FEATURE6 December 2011

Waste not, want not

Features North America

Brian Tarran meets John Dick, the CEO of CivicScience, a company on a mission to tease the value out of the millions of online polls people take part in each day.

John Dick (far right) with the CivicScience senior engineering team

John Dick (far right) with the CivicScience senior engineering team

There’s no such thing as a wasted vote. In elections, a vote cast for a hopeless candidate might be considered squandered, but expressing a preference – whatever the outcome – is crucial in a democracy. The only thing that matters is that your vote is counted and recorded.

But often that’s not what happens to the tens of millions of people taking part in the millions of polls littered around the web. Votes are wasted as opinions go unrecorded. News sites are particularly guilty of this. They poll their readers for their views on the topics of the day, but to what end?

It was 2007 when John Dick, a former special assistant to the United States Senate, set up a company with the goal of aggregating and mining all this opinion data to identify trends and patterns in public attitudes – only to discover that most websites weren’t storing any of it.

“We are building this very deep and longitudinal profile of people over time across all of the sites that we are on, and from that we can deliver very directed and targeted questions to specific subsets of that population”

“There was no sense that the information being gathered from these polls was useful,” says Dick, the CEO of CivicScience. And there were problems still with those few websites that were storing polling data. “The questions weren’t being written by researchers so they weren’t particularly sound, and most websites allowed people to come back and vote over and over again,” he says.

Shared learning
Working with engineers from Carnegie Mellon University, Dick and his team set about building technology that would allow the company to plug their own polls into websites. With the agreement of the site publishers, CivicScience would provide the questions and collate the answers.

“We started with a very large library of well-written, professional questions that were designed to still be engaging to the consumer on the page,” says Dick. “We’ve done a lot of research on which types of questions different types of people like to answer; optimal question length; optimal numbers of answers in a multi-choice question, and so on.” He points us to the Pittsburgh Post, a long-time partner of CivicScience. “We found in our research that a user of a website like this will answer up to three questions 90% of the time, and that percentage gets better as they get used to two-or-three-question polls.”

A poll might start with a question relating to a news topic, a brand or an individual – a politician or a celebrity, maybe. Answering that will lead on to a follow-up question that might ask the user to provide some demographic information: income level, age, marital status. Finally, the respondent will get to see how others voted – and this is the pay off. CivicScience offers no other incentive other than the chance to see how you compare to other people.

Clicking through to see more detailed results takes you to the CivicScience website, where users can drill down further, to see how others of similar age, employment or political persuasion voted on any given topic. While on the site, users can answer further questions to build up their profiles, which are tied to an individual using a browser cookie. This cookie can be identified by the CivicScience technology wherever it resides across a network of partner sites.

“We are building this very deep and longitudinal profile of people over time across all of the sites that we are on, and from that we can deliver very directed and targeted questions to specific subsets of that population,” says Dick. “So if somebody shows up on one of these websites and they fit a specific parameter we can deliver a question only to that person.”

The massively distributed omnibus
Dick wouldn’t discuss the mechanics of how it targets questions beyond referring to a technique for “digital credentialing”. But he says users are free to opt-out of having their “anonymous voting history” tracked at any time by downloading an opt-out cookie. “The carrot we dangle in front of the consumer is, you are welcome to not participate, you are welcome not to have your results indexed, but your result won’t be counted and you won’t have access to all the tools we are giving you to compare yourself to other people.

“The goal behind our website is to make all of this information available to the consumer. A lot of the aggregate level data we’re collecting about brands and celebrities – we want this to become a resource to consumers, maybe just to answer a trivial question or to see which brand is trending most popular.”

“Too few people are participating in the opinion sharing process. People are busy, there are more strains on their attention. I might have the time to complete a survey, but I’d rather spend it with my kids”

But CivicScience is still a business, and businesses need to make money. CivicScience does so by selling custom questions. “Think about it as a massively distributed omnibus survey,” says Dick. The data the company provides free to consumers is “fairly raw information, it’s unweighted”, says Dick. “Customers pay us for a much more sophisticated view of that information. Weighting, time-trend analysis, lots of decision tree capabilities. Our data is well-structured because we are limiting things to only single answer multiple-choice questions.”

Poll overload?
CivicScience has delivered over 110 million polls in the last 15 months. Add that to the mass of DIY surveys, online panel requests, website pop-ups and telephone polls people are asked to complete and we put it to Dick: Are we not in danger of over-polling consumers?

He disagrees. It may be that the research industry is over-surveying those relatively few individuals who have opted to join survey panels, but that suggests a deeper underlying problem. “Too few people are participating in the opinion sharing process,” says Dick. “People are busy, there are more strains on their attention. I might have the time to complete a survey, but I’d rather spend it with my kids. But give me a question or two to answer somewhere I’m already travelling on my digital footprint and I’ll do that.

“We’re not invasive in the way we present questions to people. We are in a discreet location on a website they visit everyday, and we do a lot of work to make sure the questions we deliver are relevant to people.” More importantly, CivicScience has hit on a way to make sure the opinions people share through website polls are counted and recorded – and with the right analysis, they have a chance to make a difference.

“We want people to understand that we are collecting anonymous opinion data from the world and we’re sharing it with decision makers,” says Dick. That’s better than a wasted vote.

0 Comments