How we decide what to build

This post is scheduled rather than published, kept as an example of what that looks like. It will appear on its own when its date arrives, with nothing running.

Two questions

Every feature here has to answer both:

1. Is it true? Does the screen say only what we actually know? A number we cannot source, a badge we cannot explain, or a claim about somebody we did not check does not ship.
2. Whose side is it on? A feature that helps one party by quietly costing the other is a feature we have to be able to defend out loud.

Things we keep deciding not to build

• Activity monitoring. No screenshots, no keystroke counting, no idle timers. Hours here are what a freelancer says they worked and a client approves.
• Bidding. It sets a price by auction, and an auction on labour has one direction.
• A quality score. We compute a completeness and reliability score and we are careful to call it that. Ranking people by how good they are, for money, on a rubric they did not agree to, is a different product.
• Anything that makes leaving expensive. The flat fee is what makes this possible: it costs us nothing when two people who found each other decide to work together directly.

The test we use: would we be comfortable if the person this affects read the reasoning?

What we do build

Whatever removes a question a client or a freelancer would otherwise have to guess at. Most of this product is that.