Since I created this page, the SPA have published an FAQ page on their web site about SPA software sales figures. According to the SPA, my tables below are worthless, as "Each year's press releases have been constructed to stand alone and compare only to the previous year's data contained in that release.". The SPA also state that "Growth rates reported in press releases are the exact rates reported in the Data Program reports.".
This is very puzzling to me. For example, in their March 1996 press release, the SPA said total sales "reached US$7.53 billion in 1995". In their March 1997 press release, the SPA said total sales "reached US$10.6 billion in 1996, an 8.3% increase from a revised US$9.8 billion in 1995.". Now, as far as I understand their FAQ, the SPA are saying that the percentage change reported in this press release is correct, but that I shouldn't worry about the difference between the figures given for 1995 sales as the US$9.8 billion figure was only given for comparison with the US$10.6 billion figure for sales in 1996.
So how much software actually was sold in 1995? In March 1996, the SPA said US$7.53 billion. In March 1997, the SPA said US$9.8 billion. Which figure is correct? Is either figure correct?
The only way I can make any sense of the published FAQ and the published figures is to think that the SPA is saying that the dollar values they put on sales are not accurate at all, and that no-one should pay any attention to them, only to the percentage change figures. It seems to me that the SPA don't know how much software was sold in 1995 in absolute terms.
If this is the case, they certainly don't give that impression in their press releases, where the percentage figures are given after the dollar value, and the headline they give for the 1996 sales press release isn't 'Software Sales up 8.3%' but "Software Sales Pass $10 Billion For The First Time".
But, if the SPA don't know how much software has been sold in actual US$, how can we trust their percentage change figures?
So, here's the original web page, although it should be read in light of the above ramblings and the SPA's published FAQ. Make of the figures what you will. I'm increasingly of the opinion that any analysis of the figures is pointless because the figures are so untrustworthy to start with.
I was inspired by Joseph M. Ragosta's web page and articles in MDJ (MDJ: The Daily Journal For Serious Macintosh Users) to investigate the SPA (Software Publishers Association) figures for Macintosh sales myself.
I decided that trying to analyse the methodolgy was too hard, given the lack of information publically available on their methods and sample sizes, but to concentrate solely on their own published figures. The SPA web site has press releases going to Q3 1994. By trawling the web, I found contemporary reports of their releases for Q1 and Q2 1994 as well.
In the table that follows, I have followed Joseph's style and put published figures in blue, and calculated figures in Red. I have also provided links back to the original sources for published data. The calculated figures are all derived from the published figures. The calculated 'Final Estimates' are obtained by using the figures for sales and percentage change from the next year. The calculated year-on-year change is simply the change from one year's 'Final Estimate' to the next.
All sales figures are in millions of dollars and cover North American sales, only.
| Sales Quarter | Mac Software Sales given in SPA Press Releases | SPA Final Estimate of Mac Software Sales | Year on year change given in SPA Press Releases | Year on year change using SPA Final Estimates |
| Q1 1993 | 217.8 | |||
| Q2 1993 | ||||
| Q3 1993 | 270.1 | |||
| Q4 1993 | 321.2 | |||
| Q1 1994 | 259.0 | 279.5 | +18.9% | +28.3% |
| Q2 1994 | 276.0 | 243.5 | ||
| Q3 1994 | 316.0 | 329.3 | +17.0% | +21.9% |
| Q4 1994 | 397.0 | 361.3 | +23.6% | +12.5% |
| Q1 1995 | 307.5 | 335.0 | +10.0% | +19.8% |
| Q2 1995 | 226.5 | 376.1 | -7.0% | +54.5% |
| Q3 1995 | 242.5 | 375.4 | -26.3% | +14.0% |
| Q4 1995 | 306.4 | 479.1 | -15.2% | +32.6% |
| Q1 1996 | 318.4 | -5.0% | ||
| Q2 1996 | 296.8 | -21.1% | ||
| Q3 1996 | 239.9 | -36.1% | ||
| Q4 1996 | 333.6 | -30.4% |
Please note that there is an error in the Q4 1995 press release. It states 'Macintosh application sales continued their slide begun in the second quarter, falling 15.2 percent from Q4 94 to $361 million.' The $361 million is the revised Q4 1994 figure, not the Q4 1995 figure, as is obvious from the tables included with the press release.
I hope this table is sufficient to show that for the last three quarters of 1995, the SPA press releases were grossly mis-leading with respect to Macintosh Software sales. From Q1 1994 to Q1 1995, the SPAs initial figures are very close to their final estimates (all within 13.3% of the final estimate), with no consistent direction to the error. The initial figures from Q2 1995 onwards have all been below the final estimates, and by a minimum of 35.4% of the final estimate. I have seen no evidence that the SPA have taken any steps to correct the problems (whatever they are) with their data collection, and so I expect that the 1996 figures will be shown to be equally inaccurate. We shall start to find out when the 1997 Q1 figures are published. Of course, the SPA should have the 'Final Estimates' for Q1 and Q2 1996 pretty much sorted out by now, unless their system is even more inefficient than it appears. Unfortunately, they do not publish their 'corrected' figures until making a comparison with new figures.
At the suggestion of Dan Sokol, I have also prepared a similar table for Windows software sales. Windows includes both 16-bit and 32-bit Windows applications.
| Sales Quarter | Windows Software Sales given in SPA Press Releases | SPA Final Estimate of Windows Software Sales | Year on year change given in SPA Press Releases | Year on year change using SPA Final Estimates |
| Q1 1993 | 619.4 | |||
| Q2 1993 | ||||
| Q3 1993 | 711.4 | |||
| Q4 1993 | 1182.8 | |||
| Q1 1994 | 889.0 | 1000.0 | +43.4% | +61.3% |
| Q2 1994 | 1050.0 | 974.3 | ||
| Q3 1994 | 1060.0 | 985.3 | +49.0% | +38.5% |
| Q4 1994 | 1650.0 | 1578.8 | +39.5% | +33.5% |
| Q1 1995 | 1430.0 | 1481.6 | +43.0% | +48.2% |
| Q2 1995 | 1158.8 | 1477.3 | +18.9% | +51.6% |
| Q3 1995 | 1857.3 | 2335.2 | +88.5% | +137.0% |
| Q4 1995 | 1524.8 | 1947.4 | -3.4% | +23.3% |
| Q1 1996 | 1776.9 | +19.9% | ||
| Q2 1996 | 1684.0 | +14.0% | ||
| Q3 1996 | 1905.5 | -18.4% | ||
| Q4 1996 | 2859.4 | +46.8% |
Interestingly, this tables seems to show a similar pattern to the Macintosh software. From Q1 1994 to Q1 1995, the SPA's initial figures are again very close to their final estimates (all within 11.1% of the final estimate), with no consistent direction to the error. The initial figures from Q2 1995 onwards have again all been below the final estimates, and by a minimum of 20.5% of the final estimate.
In light of this, I think the SPA figures must be viewed with great caution. Some change appears to have happened in Q2 1995 to make their initial estimates much more unreliable, and this change has affected both Macintosh and Windows sales figures, but has affected Macintosh figures more than Windows figures.
Email me with any comments about this page - especially if you have additional figures (with source please!).