The purpose of data analytics in business is often to determine some optimum strategy. But what is often the effect of a purely data-driven business design is the identification of local optima rather than global optima.
Would offering customers a one-week free trial increase conversion? What is the effect of rewards on referral rates? To answer these, experiments are designed and the results analysed through A/B testing, or perhaps by using predictive models. Through repeated experiments at the margins, small improvements are constantly made to bring the business closer to its optimum.
But there are limits to data analytics in identifying optimum strategies.
Local optima vs. global optima
In the simple problem illustrated given below, we are trying to find the point on the horizontal axis that gives us the highest vertical point. A real life question might be “What is the price that maximises sales?”
A business currently operating around the local optimum, and that is testing alternative strategies at the margin, will not escape the local optimum. It will eventually reach the local optimum where any single small step left or right will not produce a better result, despite a much better global optimum being available.
This is a limitation of data analytics that must be accepted. Whilst it may be possible to escape the local optimum by conducting A/B testing on two very different price points (say £10 versus £100), this has considerable risks to goodwill and reputation. Nor can we conduct tests at one level (e.g. £10 versus £12) and try to extrapolate results out-of-sample (to predict the effects of a £100 price point) as it is a clearly statistically unsound practice. In any case, it is difficult to test a £100 price point effectively without some fundamental change in the way a business operates or positions itself.
The role for business vision
In the example above, a purely data-driven business runs the risk of missing the proverbial woods for the trees. The role of data analytics is to inform key decision makers, and only drive optimisation at a local level.
How should a brand be positioned? What brand values should be championed? What strategic partnerships or markets should a business be in? Identifying global optima requires far-sighted vision and human judgement. After all, no amount of data analytics could have outperformed the strategic vision of the late Steve Jobs and his pivotal role in Apple’s phenomenal success.
Visionaries, CEOs and strategists can sleep well tonight knowing there is still a much-needed role for them.