So how are we all? Enjoying the two days of holiday this week? Well some people have been hard at work it seems, most notably in the offices of the Official Charts Company who sent an unexpected missive to the recipients of the weekly charts data earlier this afternoon:
DECEMBER 28th: Last night, Millward Brown discovered a bug in the weighting software used to compile the charts, which has affected a number of positions in the charts published on Sunday December 26. As a result, the OCC has decided to re-run all of this week's Official Charts. In relation to the Top 40 Singles and Albums Charts, the errors are minimal. But if you wish to correct any charts which you publish online, or simply use the attached charts for reference purposes, please feel free.
Millward Brown is conducting a thorough review of the processes and systems in light of this error. OCC and Millward Brown apologise for any inconvenience caused.
Now a word on what the “weighting system” mentioned above refers to. The average market share of each of the retailers reporting data into the chart system is sometimes used to correct anomalies in the data set to compilers Millward Brown. This arrangement dates from the mid-90s when on two occasions just a few weeks apart, one retailer submitted duplicate data following a system error, forcing an embarrassing re-publication of the affected chart countdown when the corruption was discovered earlier in the week. To forestall these kinds of problems in future, the chart rules allow for retailers to occasionally fail to submit data for certain days, the weighting rules used to instead calculate what their sales figures for those periods were likely to have been and those numbers used in calculating the bestseller tables. This, incidentally, is one of the main reasons why music charts don’t reveal the actual sales tallies for all the singles in the countdown. Aside from the commercial sensitivity of such information, the numbers used to rank singles and albums often contain fractions of sales thanks to upweighted calculations. To declare that such and such a single sold 53,424.6 copies would be completely meaningless to the general public, hence you normally never get to see the numbers behind the chart positions.
This is also why when I quote sales figures on Yahoo! Music I tend to talk in vague terms unless the numbers have been formally announced by the Official Charts Company themselves. Apart from the fact that the data isn’t really mine to reveal, however, I might have come by it, declaring unequivocally that a single has sold an exact number is fraught with danger. For all I know it may not have done in such a precise manner. Steering clear is the best way forward.
So what went wrong this week? Well in the charts published last weekend there should theoretically have not been any weighting applied. The cover sheet for the chart declared “All expected multiples data was received and available for use in compiling these charts”. I’m sure Music Week will have the full story next week, but it seems entirely possible that some upweighting was applied where it was not required and hence some figures were inflated - particularly when you dig down and note how the revised chart differs from the original version.
So what changes took place? Well as it turns out just one single, in particular, is affected. It is the X Factor charity single Heroes, originally Number 18 on the published chart, is relegated all the way down to Number 22. In consequence the singles from N-Dubz, JLS, McFly and Jessie J are all promoted one place. That would kind of imply that the error in the data came from a retailer such as a supermarket who would theoretically only be stocking the X Factor single as a one-off. If there was some confusion as to what sales data they had submitted then it is I guess entirely possible that some upweighting was applied when it was not required.
Further clues as to what actually happened can be gleaned from the changes to the album chart. The Ellie Goulding and Barbra Streisand albums swap places at 23 and 24 respectively whilst Tinie Tempah and Rod Stewart also swap around at 26 and 27. Cee Lo Green moves from 36 to 34, relegating The Beatles’ 1967-1970 down to 35. The Beatles’ 1962-1966 drops 35-38, promoting Mumford and Sons and Biffy Clyro up a place as well. Finally Take That drop to 40 with The Circus, swapping places with Lady Gaga.
Spot a pattern here? Rod Stewart, The Beatles and Take That are the artists whose sales appear to have been overstated in the original chart rundown – all of these acts who one would expect to be selling in supermarkets rather than specialist music shops. The fact that lower down it is Daniel O’Donnell and the Chelsea Pensioners who slip places only reinforces that suggestion. The data from one (or maybe several) supermarket chains got accidentally scrambled and forced this extraordinary correction.
I’m sure this will all come out on Monday and to the relief of many no significant chart positions were affected, leaving this to be a minor wrinkle in the grand scheme of things. Imagine the horror if it turned out a single had been promoted to Number One in error. Expect questions to be asked though, the OCC pride themselves on supplying one of the most rigorously researched and scrupulously accurate sales reports of any industry in the world. The fact that they have been required to publish a correction, and in the middle of a holiday as well, will cause a fair number of red faces on the South Bank this week.
UPDATE 30/12/10: As the revised chart permeated into various sales databases over the past 48 hours, it has become possible to get more of a picture as to what changes were required. Many albums, several of them in the higher reaches of the chart, have had their sales totals for the week revised downwards, as much as 20% in some cases. Whilst Matt Cardle's status as the performer of the Number One single was never at risk, it too has also seen its reported sales slashed, to the extent that the single actually slipped from 2 to 3 in the year to date rankings and raising the possibility that he may not after all have the legs to overtake Love The Way You Lie as the biggest seller of the year.
Now I stress that much of this is guesswork, but what seems likely to have happened is that data from at least one supermarket chain has indeed been revised in error. Thinking it through, the supermarkets would have reported strong sales for the whole of last week and then exactly zero on Saturday - it was Christmas Day and they were closed after all. Rather than this being flagged as an expected sales pattern, the computers compiling the chart interpreted this as "failure to report" and applied an appropriate level of upweighting to sales from the rest of the week - quite a large chunk given that many albums have had a fifth of their sales wiped out. This was such a serious error that you can understand totally why an extraordinary midweek correction was issued. The number of actual chart positions affected may have been few and far between, but such an error in sales totals simply could not be allowed to remain on the database.
[The official and more detailed explanation that the OCC communicated to the industry rather than media partners reads: "The cause of the error was a bug in the weighting software used by Millward Brown that only applied when a bank holiday falls on a Saturday, causing over-compensation. (Last weekend was the first time this situation has occurred with the current software). Millward Brown is conducting a thorough review of the processes and systems in light of this error." So, in conclusion, the issue was indeed a problem with upweighting but rather than human error simply an oversight in the coding, as it never occurred to anyone that a Saturday might be a "bank holiday" for sales purposes and so requiring special treatment.[