The Perils of Predicting the Internet
Estimated reading time: 3 minutes, 41 seconds
Forecasting the future is a difficult path to tread. Predictions are often met with a healthy dose of cynicism, and there is no lack of critics if your carefully calculated forecasts turn out to be wrong. And yet people crave to know what the future holds. Despite its perils, forecasting is one of the most interesting parts of my job, and why I love being an analyst.
The basic premise of forecasting is that stuff happens in repeatable patterns. By looking at historical trends of what has happened in the past, it’s possible to predict what will happen in the future. Therefore, if March has been the biggest month of the year for ISA searches for the last three years (and it has) then it makes sense that next March will be the biggest month for ISA searches in 2013.
The essence of these types of predictions relies on seasonality. A lot of online industries have clear seasonal peaks, from finance (January and March) to travel (January and July) to retail (December). Forecasting becomes that much harder when you don’t have seasonal trends to rely on.
Which brings me nicely to reality TV shows, The X-Factor is a guilty pleasure of mine. In the opening weeks it’s fun to see the mixture of talent and torturous embarrassment of good and bad contestants respectively, but in the final weeks it is always a challenge to try and predict who will be the eventual winner of the show. I have always thought of the Internet as a barometer of the mood of the nation, so it stands to reason that if one act is consistently searched for more than any other, more people are likely to vote for them.
This backfired spectacularly a number of years ago for a colleague of mine in the US who predicted actress Stacey Kiebler would win the second season of Dancing with the Stars. What he failed to take into account was the extraordinary volume of searches that were not wholly representative of the voting audience: ‘stacey kiebler topless’ being amongst the most popular. This led to the creation of the “Stacey Kiebler Correction Coefficient”, something we still use today at Experian Hitwise to eliminate unwanted noise from search projects such as these.
Two years ago I predicted One Direction would win The X-Factor. Not only were they the most searched for act of the 2010 finalists, they were also attracting the right demographics online of people who would have a high propensity to vote and ultimately buy their music. You can see a more detailed account here of my reasoning, but on the night itself it was Matt Cardle who walked away with the recording contract, and One Direction came in third place.
To this day I am still not quite sure why there was such a disparity between the online data and the final result. All I can say is that predictions are not perfect and no matter how good your forecast model is you are never going to be right all of the time. However, I do take some solace from the fact that since then One Direction have gone on to have two hit albums, broken America, sold 15 million records and been heralded as the resurgence of boy band culture. Matt Cardle by comparison has had just 2 million sales of combined album and singles.
When we look at the search data, One Direction is 40 times more popular online than Matt Cardle, and consistently account for 1 in every 2,000 searches conducted in the UK.
Among other accolades, One Direction were the most searched for band of February 2012 and had the third most searched for lunch box this August before kids went back to school (Lego and Batman being first and second respectively). Not only is their music popular, their merchandising potential is extraordinary. And it was all there in the data! So although I didn’t get the prediction right on the night, long-term I think the search data has been proven true.
Which leads me to this year’s X-Factor predictions, Union J are the clear leaders in terms of online searches, but having been burned once before, I am plumping for James Arthur – who is the second most searched for finalist and saw a 30% increase in searches in the last week alone. The momentum is with him – let’s see if he wins.
Until next month.