In 2012, British cycling's performance director Dave Brailsford explained that the team approach to winning medals was based around Marginal Gains. While having good funding for our athletes has certainly helped, the big differentiator for Team UK success during the 2016 Olympics has been the Marginal Gains approach.Marginal Gains assumes there is no magic bullet or secret sauce to winning. The approach is to look individually at each and every possible component that might contribute to performance. Then, for each and every component, analyse it, measure it and then with a scientific method, optimise it. Each little component might only improve things by 0.1% here or 0.3% there but the final result of all this performance tweaking will take you from the back of the pack to top of the class.

At INDEZ we have followed this exact same approach to progressively optimise businesses such as Toolstop that we have worked with from a pure play ecommerce startup through to one of the largest and most successful online tools suppliers in the UK.Key to this approach is being systematic with attention to detail throughout. While part of this is down to business processes and quality control, another factor boils down to the personality type of the individuals delivering the solution and, in particular, the ecommerce managers.The process for Marginal Gains analysis in ecommerce follows the following 5 steps:

1. Break-down and define each and every component that impacts ecommerce performance.
2. For each component, establish a measurement protocol.
3. Measure and analyse each component across a wide variety ecommerce interactions and sales.
4. Optimise directly or following A/B or multivariate testing.
5. Continually repeat with new data and with a focus around those components with the largest impact on your profit.

As the business grows, you will find that you continually take something that you originally classed as a single component and segment it into ever more sub components and sub-sub components. The process terminates when the cost of cost of undertaking the marginal gains process is higher that the profit yielded by that component of the overall performance. For those with an interest in AI, this is exactly the same approach that’s used for Inductive Rule Generation where the termination calculation is a simple entropy function.