A card store, run like an operations problem.
MetaStack is a trading-card storefront I run on about an hour a day, next to a full-time job. The hour is the point: every process that didn't need me was automated until only the judgment was left.
Gold Star seller in the first month, TCGplayer's own badge, earned on shipping speed and accuracy, at a 100% rating. That's the receipt: a store this automated still ships accurately and on time, because the automation stops where accuracy is judged. Check the storefront yourself.
It also set the pricing rule that followed: quotes are built only from Gold Star sellers and Certified Hobby Shops, because a badge I had to earn is a standard I can hold other listings to.
Bought first. Built where there was nothing to buy.
None of these started as a project. Each one started as an afternoon lost to something a computer should have been doing, and a search that came back with either nothing or a subscription priced for a company with staff.
The pricing tools are the clearest case: TCGplayer's catalogue API needs a developer key effectively closed to new applicants, so quotes are built from a community catalogue mirror and the storefront's own product requests. That's unofficial, and I'd rather say so than imply an integration I don't have. A price I can't defend is worth less than none.
An inventory priced before the coffee is done.
Pricing cards by hand is the same lookup performed thousands of times: current market, condition, my floor, list. A person doing it is slower and less consistent than a program every single time, so the program does it, for the whole inventory, in minutes.
What I keep is the strategy: where the floor sits, when to follow a falling market down, and when a spike is real. The machine executes the policy; I only revisit the policy.
Small margins aren't found. They're computed.
Card sellers argue about envelopes on forums, with vibes. I priced every protection option against every order profile (weight, value, count) and let the numbers pick. The answer was worth about eleven cents an order.
Eleven cents is nothing once and decisive at volume: the difference between a hobby that loses money and a store that doesn't. Most businesses have an eleven-cent decision sitting unexamined; the habit of running the math is the actual skill.