Monday, July 18, 2011

Qriously - real time sentiment polling, but do the results make sense? (updated)

Qriously logo

Qriously is a relatively new service that works by replacing ads with questions in smartphone and tablet apps. It describes itself as “a service for measuring location-based public sentiment, in real-time” and calling itself “pre-release alpha” it is still some way from being fully product ionised.

People viewing the ads questions can quickly stab their answer – usually by choosing between two absolute options, positioning an analogue slider between two options, or rating on a scale of 0–5 stars.

Having read Jemima Kiss’ piece about Qriously in this morning’s Guardian, I signed up to experiment with the limited about of free credit which you can use to try out the service. An obvious question – and before later in the day when “Yates of the Yard” threw his resignation into the big bin of News International-ended careers – seemed to be:

Will James Murdoch resign before the end of the summer?

Within a few minutes Qriously had hit the phone banks pushed the question out to apps using their service and 50 people had responded.

Qriously result screen shot

Of course, the 50 people were spread right across the world. A pretty dopey question in retrospect. What proportion of people out there even know who James Murdoch is? So I asked.

Qriously result screen shot

A similar proportion (but not the same people) think he will resign! Ummm ...

(For the purpose of this post, we need to leave statistical accuracy firmly to one side. A sample size of 50 people worldwide, all users of Android or iPhone smartphones, and awake at the time of the survey, isn’t the most robust methodology, but it was the smallest sample size I could use to eek out the free trial.)

Questions can currently either be distributed to users worldwide, or can be limited to US users (ie, US IP addresses). Worried about the quality of result, I came up with a US-centred question that most people in the US would be able to answer:

Will Obama be re-elected for a second term? Yes / No.

Qriously result screen shot

A third of people thought he would.

But when I ran the question again, but swapped around the order of the responses – No / Yes – a third of people though he would not!

Qriously result screen shot

Or put it another way, no matter which way round you ask that question, two thirds of people in the US prefer the right hand option when asked about Obama’s chances of re-election! Do smartphone users hold handsets in their left hands and prefer to stab the right hand side of the screen with their right hand index finger? Or maybe surveying Americans in the middle of their night is asking for trouble.

Qriously map of results

Qriously has a great UI, and the real-time aggregation and display of result is entrancing. It can even provide a breakdown map of responses by country (or state for US-only polls).

As the number of smartphone applications supporting Qriously spreads, the simple charging model (you just pay for a certain number of answers to be collected) will be attractive to small organisations that could never afford to commission their own poll or to buy questions in a national survey. It works out at about US$0.10 per worldwide answer, rising to US$0.12 for US-only answers.

It will also appeal to media companies who impatiently want a near-instantaneous method to gauge public reaction to a breaking story.

However, until Qriously improves the ability to tie down questions to specific geographies (a UK and Ireland option would be good) and until the demographics of smartphone users in those geographies are really well understood, opportunities to use the service to generate serious results may be limited. The ability to target specific languages (a large number of Chinese users replied to the James Murdoch poll) might prevent spurious results too.


Qriously were in touch to say that real (non-demo) users can much more accurately target questions at geographic regions (as well as aiming at particular genre of apps).

And when repeated later in the day with a larger sample size, the strange results for Obama (where the right hand option got two thirds of the votes whether it said ‘Yes’ or ‘No’) disappeared. So it is likely to be down to the small sample size. Right-hand bias panic over!

A more sane set of Obama results

No comments: