Over a million developers have joined DZone.

Study explores the success of Tesco crowdsourcing

DZone 's Guide to

Study explores the success of Tesco crowdsourcing

· ·
Free Resource

One of the most enduring examples used to highlight the potential wisdom of the crowd comes from one of the earliest explorations into MOOCs that saw Stanford university offer a computer science module online.  It emerged that some 400 or so of the MOOC participants were able to beat the official students in the same class at Stanford.

It’s a nice example of how even the finest minds can, under the right circumstances, be beaten by the crowd.  So it’s interesting to read a recently published paper that explored the foray into crowdsourcing taken by British retailer Tesco, who wanted to see if the crowd would be better able to predict shopping behaviours than their existing methods.

Now, we have to bare in mind here that retailing is already awash with data on shopping patterns.  Through the use of things such as loyalty cards, retailers have never had more data on what we do, when we do it, and even where we do it.  Nevertheless, Tesco wanted to see if the crowd could make better sense of this data than their algorithms could.

Tesco took a similar, competition based, approach to that of Netflix, who famously asked the crowd to improve its recommendation algorithm.  The crowd were granted access to real world data from around 100,000 Tesco customers over a year long period, and were asked to predict both the data and the purchase amount of each customers next visit.  Heady stuff.

The competition attracted 537 entrants, who between them submitted over 2,000 different models, with the winning entrant a statistics professor from Moscow.  The winning entry managed to improve the predictive model used by Tesco by around 100%, but what made the project work so well?

Interviews with participants revealed a number of factors that the authors believe were key to the success of the project.  Firstly, the competition took advantage of an existing crowdsourcing platform, so could tap into a ready made community of willing minds rather than having to build one of their own from scratch.

The paper also suggests that getting motivation right was crucial.  The Tesco competition provided entrants with a live leaderboard so they could see in real-time how their efforts compared with their rivals.  The competition did however provide some interactive facilities to allow entrants to collaborate with one another, which the authors believe was crucial in adding an element of fun to the competition, as well of course as providing a means for learning from others.

As we’ve seen with other studies exploring the motivation of crowdsourcing participants, money was not really all that attractive in this instance.  The competition used an interesting model whereby the company took ownership of the top three models, with prize money thus divided equally between each.  The aim was not only to secure a wider spread of intellectual property but also to encourage greater collaboration between participants.

As with the previous studies however, the entrants were much more concerned with testing their smarts than they were about the pot of gold at the end.

Original post

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}