Best's Review



At Large
Maximizing Analytics

Simple tools can produce significant results.
  • Bill Panning
  • December 2018
  • print this page

Bill Panning

Bill Panning

Firms that are cautious about investing in new and unproven technologies should consider ways of extracting additional value from their existing algorithms.

New and sophisticated analytical techniques with intriguing names (and uncertain benefits) have captured the attention of many firms eager to increase profits.

Firms that are cautious about investing in these new and unproven technologies should consider ways of extracting additional value from their existing algorithms (prescribed numerical procedures) for making complex decisions, whether implemented by software or by humans. Here are four such opportunities, with examples of their value.

Reality checks focus on algorithms that respond to external conditions and produce results that affect a firm's decisions. Are these algorithms correct? Have they been made obsolete by changing circumstances? Preceding the financial crisis, financial firms assumed portfolios of home mortgages in geographically dispersed locations were diversified. But this situation changed, due to the evolution of nationally integrated mortgage markets. Consequently, home values and mortgage rates in locations as distant as Phoenix and Philadelphia became highly correlated. This change was not recognized by financial analysts nor incorporated in their pricing models, which significantly underestimated the riskiness of their mortgage-laden investment portfolios. This increased divergence between actual risk and the prevalent estimates of its magnitude played a huge role in precipitating the financial crisis.

Accuracy checks examine the internal assumptions on which decision processes are based. For many years, insurers regularly altered their mix of taxable and tax-exempt bonds to maximize their after-tax net income in response to forecast changes in underwriting income. A colleague and I were assigned to evaluate that human, but nonetheless algorithmic, process at a large firm. We discovered that these seemingly routine decisions failed to consider the significant transaction costs of changing the portfolio composition. In most situations these costs outweighed the prospective tax benefits. We also discovered the forecast changes in underwriting income, obtained at great cost, had no predictive value whatsoever. Scrapping this whole process saved millions annually.

Focused reports are customized versions of various monthly financial reports that insurers routinely produce. They focus recipients' attention by omitting or condensing numbers that are of little interest to the recipient, and using specific fonts and colors to direct attention to what matters most to them (e.g., numbers that significantly deviate from past experience or from target ranges). Their purpose is to enable decision-makers to make better use of a valuable and limited resource—their attention.

Feedback meetings with the recipients of analytic reports can be valuable. At one such meeting, a department head confided he had growing misgivings about a pricing model his department had used successfully for years. Recently the model had begun to produce bids that were significantly lower than competitors. We reviewed the model and discovered a subtle programming error that failed to adequately reflect recent changes in the business environment. Correcting that error solved the problem and improved business performance.

Firms can implement such practices without help from expensive consultants. They may not be newsworthy, but can boost the bottom line.


Best’s Review columnist Bill Panning is principal of ERMetrics LLC. He can be reached at

Back to Home