Keith MacDonald 6:05 pm on May 2, 2012 Reply
Tags: , , , report interpretation   

For those who saw our booth at one of the recent conferences in which we participated, hopefully you noticed the big bin o’ cash sitting on our table.

Keep the Cash contest at the Unilytics Booth

Keith at the Unilytics booth with the bin of cash

 

We launched the “Keep the Cash” contest to open a conversation with conference attendees and prospective clients and partners and so far it has proven to be a huge success!

Crowd around the booth

Keep the Cash contest drew a crowd at the booth

There’s been a lot of interest not only in winning the cash but in details about the contest, especially at last week’s eMetrics Toronto.  Here’s a quick analysis rundown:

  • Actual money in the bin:  $362
  • Winning guess:  $352
  • Number of entries:  61
  • Lowest guess:  $96
  • Highest guess:  $2000
  • Average guess:  $365
  • Median guess:  $275 (half of guesses were less than $275, half were more)
  • Most common guesses:  $225 and $400 (guessed by 3 people each)
  • Average of variances (between guess and actual):  $184

Despite the wide range of variance in the guesses, the average of guesses was only $3 off actual and would have won the contest!  This is particularly interesting given the median of guesses was $275, significantly lower than what was in the bin.

The difference between the average and the median clearly illustrates the impact of the high-dollar-amount guesses on the raw data set – they significantly increase the average guess.

  • Two guesses were within $20
  • Only 13% of guesses were between $300 and $400
  • 87% of guesses were between $100 and $500
  • A quarter of guesses were either less than half of the actual cash or more than double the actual cash
  • The highest frequency of guesses was between $100 and $200

Distribution of guesses

Distribution of variances

Applying a 90% Winsorization to normalize the data and reduce the impact of outlying guesses:

  •  Average guess:  $339
  • Average of variances:  $157

Normalizing the outliers significantly reduces the average guess to less than actual.  Otherwise the impact of normalizing the data is relatively unnoticeable – the data set is relatively evenly distributed.

So what are the take-aways?

Most participants in the contest have a poor visual concept of money.  This isn’t at all surprising:  visual models of money have almost no value in our day-to-day lives (money is a finite measure of value so a subjective visual measurement is rarely worthwhile).  Practically speaking, we crumpled the bills and threw a couple of $5 notes in with the $1’s to make the contest more difficult.

Despite that however, most participants made a reasonable guess of between $100 and $500 dollars and almost a third of participants guessed within $100.  Perhaps most interesting is that the average of guesses, including the outliers, was almost bang-on.

For those of you interested in doing your own analysis or double-checking ours, the Excel file is posted here - Keep the Cash Analysis – Excel Workbook

See Dean Abbott’s analysis of the contest data in his blog: Another Wisdom of  Crowds Prediction Win at eMetrics.

We congratulate the winner, George Ling, on his successful guess and for those of you who didn’t win, we change the amount in the bin for each conference.  :-)

Good luck next time!