Working Software had 8 to 12 employees during the three years I worked there during the early nineties. When I hired on, we sold four retail products and one OEM codelibrary (OEM Spellswell) for sell to other developers. By the time I left, we had nine retail products, three OEM libraries and an Open Source SDK that we gave away for free.<p>We worked closely with an Apple Developer Evangelist on that Open Source product, as the Word Services Apple Event Suite enabled word processors to communicate with our Spellswell spellchecker, so the word processor vendors didn't have to license a speller, or deal with integrating one, and users got one common speller UI, one single dictionary. The protocol was also open, in that we cooperated quite enthusiastically with all of our direct competitors.<p>One day the evangelist, Sue Layman, came to visit our office. She had not realized how small we were. She was quite envious at our very smallness, because we could "move quickly", for example ship a new product in response to changing market conditions.<p>At Apple, one could never write any code that shipped to end-users without an ERS - an Engineering Requirements Study. In general, ERSs were good things, in that they would explain the proposed design of the code, what it could do for the company and so on.<p>However, lots of people had to sign off on them before a project was authorized.<p>At Working Software, I once wrote a product called Toner Tuner. All it did was make your hardcopies lighter - that is, an adjustable grey rather than black - so as to save money on toner or inkjet ink.<p>We did that by hanging the controls off the Mac OS print dialogs, as well as injecting extra data into the stream going to the printer.<p>One day I happened to point out to our President that I could, with very little work, build a "watermarking" program out of Toner Tuner's codebase. It needed different print dialog controls, and to inject different data into the print stream, but the hard part was getting any controls at all into the dialogs, as well as getting any data at all into the print jobs.<p>I discussed this with Dave for maybe five minutes, it took a few days to produce a beta (starting with Toner Tuner's shipping code), we were able to charge quite a lot more money for Working Watermarker than for Toner Tuner, and it sold like hotcakes.<p>The total amount invested in paying me to write it, Dave to manage me, for our tech support and office staff to QA it - we didn't have a dedicated QA team - to design and print the product box, floppy disk label and manual was quite a lot less than what Apple would have paid, just to have a bunch of engineers and managers attend the meetings where new ERSs were discussed.<p>Apple and I expect quite a few large companies do some really stupid things sometimes. For example, while I was with WSI, Apple once dropped one million pieces of direct mail without testing anything at all.<p>When we dropped direct mail, we would start with a bunch of carefully controlled tests, where we would rent lists from other software publishers, trade magazines and the like, compose different offer letters, try out different text on the outside of the envelope, different prices for the product we were selling in the piece, even send the mail from different post offices.<p>It was only when we were dead certain we had a offer that would profit consistently, that we would roll out a large drop. Even then we would do it in stages, starting at say 10,000 pieces, then 50,000. Our very largest single drop was 250,000. It profited handsomely.<p>Steve Jobs did a lot to restore Apple's profitability, but I still see the company do really stupid things, like changing the UI of its software in ways that I cannot tolerate, while at the same time failing to fix quite serious bugs that I myself reported to them.