Accumulating marginal gains over time, by removing less probable ideas first.

Who knew that small nudges, easy to implement, could make you the winner in a photo-finish innovation battle?

At roughly age 20, David Brailsford gave up his job as an apprentice draftsman, and he became a autodidactic student of physiology and sports psychology, reading whatever he could. After earning a couple degrees, Brailsford became program director of British Cycling, and eventually he became performance director in 2003. The came up with the idea of focusing on “marginal gains” to improve his team’s performance. The idea was to analyze every aspect of riding a bike, and to improve each by even 1%. Separately the pieces are small, but done over many factors, he hypothesized that the gains would be substantial. He took many measurements, and enforced small nudges in a disciplined way. The hypothesis was tested in competitions, and Brailsford teams won many medals in the 2004, 2008, and 2012 Olympic Games, and his team had many of the Tour de France winners while Brailsford was performance director in the 2010s. A 2013 book Mastermind was published by Richard Moore.

Making small gains, whether it’s in bicycling, mathematics, sales, investments, or innovation, might seem onerous. But if done with discipline repeatedly over time, substantial gains can result. For instance, if you improve 0.1% on something for each of 240 working days/year on something, you would improve by 27%! (1.001)^240 = 1.27.

One of the surprising areas in innovation where small gains gain occur is in choosing which of your hypotheses to test first. It’s so tempting to “verify” or “validate” the most probable hypotheses, but an analysis using Shannon information theory shows two surprising results:

1. Testing low probability hypotheses first increases the amount of information you gain much more (e.g., 5-10 times) than testing higher probability ideas.

2.  Trying to “validate” your hypothesis is always a loser! You ironically end up with more uncertainty than when you started!

CONCLUSION:

- If you have a question, and have 3 or 4 or 5 ideas of how the answer might go (i.e., hypotheses, or educated guesses about the answer) …

- Use your team’s best hunches about the relative probability of each hypothesis …

- And rank them from low to high.

- Start testing the low probability hypothesis first, with the intent to show that the hypothesis does not work (or work well), and remove it if correct.

- Of course, if the low probability hypothesis survives the test, you’ve learned a lot, and perhaps saved a winning idea!

This method might gain you just 0.1 to 2% per question. But if you ask one question per day for a whole work year (240 days), you’ll gain 27% or more, perhaps even more than 100x! And that might be more than enough for you to finish the photo-finish innovation race -- it might be enough for you to be highly successful over the long term!