From today on, we have a new policy in our company.
It is about the right to make mistakes: the fundamental and inalienable right of a startup. Every employee of a startup has the right to make a mistake (to err is human).
All startups operate in a zone of uncertainty and we cannot hope that any hypothesis will be confirmed with a probability of any close to 100%. Moreover, we must be aware of the fact that most of our hypotheses will ever be confirmed.
As an investor, I believe that each question has 1000 possible answers, of which 900 are incorrect, and the other 100 are correct. Out of the 100 correct, you need to choose the best one suitable for these circumstances. We have no methodology to define what are these 100 correct and the 1 that actually fits. Except for one. This methodology involves conducting experiments to go through all the options. But at the same time, experiments should have two effects — to increase the speed and efficiency of the variation verification, and each mistake should increase the team's expertise. It is non-linear. That means that verification of the 10th hypothesis should be many times faster and more effective than the first, and the 20th hypothesis is many times more effective than the 10th. Thus, with the growth of our expertise, some answers begin to fall off on their own, as mistakes become more informative.
Thus, an investment in mistakes is an investment in the growth of the team as a whole and of each team member separately.
Here are some requirements for experiments to test hypotheses where we will make mistakes in the near (and not only) future:
- 1. An error should cut out erroneous hypotheses, or rather branches of hypotheses
- 2. The error must be planned and parameterized. This implies that we form a hypothesis, describe an experiment that confirms or refutes it, with metrics on the basis of which conclusions can be drawn.
- 3. We work in Extremistan. This means that if an error is confirmed, we normalize the damage (all experiments should be normalized in terms of time, budgets and the diversion of company resources), but the benefit should be asymmetric. If the hypothesis fails, we should not lose much, but if it is confirmed, we should gain a lot. In this case, "a lot" is subjective, so you need to have a consensus among people who are interested in the experiment.
Failure of the experiment should not lead to negative perceptions. If we work in the x10 multiplier, then every mistake should bring joy. If we are now mistaken, then tomorrow we did not spend x10 money, time and effort in this direction.
The right to err applies to all areas of our company: technology, product, marketing, sales, business and management.
The above will be relevant for our company for at least 36 months, and preferably always.
