FEATURE15 June 2012

‘We didn't set out to build a market research tool’


Brian Tarran talks to Google’s Paul McDonald and Monica Plaza about the origins of Google Consumer Surveys and where they plan to take it next.

Research: How did Google arrive at the idea of developing a survey product?
Paul McDonald:
We were trying to figure out a way to keep premium content online and free for web users. We saw paywalls going up at places like the New York Times and the Wall Street Journal and we decided that we didn’t really like that experience, so we thought about ways people could pay for content without actually having to pull out their credit cards.

The original idea was that we could get them to complete a small task, something that humans are good at doing but computers aren’t, like labelling images for our image search results or picking the best set of search results for a given search query. So what we did was a small trial of this on the Huffington Post for about a month in July 2010. Half were tasks, the other half were market research questions, so we could see what the reaction was to the ideas, what the completion rates were like – that sort of thing.

“We didn’t set out to build the best market research tool; we tried to build the best monetisation option”

It turned out that the tasks were much harder for users to do – they took a lot longer and the completion rate was much lower. We didn’t think there was a viable business there because the market for these tasks was fairy small but we knew that the market research industry was huge. Google does market research everyday, we have millions of advertisers that could benefit from this type of service, so we really focused our efforts there.

That’s why you see a product that’s a little bit different from what you typically get from a market research firm. We didn’t set out to build the best market research tool, we really tried to build the best monetisation option for a publisher. It just turns out that the solution we came up with solves a lot of the needs of the market research industry.

Specifically, what needs do you see GCS filling?
Just like we did for AdWords and Analytics, we are trying to bring professional quality tools and methodologies to the masses. Folks that couldn’t afford this sort of thing before – that’s our initial market.

Monica Plaza: And as much as we are looking to democratise research for the masses, from the customer side there are definitely gaps we fill for companies of all sizes, from Fortune 100 all the way down, simply because they lack the ability to do quick and accurate research: for incidence testing, or to follow up on a larger study or just to do things fast. Larger companies are finding a lot of value in what we’re offering.

PM: I’m sure you are also familiar with the issues going on in the market research industry today, especially around random digit dial polling and the difficulty research firms face in getting a representative sample using telephones.

Landlines are going away – it’s more expensive and time consuming to call mobiles phones and there’s no real solution on the internet for getting a probability-based sample. What we’re trying to do is figure out a way we can do that while focusing on improving the experience of the person answering the question.

Not only do we have a huge publisher network in our AdSense product – with hundreds of thousands of websites signed up to display ads – we also have millions of advertisers that have market research needs, so we can create this marketplace that is sustainable and scalable. We can have millions of people answering surveys on a monthly basis, larger than any other sample provider out there. Google has the scale to do close to probability-based sampling.

Are the AdSense and Google Consumer Surveys networks separate?
The publishers in the GCS network are a subset of the AdSense network.

Can you put a figure on the size of the GCS network?
We don’t have a figure to share with you, but we do see almost two million unique users a month answering surveys.

Google touts the simplicity of its survey tool as one of its strengths, but when it comes to building out the survey offer will simplicity become a limitation? For instance, the one-or-two-question survey – are you wedded to that format or will you have to get rid of it at some point?
The one-or-two-question surveys will hold for quite a while. The research we did before launching this product showed that companies who do a lot of market research rarely need correlation across two or more questions, especially when they already have demographic information. There are definitely cases – longitudinal studies in particular – where we don’t serve a need, but we think that from the end-users’ point of view – the person who is answering the survey – they have a much better experience and are much less likely to either straight-line or provide false or inaccurate answers because our surveys are so easy and quick to do. I think you’ll see more and more providers going this way because it is a much better experience for the respondent.

“We will consider anything so long as it keeps our eco-system of users, publishers and researchers intact”

Monica, at a recent event you were asked about the possibilities for different question types. Where are you with that?
There were plenty of questions in the Q&A after my presentation about what was coming next. In particular, lots of interest in open-ended questions and whether we’d be introducing other kinds of stimuli, like video and audio. The answer I gave is that we will consider anything so long as it keeps our eco-system intact – where the user is still willing to answer the question, publishers are monetising their content and researchers are getting the information they need. But there are definitely question types that, if introduced, would cause the eco-system to fall apart.

PM: It’s also worth saying that the results and analysis interface is just as important to us as being able to serve the survey and get accurate responses back. We do all the analysis for users – we try to pull out the interesting nuggets of information that we call ‘insights’, the statistically-significant differences between sub-populations in the data. With things like open-ended responses, we need to ask: ‘What does that mean in terms of the reporting interface?’ Just showing the user a list of responses isn’t really good enough for us and so we are trying to create new and innovative ways to understand that data without the user having to do a lot of the work.

Questions have been asked about who owns the Google Consumer Survey data. Can you explain?
The person who pays for the data owns it and can do whatever they want with it. Google doesn’t use it for any other purpose.

But aren’t there opportunities for Google to take all the behavioural data it already collects about web usage and tie that with the attitudinal data it gets from GCS?
It depends on what you mean. We already tie cookie data to the survey responses – that’s how we get the inferred demographic data. Some of that inferred data is based on browsing behaviour so we know your interests based on sites that you’ve gone to and seen. That’s a possibility, for sure – to tie browsing interests to answers in the survey. But I don’t think we’ll do much beyond that for privacy reasons.

Could a publisher tie GCS data from readers to their Google Analytics data?
We don’t share the survey data with publishers, but publishers can run their own surveys. Some percentage of their traffic gets dedicated to their own surveys. The ones that they run themselves aren’t charged for, but they aren’t paid for them either. But right now we don’t have any plans to tie survey data back to Analytics data.