(live results disabled for the moment)

The experiment is live at
nfrom1to10.appspot.com »

See the ongoing results »

Write-ups about the project »

Pick A Number is an experiment to determine what numbers people prefer, out of the range from 1 to 10, inclusive. Common wisdom suggests that people prefer specific numbers, especially 7. This experiment is an attempt to verify that belief, and is so far very successful, with over 300,000 votes. The experiment garnered a very strong response, reaching the top of Hacker News (twice) and Reddit.

The user is prompted to “Pick a number from 1 to 10” or “Pick a random number from 1 to 10”, with the wording and UI chosen pseudo-randomly.

One of the metrics being tracked is the time to first picking a number, then submitting a number. This allows for determining if the user thought about the choice, or submitted it quickly.

Being a bit of a pedant myself, I made sure to qualify the acceptable number range. This, combined with validation built into the client-side code has not prevented people from circumventing the page and submitting alternate numbers.

I considered adding server-side validation, besides ensuring integers, but it's interesting to see what people submit. Sample 'subversive' numbers include 0, -1, 100, 42, 1337, and 9999.

Once the user chooses a number, they are presented with two followup questions for determining their method…

…and thanked for their participation.

While initially developing the experiment, I had difficulty settling on a method for having the user submit a number. This resulted in presenting four different methods, adding an additional dimension.

The four UIs are:
  • input - standard text input
  • radio - numbers arranged on a line
  • select - number line revealed on hover
  • slider - slider along line

For a description and breakdown of the results for each UI, see the Pick A Number results page.

The look-and-feel is intended to be simple and clean, focusing purely on the experiment, with a little bit of extra personality to invite participation. The 'For SCIENCE!' banner works especially well as a rallying cry of sorts. Various sites linking to the experiment use wording like 'do it for science'. Also, other experiments use similar banners and slogans.

The experiment is ongoing. The first iteration did not have the follow-up questions, so the additional data is still being collected and processed.