but isoften a visual grid with rows and columns that sifts out eachuploaded data source file to locate what 'necessary' components areavailable, missing or align with what a specific predictive modelneeds. Many platforms support data uploads, but isthere opportunity for greater analysis? In the example below, thegreen check marks show a match between client data and systemexpectations. The yellow marks go further, taking data initiallyunaccounted for into consideration, providing an opportunity formore in-depth analysis. This tenant of structural data makes sureoptimal insights are gleaned from the data. 

Technical inventory establishes an initial level of data qualityto reuse as the data scientist iterates through the process.Ongoing monitoring and performance evaluation of a solution isimportant, as data acquisition is a continuous process, not justfor the initial model build.

Related: 3 keys to calculating loss adjustmentexpenses

Data cleansing

Once technical inventory is complete, the next step is tounderstand what the data will support from the functionalperspective. Does the data support the conclusions an insurer wantsto see? On the business side, these conclusions may mean "how muchmore profitable will we be," whereas the data scientist maydetermine they want to understand various effects of specificcharacteristics. This discussion of "why something impactsprofitability" vs. "how much it costs" is an ongoing discussionbetween business and data teams.

However, both can agree that a model aimed at providing insightsinto new geographic territories or performance over time, must begiven clean data with which to work.

For example, a common method to see both functional perspectiveand data cleanliness is a heat map. This groups and displays dataalong expected attributes, making it easy to identify data that mayneed further cleansing.

The key is to start simple. Don't use all methods of datacleansing, but rather focus on which technique gives the most valuebased on the data.

Data lineage

As insurers go through the processes of technical and functionalevaluations of data, they will encounter items that requirere-processing of the data. For the business side, this should feelfamiliar. It is the equivalent of providing ongoing support forquestions that come from the finance or auditdepartments. 

During repeat the validation and cleansing, it's important tomaintain visibility into the data history to understand the changesthat may occur over time. Retaining profiles for each iteration ofdata provides a history of data quality.

Even in the absence of a formal data governance, simply trackingdata profiles will answer questions about what was considered, howmany iterations of cleansing an insurer went through, and whichversion was used in the model data set. These are especiallycrucial if a carrier uses third-party data to supplement their ownto avoid sample bias. This information must be requested of thevendor before using their data. The image below is an example ofhow Valen shows data lineage.

This list shows what files were received, who uploaded them,current status and details for where the files are within our dataprocessing workflow — including final approvals of thedata. Another important view of the data as an insurer iterates isthe impact of each iteration on the resulting approved dataset.

Although the journey to connecting business leaders with dataleaders is still in its infancy, there are universal steps insurersshould take to ensure data initiatives are positioned to drive thehighest ROI. After all, the two sides of the business can certainlyagree that business that's both within regulatory guidelines andmore profitable, should be the goal.

Following the steps above, and communicating within theirframework, can help to bridge this departmental divide.

Dax Craig is the CEO and presidentof ValenAnalytics®. He can be reached by messaging [email protected]. The opinions expressed here arethe writer's own.

Read more columns by Dax Craig:

Variety and context are critical for predictiveanalytics success

A proactive approach to working withregulators

Barriers to the innovative mindset