In my last column, I told the story of the humble origins of my company, Fog Creek Software. My co-founder, Michael, and I envisioned the business as a high-end Web consulting shop, until the implosion of that industry forced us to change plans. By the spring of 2001, we had laid off the company's two salaried programmers. It was just the two of us, working in an old Manhattan brownstone we rented from my grandmother. We had only two products, the flagship being a simple Web publishing program called CityDesk, which, of course, we thought was going to be as successful as the next Ford Model T, or Cinnabon at the very least. Unfortunately, success proved elusive. In May 2001, we made a scant $6,463 from the licenses.

With only two products and with no pesky consulting clients, we had plenty of time on our hands. We could have focused on writing code to make our software products better. But instead of doing that, we wasted a lot of time trying to solve all the wrong problems.

One day, for example, Michael was browsing eBay when he found someone selling "THE ULTIMATE SOFTWARE CD!!!" -- hundreds of dinky shareware programs for $2.49, including, it appeared, CityDesk.

Software pirates! Now this was a serious threat to our business.

We began to investigate, and we learned that there was a whole cottage industry of people on eBay trying to make a living by copying CD-ROMs on their home computers and selling them. This practice blatantly violated the auction site's rules. We spent a few days tracking down the right person at eBay to complain to, and we registered Fog Creek with the site's VeRO program, through which copyright owners can press a claim to have an illegal auction removed from the site.

THE ULTIMATE SOFTWARE CD!!! was booted from eBay, which left us feeling a great sense of victory over a few barely computer-literate eBay vendors -- most of whom seemed to live in Florida (I'm just saying) -- who fell for a make-money-from-home scam.

Auctioning bootleg copies of software was a hopeless business, of course, as CD-ROMs were being displaced by software that you could download. Trying to make money selling THE ULTIMATE SOFTWARE CD!!! on eBay was like trying to sell a pager to a person who uses an iPhone.

In a final irony, it turned out that the discs actually weren't bootleg copies of our CityDesk, a state-of-the-art Web publishing application that sold for $350, but instead a truly ancient shareware utility with the same name that allowed ugly dot-matrix printers to print pages that were slightly less ugly.

So this was one example of how we wasted our time. Another way we spun our wheels: putting together "strategic partnerships." In general, when someone contacted Fog Creek, we had a hard time figuring out whether the person was proposing a relationship in which money went from them to us or vice versa. For example, one day, we got a call from a guy who worked at a site that catered to webmasters. He offered to recommend CityDesk to the site's users. "We love CityDesk," he told us. Besides, the product the site had previously recommended had gone up in price, and the site didn't want to recommend it any longer.

"Wow!" I said to myself. "This could really make CityDesk take off!"

Naturally, the recommending site wanted a commission on every copy of CityDesk it sold. Fair enough. So our friend Neil, taking advantage of the dot-com crash to be between jobs, came into the office and spent a few days negotiating with the site on our behalf. In the meantime, Michael stopped what he was doing to develop the software to make this commission business work. He decided to build an affiliate system, modeled on Associates. The idea was that anybody with a webpage could create a link from his or her site to ours. The link contained a special code so that when somebody clicked on that link and bought our software, we would know to pay the affiliate a commission.

It took a couple of months of programming to get the affiliate system to work right, but we figured it was worth the effort to close the deal. We also thought we could incorporate affiliate marketing into our broader promotional scheme. We imagined millions of websites around the world all promoting CityDesk to make those referral bucks.

As the talks progressed, we hired a lawyer and spent a week putting together a contract, which we submitted to the publisher. But our would-be partner balked at this suggestion. "They just want us to sign their standard contract," Neil said.

Standard contract? OK, I'll take a look. The first line of the "standard contract" said, in all capital letters, INSERTION ORDER. And somewhere in my head, a light switched on.

We hadn't been talking to a potential strategic partner at all. We had been talking to an ad-sales guy. Here we thought we were negotiating an agreement to have a website promote our products in exchange for a commission when, in fact, the site was trying to sell us advertising.

Like I said, in the early days, we were young and naive and didn't always know which way the payments were supposed to flow. After that, I decided that when I answered a cold call, I would ask directly, "Are you proposing a relationship in which you write checks to us or one in which we write checks to you? Because, you know, we took home $812 last month, so we won't be writing any big checks."

We didn't do the deal with that website, but we did continue to plug away at our affiliate program. More than 200 people signed up, and 40 of them actually sold software for us. But the actual revenue was minuscule; only three affiliates made more than $1,000. The rest of them sold one or two copies, and the whole program was an administrative nightmare. Eventually, we shut it down.

The first three years of Fog Creek were like this. We tried in vain to find the magical sales and marketing formula that would make our software successful. We sent letters to existing customers (on paper, with a stamp!) asking if they needed more licenses; that cost us $300 and made $500. I added a feature to the trial version of CityDesk promising a 25 percent discount to a customer if he or she purchased a license within 72 hours, which required a whole coupon/discount infrastructure that took another couple of months to code up. That had some effect, but it was by no means a home run.

We couldn't afford advertising, but I was able to strike a deal with Programmer's Paradise, a mail-order catalog of software programming tools. I would write a one-page column to be printed in the catalog every month, in exchange for a half-page ad for our other product, FogBugz. That sure saved us a lot of money on advertising -- and, I think, even sold an entire copy of FogBugz.

And finally, we tried something else: We made our software better. In late 2002, we released FogBugz 3.0, a vastly improved product at a higher price. The launch was a huge turning point. Our sales, which had been averaging $7,000 a month, shot up to more than $30,000 a month.

The effect of making software that had more and better features was dramatic compared with anything else we tried. All along, we had thought we had a marketing problem. In reality, what we had was a product problem. Once we focused on improving the software itself, our business finally found success.

With higher sales, we could finally pay ourselves regularly. We moved the business into a very nice loft in the Garment District. And we could soon afford to hire employees again; I relied on a rule of thumb that we could add another full-time employee each time we produced an additional $10,000 a month in revenue.

Through this experience, I learned an important lesson: When in doubt, make your product more compelling. All of Fog Creek's affiliate marketing ideas, coupons, discounts, direct-mail pieces, catalog ads, and everything else we spent time on -- none of this was as good a use of our time as simply doing what we loved best anyway: creating useful software.

Joel Spolsky is the co-founder and CEO of Fog Creek Software and the host of the popular blog Joel on Software. For an archive of his columns, go to