It's Not Just About Stats: Balancing Modeling with Stakeholder Management

Robert Horrobin, AVP, John Hancock

Robert Horrobin, AVP, John Hancock

Smart watches and other wearables; sentiment mining and social feeds; cloud-computing and open source analytics packages; with the ever-increasing amount of data being made available for deeper insights, we as an industry have wisely invested in the technical and talent capabilities of Advanced Analytics Practices (AAPs). As a natural by-product of AAPs’ maturation, many have adopted model lifecycle frameworks. Still, it is not uncommon for projects to struggle in getting past the experimental stage. To truly realize the benefit of AAPs, the leaders of these internal consultancies need to consider the non-technical aspects of their work, including a consistent process for managing internal customers (stakeholders), building alliances with key partners, and driving their team with an eye for portfolio profitability.

Know Thine Stakeholder

All too often, Data Scientists get access to data and start modeling, trying to identify some great unknown insight that will wow their internal customer. This can be a recipe for disaster in a few ways. Firstly, AAP modelers can be chasing too high a degree of precision, working feverishly for unnecessary degrees of certainty. Secondly, AAP teams may start running with research without realizing their stakeholders aren’t 100 percent certain of the problem they are looking to address; follow up questions can solve much of this. Thirdly, more dialogue with stakeholders may reveal issues to implementation once the model is complete; asking “If I have a perfect answer for you, what will you do with it?” can reveal unforeseen barriers to implementation early and often.

"We leverage advanced statistical methods to find out what customers really care about, so our leadership can make customer investments in a pointed fashion"

Furthermore, as a guide to growing relationships AAP leaders can mull over questions such as:

• What issues are facing stakeholder areas (Growth targets? Expense pressure? Customer retention?)
• Regarding stakeholder time and resources, what are AAPs competing against?
• Are stakeholders comfortable with bold solutions or prefer incremental change?

If AAPs understand how their work supports stakeholder objectives and tailor efforts to suit; they can avoid struggles that don’t arise until the point of model implementation when much of the cost has been already incurred.

Build Critical Alliances for Mutual Benefit

Stakeholders are not the only key AAP relationships. Often, models fall short of desired results because implementation relies on many different parties across the organization.

Examples of key relationships include:

• Suppliers (e.g. IT for data access & data understanding, finance for funding proposals)
• Supporters (e.g. risk/compliance for signoff of delivery)
• External Alliances (e.g. Universities for talent, vendors for new technology)

By understanding these Alliances, AAPs can explore where mutual benefit may exist, which in turn creates opportunities for economies of scale or win-win value propositions.

Source Work Appropriately and Create Demand

In speaking with peers across our industry, another persistent challenge is being able to clearly state the value proposition analytics provides and creating demand for services. Having a clear and quick value proposition statement in the language of potential clients can do wonders here. An example statement could be “We leverage advanced statistical methods to find out what customers really care about, so our leadership can make customer investments in a pointed fashion.” This value proposition can also change depending on audience.

Additionally, AAPs can decline work that doesn’t meet one of three objectives:

1. Work that is proposed from a new stakeholder community, which (if taken on) can create positive relationships and potential future work. This type of work should be “quick wins” and have low cost of failure.
2. Work that is driven through collaboration and ideation with stakeholder partners that will have demonstrable financial impact and matches effort vs benefit.
3. Work that may not have demonstrable financial impact, but will have measurable impact in some form (example: customer satisfaction scores).

Having a mix of model categories from the above helps ensure a steady flow of new work and a track-record of profitability.

Measure Success via Scorecard

AAPs’ abilities to demonstrate value is key, especially to avoid being viewed as a cost sink. As an example, an analytics team can aim for each project to have a one-year payback period at most. They can then track expenses and benefits through a scorecard that is regularly updated and distributed to senior leadership, such that they can see the benefit of modeling efforts. This measurement method also allows AAPs to triage and sequence work. They can take some risks on projects that may have a high potential of failure but with very large potential upside, but only after they’ve delivered some profitable, lower risk projects.

Test and Re-evaluate Organizational Assumptions

Lastly, AAP leaders would be wise to develop a list of assumptions about their team, the focus of the larger organization and how they support the overall mission. By constantly re-evaluating and re-testing the above, AAP leaders may see that team skill sets no longer align to the strategic direction (e.g. you have deep domain expertise in Natural Language Processing, but the larger organization is a 100 percent self-service model now). Through this self-evaluation, you may also identify new Stakeholder relationships or alliances that need to be addressed.

In summary, consistent and well-proven approaches for model life cycles are indeed key for delivering value, but equally important is a structure for managing cross-functional relationships. By valuing this component of running an internal consultancy, AAPs can be viewed as an area worthy of investment, effectively working cross-functionally deliver value in unexpected places.

Weekly Brief

Read Also

Protect or Innovate? Cutting Through the Noise When Evaluating Predictive Models

Protect or Innovate? Cutting Through the Noise When Evaluating Predictive Models

Tom Fletcher, PhD, VP, Data Analytics, North America Life, PartnerRe
Optimizing Innovation Initiatives by Artfully Managing Change

Optimizing Innovation Initiatives by Artfully Managing Change

Lori Pon, Director, Claim Contact Center and Claim Handling Unit at AAA-the Auto Club Group
 Digital Ecosystems and Insurance - A Winning Partnership

Digital Ecosystems and Insurance - A Winning Partnership

Sean Ringsted, Chief Digital Officer, Chubb
Data Governance Systems Undergoing Ongoing Evolution

Data Governance Systems Undergoing Ongoing Evolution

Paul Pries, Director – Data Governance, West Bend Mutual Insurance Company
People as Decision-Makers; Technology  as an Enabler

People as Decision-Makers; Technology as an Enabler

Ralph LaSpina, EVP, Chief Marketing & Underwriting Officer, FCCI Insurance Group