FarmDirect started as a pretty simple pitch: cut out the middleman, farmers keep more money, everyone wins. I said this in a pitch competition once and someone in the audience asked, very reasonably, “but doesn’t the middleman also know things you don’t?” And I didn’t have a good answer. So I went and built one.

The honest version of the question is: a middleman takes a cut, sure, but he also knows which market is paying well this week and which one to avoid. A farmer selling direct doesn’t automatically get that knowledge just because there’s no middleman skimming the top. So when is “direct” actually better, and when is it just “worse, but you keep more of the worse”?

I couldn’t think my way to the answer, so I simulated it instead. Using real tomato price data across 17 Ghanaian markets over about six months, I made two kinds of imaginary farmers. One sells through a middleman, takes a real markdown (grounded in actual research on Ghanaian tomato marketing margins, not a number I made up), but always ends up at whichever market is paying best, because the middleman knows. The other sells direct — keeps the full price, but only picks the best market some percentage of the time, depending on how good their information is. I dialed that percentage from 0 (basically guessing) up to 100 (as good as the middleman).

Then I just… ran it, a lot, and watched what happened.

At low information, direct loses. Obviously, if you’re guessing, keeping 100% of a bad decision doesn’t beat 55% of a good one. But the crossover comes faster than I expected. Once a farmer’s information quality passes roughly 27%, meaning they pick the best market a little better than one time in four, going direct starts winning. And it keeps winning by more the better the information gets, eventually earning close to double what the middleman route would.

That number kind of delighted me, honestly. It means you don’t need to build some incredible market-intelligence oracle for a platform like this to be worth using. You just need to nudge a farmer from “guessing” to “a bit better than guessing,” and the math already tips in their favor.

I also poked at something I wasn’t originally even trying to find: farmers near volatile markets have wildly bumpier income than farmers near stable ones, no matter how good their information is. Good information helps you make better decisions, it doesn’t flatten the market’s actual mood swings. Which means “give farmers more information” isn’t a universal fix, it helps some farmers a lot and helps others navigate a genuinely rougher ride.

None of this was in my original plan. I just wanted to answer one heckler’s question from a pitch competition, and it turned into the most interesting part of the whole project.