Methodology
YouGov's methodology is to obtain responses from an invited group of Internet users, and then to filter these responses in line with demographic information. It draws these demographically-representative samples from a panel of more than 150,000 people in the UK.
As YouGov's online methods require no field-force, its costs are lower than competitors that employ traditional face-to-face or telephone methods. YouGov has exploited its cheapness and speed to conduct more polls for newspapers and television programmes than any other organisation. Its media clients include the The Daily Telegraph, The Sunday Times, The Economist and Sky News.
Panel members, volunteer members of the public, are credited with 50p-£2 for each survey they complete (they are sent a cheque when the amount accrued reaches £50). In addition there is a monthly prize survey, the completion of which enters the member into a prize draw, and other occasional prize surveys.
Accuracy
YouGov has contended that its opinion polls in recent UK elections, e.g. the 2001 general election, have been consistently more accurate than traditional opinion pollsters who repeatedly over-estimated the Labour vote.
This pattern was repeated during the 2005 general election campaign, when most traditional polls reported Labour's support in the range 38-41%, compared with the 36% it achieved on polling day. In contrast, YouGov's nine polls during the final three weeks of the campaign all showed Labour on 36 or 37%, although NOP (published in The Independent) were the most accurate pollster in 2005.
Criticism
Critics argue that, as not all of the public have access to the Internet, its samples cannot accurately reflect the views of the population as a whole. YouGov counters that they have a representative panel and they are able to weight their poll/surveys appropriately to reflect the national audience that they are aiming to poll.
It is a function of their internet panel approach that YouGov isn't able to pick up turnout factors to the same degree as other pollsters and they exclude it from their methods. However, traditional polls use widely differing methods to take account of turnout, and these produce equally varied "corrections" to the raw data. No consensus has emerged as to what, if any, "correction" has greatest validity.
http://en.wikipedia.org/wiki/YouGov