Why I'm building Forkbell

I have logged food in most of the big apps. Each time it went the same way. Search "roti". Scroll past entries reading 71, 85, 104, 120 and 297 calories for what is, in my kitchen, the same roti. Pick one. Feel vaguely wrong about it. Fix the portion because "1 serving" means nothing. Two minutes later, one meal is logged.

By the second week I was skipping dinner logs. By the third I had quit. Ask around a gym and you will hear the same story, with small variations, from almost everyone who has tried.

The problem is the food, not the discipline

Calorie databases were built from packaged food. A barcode gives one exact answer. Home-cooked food has no barcode, and most of what people in India eat is home-cooked. So the database fills with entries typed in by other users, and the app makes you choose between them.

That is not a minor quality issue. Those entries are unverified, they describe someone else's recipe, and they skew low — people estimating their own dinner tend to be generous to themselves. The mechanics are in why two apps give the same roti different calories and how to judge a calorie app's food database.

Serving sizes make it worse. Nobody measures dal in grams at the table. People eat a katori of dal, two pieces of roti, a spoon of ghee. An app that wants grams is asking for a conversion at every meal, forever.

And then there is speed. Logging that takes two minutes gets skipped. Logging that takes three seconds becomes a habit. Every extra tap is a reason to stop.

What Forkbell does differently

You type what you ate, in your own words. "2 roti, dal, katori curd." That is the whole interaction.

Behind it, Indian dishes come from the Indian Food Composition Tables published by the National Institute of Nutrition, with household servings attached. Generic items come from USDA FoodData Central. Packaged products come from Open Food Facts and manufacturer data. Every number in the app shows where it came from, and estimates say they are estimates.

Your workout goes in the same log. Sets, reps and weight sit next to your meals and your weight trend on one screen, because for most of us those are one project rather than two apps — the argument in why food and training belong in the same log.

The part no app can solve

Honesty about this matters more than any feature claim.

No database can price the dal you made on Tuesday, because that dal contains however much oil your household used, and cooking fat can change a dish by 40 per cent. The numbers in how much oil is actually in your curry are not an edge case; they are the central problem of home-cooked tracking.

What an app can do is make it trivial to calculate a dish once from raw ingredients — including the oil — and reuse that number every time you cook it. That is the method in how to calculate the calories in a recipe you cooked, and building it in properly is most of what makes a tracker work for home cooking.

Anything that claims to have solved this without you supplying the recipe is claiming something that is not available.

What it will not do at first

The first version is small on purpose. No social feed, no meal plans, no barcode scanning, no photo estimates. Those come later, once the core loop is fast and the food data is trustworthy.

There are reasons beyond focus. Barcode scanning only helps with food that has a barcode, which is a minority of an Indian diet — what barcode scanning can and cannot do. Photo estimation identifies food well and cannot see the oil — how accurate is calorie counting from a photo?.

Shipping a small thing that works beats shipping a large thing that is slow.

Writing this in public

Everything on this blog is written while building the app, which means two things.

The nutrition articles cite their sources and label estimates, because a site that tells you to demand that of your food database should hold itself to it — the rules are in about Forkbell and how we get our numbers.

And the comparison articles are about criteria rather than verdicts. I am building a competitor to the products discussed in MyFitnessPal alternatives, compared on what matters, so the honest thing is to give you the tests and let you run them, including on Forkbell when it exists.

Where this is going

Closed beta from the waitlist in October. Public launch on iOS and Android in November. Until then I post most days about what I am learning, mostly about food data, and send one email a week to the list.

If you have quit a calorie app for the reasons above, join the waitlist. And reply to the welcome email with what you use today. I read every one.