Founding Director at YouGov
Joe was the first employee at YouGov joining the two founders in 2000 when the company was founded. He joined Yougov from Ipsos, a UK-based competitor, where he was responsible for building the infrastructure to run representative online surveys for political and social research. Joe enjoyed 20 years at YouGov running various research desks across sectors and countries, creating some of YouGov’s leading products such as the BrandIndex and the foundations of the Cube. Joe currently runs Deltapoll, a leading UK public opinion consultancy. Read moreView Profile Page
Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.
Joe, can you share some background to when you first joined YouGov?
I joined YouGov in 2000, over 20 years ago. The work that I had previously been doing was as a graduate trainee and then, latterly, as a research executive at what, at the time, was called Ipsos-Rsl. It had previously been Research Services Limited; it was then Ipsos UK and it is now part of Ipsos MORI. That was a traditional research agency that specialized in all the kind of work that a traditional agency specializes in. It did a lot of face-to-face and as well as a lot of telephone surveys.
At the end of the 1990s, however, the internet was seen, by many people over a number of industries, but particularly in research, as the next big thing. Ipsos had a reputation for innovation in the area of technology. It had done a lot of interesting things with laptops, back in the 1980s and adapted those for face-to-face research. It wanted to get into the world of internet research. But, as is so often the case, the next big thing is not particularly well understood by a lot of people. Various people at Ipsos wanted to do internet research but didn’t know what internet research looked like.
I was involved with developing internet-based research, at Ipsos. Initially, that focused on a whole range of different subjects – everything under the sun – except political work. It was political work that I was particularly interested in, having graduated in politics. The work that Ipsos was doing, crucially, was not nationally representative. They were doing a survey, for instance, of washing powder, among a specific group of users of that particular washing powder who had been recruited, rather than the national picture, to find out what they thought of it. It was really dipping a toe in.
Ipsos didn’t really take it particularly seriously, at least not at that time. One of the reasons for that was that they had an awful lot of work that was done via face-to-face and the telephone so they had vested interests in those areas. They weren’t willing to sacrifice a lot of that work, on the altar of online research, just as an experiment. That’s why someone, fresh out of university, as I was, but with a knowledge of the internet, was placed really at the forefront of their work.
When YouGov was being thought about, before it existed as any kind of actual entity, part of the offering that YouGov would provide was a polling element. I was contacted by the two guys, Stephan and Nadhim, who were setting YouGov up, to come on board from the very beginning and head up that polling side of YouGov. Back in those days, YouGov covered a whole host of different things and it presented itself as an e-democracy website, focusing on politics with various different elements; everything from an online bookshop to a recruitment consultancy for government jobs. Polling was just one aspect of YouGov at that stage and that’s the bit that I was involved in.
My role was to build up the company’s technical infrastructure, essentially to get it into a position where it could use the internet to conduct, not just research, but representative research; representative of Great Britain and the UK, for their adult population. Not just for elections, but for everything else.
Can we walk through how you would, typically, work with a client for custom research at, let’s say, YouGov, in the early days? Let’s say I am a big brand, like Pepsi, and I want to know more about my UK consumers and their habits, what is the process of conducting and organizing that online market research?
It’s really difficult to summarize things in terms of a typical survey because these things are so different. It would be like going to a car dealership and saying, I’d like a car please; there are lots of different cars that are available that suit different purposes, at different times. It’s exactly the same with research. We tended to work in terms of general approaches. We had our own proprietary technical infrastructure, which allowed us to conduct surveys among different populations. But there were key questions that we would always ask of a new client. Who did they want to speak to? What type of people did they want to speak to? How many people did they want to speak to? What questions did they want to ask? How many questions did they want to ask? What kind of output were they looking for? Were they looking just for a set of results tables or were they looking for us to explain those results? Were they just looking for a data file? It’s hugely variable. One client may want one thing, at one time and it may be very different another time.
We, as researchers, would work with the clients to really try and answer those questions. Some clients will have very clear and specific answers to those questions. You can ask them and they will be able to run through the answers very quickly. Other clients – sometimes even the same clients but at different times – would need much more guidance. The whole process would become much more of a discussion, an ongoing discussion in many cases, with the client, about what their overall aims of the survey were and how that could then be achieved through a series of questions and the type of analysis and the type of people that you go to. It really does cover a full range.
In some instances, you will have people come to you with an entire, oven-ready questionnaire, all ready to go, knowing exactly who they want to speak to and they just want to run it through our service. On other occasions, you will have clients come to us with just the vaguest of questions and we deal with everything in between. It’s all about providing the most appropriate response for those clients.
At the heart of the process, however, is this idea of active sampling. Internet research has been around for a lot longer than representative internet research has been around. The early days of internet research had pop-up ads and things like that simply saying, do you purchase socks? Do you wear socks? Come onto our survey and answer questions about socks. You would get tens or hundreds or thousands or tens of thousands, hundreds of thousands of people replying to these surveys. But they themselves were replying because they were interested or because they wished to take part. Sometimes, they could take part as many times as they liked and anyone could take part. That’s what is called passive sampling. That kind of passive sampling does not give you the kind of controls that active sampling provides.
Active sampling is the basis for all representative internet research. Active sampling means that you can contact only the people that you want to take part in a survey and only those people are allowed to take part and, crucially, they are only allowed to take part once. It means that you contact the right number of old people, young people, men, women to be representative of the country, as a whole, rather than just have people pile in. That forms the basis of all of the work that we do, even if the precise nature of the work then varies along a big spectrum.
Is that the advantage for YouGov that they can actively sample from the panel, to serve those clients?
The advantage that a company like YouGov has is the size of its panel that’s available. YouGov are not, by any means, alone in being able to offer a large panel. Various other companies, such as Toluna, Donato, Civic, can all provide a panel of British respondents but the particular advantage that YouGov has is that its panel is, firstly extremely large but also and crucially, the amount of data that they hold on some respondents goes back a very long way. As I said, YouGov started in 2000 and, from the very first day, we were collecting data on respondents and then putting that in our database, so that it could be queried, as any data in a database could be. What that means in practice is that you, for instance, are the British Election Study, the largest academic, political study of its type in the UK. They first ran a survey, with YouGov, and I was instrumental in working with them on this, so I remember it very clearly, back in 2001; the 2001 General Election.
YouGov collected a small sample of around about 3,000 respondents, who took part in the 2001 General Election and they voted and that data was recorded. Since then, we’ve had an awful lot of political events such as local elections, national elections, General Elections, as well as referenda. As time has gone on, all that data has been collected and we’re now at a stage where, I believe, 30,000 take part in the British Election Study, not just at elections, but at regular times in between elections. All that data can be linked together and, for some of those respondents who are still taking part, you can link data going back to 2001; 20 years’ worth of individual level, tracked data. That kind of offering cannot be provided by anyone else because the Ipsoses of this world only started getting their panels together and in order in the late 2000s. Even in those cases, in some instances, for some companies, they didn’t link the data up and they don’t have that historical data. A company like YouGov does have an advantage in that respect, because they have so much historical data. That’s not always going to be essential for a lot of clients but, in some cases, it can be very useful.
How do the likes of Kantar or Nielsen – with the TV or consumer data that they collect – approach active sampling versus what YouGov would do?
Essentially, everyone now approaches the sampling of surveys in, fundamentally, the same way. They all have access to panels. Sometimes, that will be their own panel. At my own company, Deltapoll, we don’t have our own panel; instead, we use proprietary panels for other companies. We will go out to specific panel companies, be it in the UK, Europe, America or, indeed, anywhere in the world, and we will use their panelists to drive respondents to our surveys. In the case of a company like YouGov, they have their own panel. Kantar has a small panel and they supplement that with other panelists from third parties. But, essentially, it’s the same process. You access panelists and you offer targets for quotas; in other words, the number of people in different groups that you want to survey, be it based on gender, age or a whole combination of different things. Those people are then contacted.
In the old days, they would be contacted simply via email. Nowadays, it can be via email, via push notifications on mobile phone apps or a whole host of different things. They are then invited to take part in the survey, in a given period of time. They take part in that survey; their data is collected. Sometimes, that data will be linked up to previous data they have given. On other occasions, it won’t be. Those respondents are usually – although not always – incentivized for taking part in that survey. The incentive varies from panel to panel but it’s usually cash-based in some respect. It might be vouchers; it might be a prize draw but certainly, it will be something you can spend.