Friday March 09, 2007

Thoughts on an interesting interview with Chris Hofstader, formerly of Freedom Scientific

[Advanced warning: this is another long posting...]

The American Foundation for the Blind's magazine Access World has an interesting interview with Chris Hofstader in its March 2007 issue. Billed as "A View from Inside: A Major Assistive Technology Player Shares Some Industry Secrets", the article/interview attributes to Chris an enumeration of a number of "myths in the field."

One of the 'myths' Hofstader cited (or more accurately stated, "the most outrageous lie in the assistive technology industry") is that the AT industry "is too poor to fix problems and make better products and that there simply is not enough profit in this small market to do as much as talent may allow." Chris is in a excellent position to comment on these things, as for many years he was a senior member of perhaps the largest AT company in the industry. And while I completely agree with him that this and other things he cites are myths, I don't completely agree with him as to the 'real reasons' for the quality and stability of AT products (especially screen readers), and the price of products.

To be clear - the price of AT is a real problem and burden. I was in Brussels last Friday (among other reasons, I needed to get a chocolate fix from Mary Chocolatier - specifically their dark chocolate mouse offerings), and during conversations at the Belgian Confederation for the Blind and Low Vision, I discovered that it costs €3,000 to get a copy of JAWS professional in that country. The blind computer science student I met who would like to use JAWS professional can't get it from his school (they only provide a less expensive and less feature-full version). His parents aren't in a position to buy it for him, even as his mom (also blind) would like to use it for herself as well. And that is in a wealthy socialist democracy. It is far worse for blind folks in places like the Czech Republic!

Likewise the quality and stability of some AT offerings are also a real problem and burden. The stability issues are most significantly present in the most complex AT - screen readers. But where I disagree with Chris - at least as is attributed to him in the AFB AccessWorld article - is with his reasons for why this is the case.

In the article, Chris claims that annual revenues from all screen readers sold worldwide - a stated 2,000 copies/month - is between $18 million and $20 million. He then claims that this revenue is 10 times that of "all the software development people at all the leading screen-reading software companies." In the following paragraph, this is compared to ExxonMobile charging $75/barrel of oil that costs the company $25, and public outrage at that 3x margin (compared to his stated 10x margin for screen readers).

But I really must question these numbers. Chris claims the average selling price of screen readers is $800-$900 (so we will presume that most sales aren't the Belgian edition of JAWS professional). Chris might know the dealer markup, and percentage of sales that go through dealers, but I will just make a simplifying guess that actual revenue back to the screen reader developers is 65% of retail sales (a generous amount). Applied to the higher of Chris' estimates, we now have $13 million/year in revenue to the screen reader company from screen reader sales.

Next, we need to examine the assumption that the only employees important to getting a screen reader into the hands of a satisfied customer are those in software development. Even counting testing and documentation as part of the "software development" organization - they aren't in all companies - you still need some manufacturing, shipping, accounts payable & receivable, technical support, marketing (all those accessibility conferences), sales, and management. Furthermore, with something as technical and complex as a screen reader, training staff is important. And you also need to factor in things like rent on office space, buying computers and other equipment, and travel costs for that sales and marketing work.

Typically in a pure software development company you will have around 25% of the budget actually devoted to software development - 33% if you are very lucky. So that $13 million becomes $4.3 million (assuming the lucky 33% figure). After you factor in facilities and other equipment costs (again lets be nice here in our assumptions), let us presume we have $3.5 million for software developer salaries.

Next assumption: the average software developer salary is $50,000/year, and the "full loaded" cost is $75,000/year (including employer-paid taxes, health insurance premiums, disability insurance, etc.). That results in less than 50 folks in development, across all screen reader vendors. Seems like a lot, right? But that of course includes folks doing testing and documentation (and engineering management). And that is across all screen reader vendors: Freedom, GW Micro, Ai Squared, Dolphin, and others. If we have half of the software development staff actually writing code, divided by (let's say) 5 screen reader companies, we get a bit less than 5 coders on average for each company. And from what I know of these companies, that number isn't too far off the mark.

So if the reason these products are so expensive isn't because of outsize profits, why do they cost so much? And why are at least some of them of poor quality and stability?

I believe a key reason for both of these issues is the lack of good support for accessibility in the Windows operating system, and from the key Windows applications. 4-5 software engineers can get a awful lot done if they aren't spending the bulk of their time reverse engineering the operating system and software applications - and re-implementing that reverse engineering every new release of the OS or key application suite. For proof of this, I need only point to the small Orca screen reader team. 4 core people in less than 3 years have built a very powerful screen reader - something that is far more capable than any Windows screen reader was after 3 or 6 or even 9 years of development. It runs on Solaris and GNU/Linux; and it runs on the SPARC and x86 and PPC microprocessors - that's two different operating systems and three totally different microprocessors - all from the same code.

And it is precisely that need to reverse engineer everything that causes a lot of the issues with product quality & stability. I particularly like the "Telling Press Release" from two screen reader/magnifier vendors who noted with fanfare earlier this year that both of their products could be installed on the same machine at the same time, and both actually continue to function while doing so!

Of course, another reason for the high price comes from the quantity of sales. Worldwide software sales of 24,000 copies/year is minuscule compared to many other software products (cf. the many millions of copies of StarOffice that Sun sells each year - at a much lower price, and with a much larger development staff). Another comparison that I know very well comes from my former employer - Berkeley Systems. We had a team of 5 people making access products (1.5 developers, 1.5 folks doing testing & documentation & tech support all at the same time, and 2 others sharing all the other tasks), and we were struggling to be profitable with a $500 screen reader. Yet the rest of the company (nearly 100 of them) was doing very well selling $20-$50 products - in the millions of copies!

Which rather suggests that a different approach might yield us better quality and stability and lower cost. But this entry is long enough already; so that'll have to wait for another time... (2007-03-09 18:21:21.0) Permalink Comments [2]