At a Lean Startup Machine weekend in New York City, I worked with a team which started out trying to help new entrepreneurs speak and present better. But after finding hypotheses about these entrepreneurs interest in and willingness to invest in improving these skills, we left that and started with something completely different.
We wanted to make a system to help people count calories simply to help them lose weight. The key hypotheses that we wanted to test in the limited remaining weekend time included:
- people have tried counting calories to lose weight but the counting itself was too burdensome,
- we could crowdsource calorie estimates based on photos of food that would be within 10% of actual calories, and
- we could design a prototype mobile app that people could use to take photos and count calories.
Over the weekend, we didn’t take on whether we could economically develop and run such a system, but we believed that on the calorie-estimation side that we would be able to rely on Amazon Mechanical Turk or something like it.
It didn’t take much to confirm that many people have tried counting calories but gave up because of the trouble.
The team worked together to quickly craft and execute a clever experiment to test crowdsourced calorie estimates. Team members found recipes with photos and calorie counts, made an online survey to gather calorie guesses, and used social media to recruit acquaintances (not nutritionists) to estimate calories in the photos. Survey results were quickly tabulated and found that while there was a lot of variance in estimates for each dish, in the aggregate the estimates were within 10% for the total of all the dishes.
For the last hypothesis, I led the work to create a tappable PDF prototype with Keynotopia to try on an iPhone, with screens including these:
People who tried the prototype found it simple to use and believed that — if the calorie counts were accurate — it would be a useful way to count calories.
In the weekend’s competition judging, we were runner-up, which was gratifying.