Gambling on Innovation with Learning

To maximize your median return (i) and minimize your probability of "going bust" in innovation, you need to invest the proper amount in your innovation projects. A previous article (IECR 2021; DOI 10.1021/acs.iecr.1c00511) showed that the Kelly gambling (or investing) strategy helps to make the optimal decisions. To use the Kelly strategy, you must know the anticipated win ratio (b) of an innovation project and its estimated probability (p) of success. But both b and p will likely increase when you bet a fraction (f) of your innovation budget on a project, due to learning that increases both b and p. This article analyzes how the optimal f changes when learning takes place. I use simple linear approximations p = p₀+ Ï€f and b = b₀+ βf. As πand/or β increases, the optimal Kelly Criterion value f_KCincreases, often significantly, calling for a higher resource investment, and also giving a dramatic increase in growth rate (i). It is proposed that organizations measure a, b₀, p₀, Ï€, β, and k, where k is the rate of attempts made per time, and focus on improving them. This article provides the framework to use these parameters, enabling probability processing to improve the probability of success.