As the curtain comes down on Great Britain’s most successful overseas Olympic games, much has been made about the UK’s no-compromise approach to achieving success. One element of this success is the team’s obsession with marginal gains: small incremental improvements across the board adding up to a significant improvement to the final result.
First implemented by Sir Dave Brailsford during his time as Head of British cycling, the strategy of marginal gains changed the reputation of our cycling program from a laughing stock to what is now known as The Medal Factory. The basics were simple: by analyzing every process that takes place in the performance, and improving each element by even just 1%, organizations can affect large changes on much larger scales.
This didn’t start and end with training and racing. The level of analysis went further and deeper than anything ever done previously. Pillows were tested in an effort to improve quality of sleep, and riders were taught to wash their hands correctly to avoid getting ill. The effects of certain types of socks (or no socks at all) were studied, custom-made carbon shoes were commissioned… the list goes on. All small adjustments, but in sum they made for significant changes in performance and success at an organizational level.
This willingness to look at every aspect of performance clearly pays dividends, and not just in sports.
So why is this not approach not widely applied in the business world? Particularly, why don’t we see more of it in high frequency trading? Like cycling, it is a realm where speed is of paramount importance. If the marginal gains philosophy were applied to high frequency trading, who knows what improvements could be made.
Where to improve though? Or, before we even ask that question, how can we see what needs improvement? One needs full visibility into the operations of the network. And because of the nature of the industry, this reporting needs to be instantaneous and granular. As the bike tires were adjusted for minute changes in PSI, so must we be able to understand at nanosecond granularity the specific issues that the network faces. Every hop along the wire must have its own metadata to compare, to pinpoint problems. The data should be accessible from every vantage point with appropriate contextual information, whether wire data, app data, or anything else.
In the realm of HFT, nanoseconds can mean billions. The aggregate result of these seemingly small changes can mean winning the gold medal—or winning at gold futures.