Why applied analytics projects have more success
Having benefitted from listening to a CX leaders’ perspective on vision for data & analytics, lets consider applied analytics projects.
In fact, this focus is an opportunity to hear the voice of another key stakeholder in major analytics implementations. That is, the perspective of the software provider or IT supplier.
Now, I know, these people can have a bad reputation. Too many ‘shiny suited’ salespeople and unfulfilled promises, post demo, can leave many leaders very sceptical.
But, there are also good people working in technology companies & the experience they gain from helping multiple companies implement analytics can be a godsend.
So, I’d like to welcome a new guest blogger for our blog. Dax Craig is co-founder and CEO of Valen Analytics. Our old friend Paul Carroll (of Insurance Thought Leadership blog) recommended Dax, to share these thoughts on why applied analytics projects are the way to go.
Let’s suspend that skepticism, long enough to hear Dax’s useful advice for your projects…
Applied Analytics are key for progress
While most carriers collectively understand that predictive analytics is necessary to remain competitive with other data-driven and VC-backed companies, progress isn’t as fast as it should be. Some are still not using analytics at all despite knowing its importance, while many others limit use to one area of the business. One main obstacle is a disconnect between the C-suite and analytics staff.
The C-suite will not invest in initiatives they cannot measure, manage and understand. Therefore, how a business handles the implementation of analytics is just as important as the predictive models themselves. This is why applied analytics programs are needed to push the insurance industry into the future.
Applied analytics recognises the value of data in specific business processes and basing key decisions off the insights. Carriers must define their strategy and goals around the initiative, decide on an implementation approach and convey it successfully to the rest of the organisation to ensure buy-in and adoption. Executing this can be difficult, often a result of balancing sound technical decisions against each individual organisation’s company culture.
Strategy and Goals
The first step a carrier should take when applying predictive analytics is building the strategy and selecting the goals. Beginning with smaller projects is often an excellent way to build confidence for future implementations across different areas of the business. In this phase, carriers decide which specific challenges they hope to tackle using predictive analytics. For example, a carrier’s main goal may be to improve risk selection and pricing, and a secondary goal would be to achieve underwriting consistency as the carrier grows its existing business in new states. By defining what the organisation hopes to gain from a particular initiative, it avoids misunderstanding and confusion during the implementation and duration of the project.
Next, the carrier must choose their success metrics. Using the same example, if the target for a predictive analytics initiative is improving risk selection and pricing, loss ratio improvement is often the best measurement. If business growth is the primary concern, profitability and aligning price-to-risk is an important metric to use. The C-suite is results oriented and this step sets the foundation for the implementation that follows.
Once a project has been selected, its goals determined and success metrics defined, an implementation plan should be created to strategise the role governance will play within the corporate culture. The way a carrier must address this step largely depends on the type of predictive analytics initiative in place. If a commercial auto carrier uses analytics to improve pricing with its long-haul trucking business, it’s key to find the balance between the model score and your underwriters’ expertise. Don’t leave it to chance, carriers should develop clear rules for how to use the predictive model that matches each company’s weaknesses and strengths.
By using the predictive model, policies are scored individually and assigned to a specific risk “bin”.
As one example, a carrier may decide that the best performing risks for smaller policies (bins 1-3) are approved for straight through processing, the average risks (bins 4-7) must always be reviewed by underwriters, and the poorer performing segments (bins 8-10) should be avoided altogether. A common and more granular form of implementation is determining how much credit or debit can be added to each policy depending on the model score, before needing management approval. By implementing rules of engagement that correctly fit a specific organisation, it will not only boost the effect of the predictive analytics project but make it easier to manage and make necessary adjustments long-term.
Even if a carrier has an excellent strategy and model, it means nothing if those who will be using it on a daily basis fail to do so correctly. During the implementation process, all members of the organisation must be aligned with project goals and comprehend its importance for the company. This means undergoing training and support by those involved closely with the initiative — whether the model is being built in-house, through a third-party vendor or a consultant. There should be complete transparency throughout the process, and room for adjustment based on feedback from the staff. Predictive analytics is an imperative tool in its own right, but just like any tool, it requires a skilled individual to obtain the best results. In fact, Valen research shows the best results are found when combining human judgement and predictive analytics.
The graph shows the lift of a predictive model. The greater the lift, the more effective the model is for the carrier. The blue line represents the loss ratio improvement based on a combination of the underwriter and the model when making decisions on pricing policies. There is clearly a more significant lift here when compared to both the underwriter (green) and predictive model alone (red).
In order to sustain long-term plans and goals, the predictive analytics strategy must converge with the overall corporate strategy. That can’t happen for any real length of time without executives confidently making important decisions using the insights that come from predictive modelling. Only when a carrier and all of its members fully understand those insights and trust in data are they able to become a data-driven organisation.
Response from Customer Insight Leaders
Thanks for that advice, Dax. Much appreciated. I completely concur with a number of your key points:
- Need for alignment with corporate strategy – ensure applying analytics where matters
- Start small & grow through proven delivery of results
- Even predictive analytics produces best results when combined with human expertise
What about your experience of applied analytics projects? As a Customer Insight Leader have you seen the pitfalls & benefits that Dax outlines?
I hope this post helps & can in small way foster the kind of knowledge sharing and partnership between suppliers & insight leaders that can result in the most successful projects.