Comment by ABrandt | Hacker News Reader

Comment by ABrandt | Hacker News Reader

Expand to the other major chain restaurants and source your pricing from Mechanical Turk (except for the dollar menus of course). I'd take a few precautions to validate the data. For example, you could put up an HIT to get Burger King's entire menu and have 5 people complete it. Compare, lather, and repeat. But I see that as only the first step.

Add a dietary component. First, scrape all the various nutritional values (calories, grams of fat, etc.) from each major restaurants website. They all publish that stuff. Then tie into people's bank transactions through something like Yodlee or finkin(a "Show HN" project from a couple weeks ago[1]). You'll then be able to identify when a user buys food at a given restaurant for a given amount.

With a little tweaking, you could give people an automatic and fairly accurate assessment of their food intake. For example, your current site is telling me that 2 sundaes, a medium fry, and a filet-o-fish cost exactly $5.89. According to the McDonald's site, that meal is a whopping 1420 calories and 20g of fat. People would pay money this. Bonus points for sending customized tips on how to quit eating so horribly based on my data.

You'll have to factor in taxes, different promotions, identical price combos, and more to improve accuracy. I think it could have massive potential though. I've done a little research into a concept like this so feel free to drop me a line if it interests you.

-- [1] http://news.ycombinator.com/item?id=1783823

EDIT: for clarification

> Expand to the other major chain restaurants and source your pricing from Mechanical Turk (except for the dollar menus of course). I'd take a few precautions to validate the data. For example, you could put up an HIT to get Burger King's entire menu and have 5 people complete it. Compare, lather, and repeat. But I see that as only the first step.

I have an API that does this automatically for you, http://houdinihq.com. I'd be happy to help the OP if there's an interest in doing something like this.

Someone's clearly excited about the idea lol

Guilty. I'm most interested by ideas that can produce immediate and tangible results for users while requiring as little behavioral change from them as possible. This is especially important when it comes to something like health--people are always trying to reap the greatest benefits by exerting the least effort.

This is very insightful feedback, thanks!

Do you really think listing what people eat at restaurants/bars/coffees be enough to estimate their calorie intake? This might be my cultural bias talking, but here in Europe it would definitely not be accurate because people eat and drink a lot of "homemade" stuff.

I still consider SoWhat... as a gadget solely created to learn Ruby, but your ideas are intriguing. I'll drop you a line soon (busy schedule this week :) ).

Also, you'll want to factor in availability.

Have you considered using varnish + cassandra? Take care to avoid the thundering herd problem.

I looked up what the thundering herd problem is but still don't really understand how it applies here? Could you explain further? (Sorry if thats a dumb question, not a computer scientist here).

Hmm. If this project ever expands to a larger audience than the one intended here, I will definitely look that up. Until this happens, I'll apply the YAGNI principle. :)

Reply on news.ycombinator.com