How do you make decisions when you have incomplete information?

How do you make decisions when you have incomplete information?

For someone coming out of a PhD, that's the hardest adjustment. Research trains you to keep gathering until you're confident. Running a company doesn't give you that option, because the cost of waiting is usually higher than the cost of being wrong.

And the decisions come from every direction. A technical question about a method you haven't worked with. A go-to-market question nobody on the team has faced before. A strategy call that depends on things you won't know for another year. When you're building something new, there often isn't a playbook.

One of ours was whether to build our own HTE lab.

For a software company that's a strange decision. Hundreds of thousands of dollars in equipment, plus chemists on payroll, and every dollar spent there is a dollar not spent on engineers. No spreadsheet made it look obviously correct.

We did it anyway. And it took a while before the real reasons became clear.

We became our own users. Every day our team runs campaigns on our own platform, and that surfaces things a customer interview never will. The small inefficiencies. The repetitive steps you only notice when you're the one doing them for the tenth time. We feel those ourselves, and they go straight into the product.

And then there is the HTE data our chemists generate - giving us a true competitive edge.

Neither of those was in the original justification. We found them by doing it.

So, the rough approach:

If you don't have the information, go and find it. Talk to users, read what exists, ask someone who's done it before. This is faster than people assume, and skipping it is laziness dressed up as decisiveness.

If you can't find it, make your best call and accept that you might be wrong. Some questions only get answered downstream of building the thing.

And when you're wrong, be fast about it. The cost of a bad decision is rarely the decision itself. It's the months spent defending it.

The only unacceptable version is making the same mistake twice.

‍