Published: September 13, 2018
Does early-stage startup try highly risky things to explore breakthrough, or does startup exploit current situation and knowledge to make the best out of it?
Early-stage startup is definitely designed to implement the former, but average people feel uncomfortable on unstable situation startup faced with repeatedly. We cannot ignore cost of exploration. It could pressure team heavily.
In my hypothesis, startup with qualified product tends to show more robustness than startup without it. Qualified product allow us to reduce exploration cost. Early-stage startup is exploration machine, so efficiency of exploration and enough number of trials is inevitable.
Qualified product solved problem on member’s emotional aspect. Anxiety or doubt by members is one of the most dangerous factors that halt the business. Performance of team is relying on how members sees situation it is included. If members of team consider it hopeless, it is the time of exploration.
Startup should take Upper-confidence bound strategy. Upper-confidence bound strategy is known as the algorithm is based on the principle of optimism in the face of uncertainty, which is to select your options as if the circumstance is as nifty as is naturally possible. It is about being optimistic for uncertainty and keeping choosing the most plausibly best options.
Actually we don’t know enough about the option, we can not precisely estimate the return the option will give us. we are bound to learn more and improve our future estimations by pursuing the option.
Options with high uncertainty usually lead to a lot of new knowledge. We can expect reward from failures startup made. It is desirable to share new knowledge as soon as failed startup have it, but startups are competing each other, so we can’t enjoy fully sharing of the Information.
I guess most of startups never finish comprehensive exploration process in the field they targeted, since team is easy to lose capability after series of failure. So, I think very small party is favorable to early stage exploration, for the purpose of reducing possibility of Human-like disorder and move quick.
Bandit Algorithms “The Upper Confidence Bound Algorithm”