Becoming analytics-enabled is perhaps one of the most importantevolutionary steps forward an organization can make. The rewards ofdata-driven decision-making can be a powerful boost to the bottomline. For insurance companies, this may include using underwritingpredictive models to increase profitability through more granularpricing, driving a six to eight-point reduction in loss ratios. Onthe claims side, predictive models have helped insurers bettersegment and triage high severity workers' compensation and bodilyinjury claims, driving a four to 10-point reduction in claimsspend.

An important part of the analytics journey is overcoming thenumerous challenges an organization encounters when experiencingthe end-to-end development and deployment of predictive models.Model development (e.g., data assessment, data acquisition, datacleansing), scoring engine development (e.g., scoring engine anddatabase design, development, testing, deployment), and businessimplementation (e.g., strategy formation, change management, toolsfor measuring business) are some of the common questionsorganizations should consider.

Business Issues

Some of the concerns the authors have observed in thedevelopment and implementation of advanced analytics include:

1. Executive Ownership

Without buy-in from senior leadership and a clear corporatestrategy for integrating predictive models, advanced analyticsefforts can end up stalled at model development. In order to beeffective, analytics efforts should involve the key executives whocan help drive acceptance and change throughout the organization.Senior leaders should insist there be a clear correlation betweenthe actions to be taken through model implementation and theexpected business benefits to be realized. Without accountabilityfor a targeted return on investment, organizations risk spending alot of time "doing" versus "getting things done."

2. IT Involvement

Failure to involve IT from the very beginning of the analyticsjourney can lead to significant issues down the road if technologygaps and limitations aren't understood up front. Modelers may finda way to get access to internal and external data, but without thehelp and involvement of IT, it is almost impossible to bring themodels to life in the day-to-day operation of the organization

3. Available Production Data vs. Cleansed ModelingData

Access to historical data for model development is verydifferent from access to real-time data in production, and a strongmodel is only as good as its ability to be practically implementedwithin the technology infrastructure. Real life limitations mayrestrict the data that's available for historical modeling.Sometimes a proxy variable can be used for modeling until the datais available. Analytics initiatives often risk being stymied by thebelief that data for modeling must be perfectly clean andorganized. Predictive model development is not an accountingexercise, but rather a statistical process where numeroustechniques allow the "dirt in data to be washed away."

4. Project Management Office (PMO)

Lack of clear ownership of the end-to-end journey is a commonstumbling block for organizations that have struggled (and failed)in implementing their predictive models. Without the right projectmanagement structure in place, a clear cadence of projectmilestones, and the ownership of deliverables by pre-identifiedbusiness owners, the project could be doomed before it starts. Mostimportantly, the PMO must be able to connect with all interestedparties and adopt an agile approach.

5. End User Involvement and Buy-In

Lack of end user involvement in the planning, design andultimate roll out of the predictive models can be detrimental tothe efforts. For underwriting or claims models, involvingunderwriters, marketing, actuaries, claims adjusters, nurse casemanagers and special investigative unit (SIU) resources early inthe process is critical. End users also have more insight into thebusiness process and may be able to better identify potential gapsor roadblocks to successfully incorporate models in day-to-dayoperations. If the end users feel as if they have a stake in thepredictive model roll out, then the company may be more likely torealize the potential financial benefits. If done correctly, someof the early doubters can eventually become analyticsadvocates.

6. Change Management

Organizations often fail to understand how predictive modelschange the current business and technology operations — policies,procedures, standards, management metrics, compliance guidelinesand the like. Without the proper design, development and roll outof training materials to address the impacted audiences in thefield and home office, the analytics journey can come to an abruptend. Educating end users and other related stakeholders on how themodel will be used on a day-to-day basis, and how their life maychange, is important. A communication plan should be developed toanswer frequently asked questions (FAQs), address common concerns,and help end users appreciate the strategic vision of theorganization. Change management doesn't start and end withtraining; it begins on day one and lasts well beyond the roll outof the models.

7. Explainability vs. the "Perfect Lift"

It is important to balance building a precise statistical modelwith the ability to explain the model and how it produces results.What good is using a non-linear model or complicated machinelearning method if the end user has no way to translate the driversof the score and reason codes into actionable business results?Experience shows that a less complex statistical model developmentmethod yields results similar to those from more complexapproaches, and a small sacrifice of predictive power can result inmarked improvement in the explainability of technical modelrecommendations for end users.

Insurance Company Size Matters

Every company is different, and designing a successful approachto implementing predictive models within the business process canvary widely. What works for a large national insurance carrier maynot work for a regional mutual insurance company and vice versa.Some differences include:

1. Communications and Training

When it comes to change management, larger companies maystruggle with executing communications and training plans across anoften siloed, functionally diverse and geographically spreadorganization. A major challenge involves providing tailoredcommunication and training to address business processes formultiple stakeholders with different expectations and officecultures. Larger companies may have to balance corporatecommunications protocols while smaller companies may not have thesame challenges.

Small companies may underestimate the effort involved in changemanagement and may discount the importance of multiplecommunications, gaining stakeholder buy-in, and conductingeffective training. There is also the possibility of beingstretched too thin to provide the adequate level of dedication andfocus, since they are involved with other aspects of the project inaddition to their day-to-day responsibilities.

2. Competing Initiatives and SystemChallenges

Large companies may also face additional challenges when workingacross new and legacy systems. Involving technical and businesssubject matter experts (SMEs) in the model-build and implementationphases is extremely important to get the full view of the currentand target state across each impacted business process. Largercompanies may also find themselves juggling multiple initiatives atonce (e.g., rolling out a new claims management or policyadministration system along with predictive models), which can be achallenge for resource allocation. Another consideration is howpredictive analytics and other initiatives will affect one another.When related projects are underway, a multi-phased analytics futurestate, starting with a semi-integrated predictive analyticssolution, is a common approach to work around initiatives that maybe affected by predictive analytics.

3. Resource Constraints

Large companies may also face challenges getting the rightresources, and especially the decision-makers, involved in theprocess. During the model-build phase, it's typically easy toidentify who needs to be involved (e.g., analytics team, actuaries,data SMEs). Making decisions on how to implement can be morecomplicated. These decisions need to involve business unit leadsand, in some cases, upper management — busy individuals who may nothave the dedicated amount of time needed for implementationdiscussions and decisions.

While it may be easier for smaller companies to get the businessunit leads and even executives involved in a predictive analyticsproject, they may suffer from having too many people involved,since project leaders may not want to leave anyone out of thedecisions.

4. End User Buy-In and Model Use

Getting the most benefit from predictive analytics is achallenge generally faced by all companies, but it may be morecomplicated in smaller companies due to sensitivity aroundexpectations for using the model. While some companies tend tothink of a predictive model as an additional tool to help makebetter decisions, the degree to which employees are expected tofollow the model can vary significantly.

At smaller companies, employees may worry about their rolesbeing replaced by predictive analytics. Models are not typicallyintended as replacements for people, but rather as helpful toolsfor making more informed, objective, metrics-based decisions. Thisbalanced approach and intent should be continuously communicated toall project participants as an essential foundational philosophyfor organizational buy-in and effective change management.

As Larry Winget said in It's Called Work for A Reason,"Knowledge is not power; the implementation of knowledge is power."Successfully building and deploying predictive analytics goes farbeyond the development of a highly predictive model and a robustscoring engine technology. It is a holistic endeavor that requiresa focused effort on practical implementation within the businessoperation as well as organizational, people and processconsiderations for what is most effective for end users and thebroader corporate culture.