Monday, July 27, 2009

The Story of StarText: 1983

(Note: This is one in a continuing series tracing the history of StarText. Read it from the beginning by using the Archive Post links on the right)

1983 -- Merry Christmas!

The jubilation we felt at the end of 1982 only grew in the first weeks of 1983. And all the credit went to Santa Claus.

On Christmas morning, 1982, Old St. Nick and his helpers delivered a lot of personal computers to residents all across North Texas. Underneath all that bright wrapping paper and pretty bows were names like Tandy, TI, Apple, Atari and Commodore. And every one of them a great candidate for StarText.

That became our “house ad” theme for January (and many Januarys thereafter): “Have a new computer? You need StarText!”

The result was more than we dared hope – 159 new signups in just one month. By the end of January we had climbed to nearly 400, a 53 percent increase! Along the way we got this letter from a new customer dated January 18, 1983:

“Thank you for helping to make this videotex service available. It is fun and easy to use. The articles are well written and short enough to be interesting. The summaries are exceptional; they are great lead-ins or side-bars (if I am not really curious). … If my assumptions are reasonable, there are currently about 20 percent of the required number of subscribers necessary to break-even. Don’t give up yet.”

He had obviously done some math even I hadn’t gotten around to yet. At the time I was focused more on surviving than breaking even.

Nonetheless, we began to grasp an underlying current that run through these early subscribers to the “online newspaper.” They were loyal. Often, fiercely so. For many, StarText was more than just a service. Just like the promise of what this new medium represented, StarText was their personal service. In the years to come, this feeling of ownership and loyalty would become one of the primary cornerstones of our success.

But while the boom in personal computers gave us plenty of reasons to celebrate, our rapid growth had an unsettling downside: The Tandy host computer was buckling under the strain of so many new users.

Not long after taking the editorial reins, I was curious about just what kind of equipment hosted the service, so our Tandy rep offered to show me. After making the trip down to the Tandy Towers, he ushered me through a highly secured area and into the main computer room. It was amazing.

There were walls and walls of what appeared to be floor-to-ceiling mainframes -- tape drives turning, red diode lights blinking, disk drives whirring. It looked like something I imagined that existed inside the Pentagon.

As we walked down a long aisle, I asked which one was the StarText host. “It’s back here,” he said. This was indeed impressive … very impressive.

When we reached the end of the room, he pointed and said, “That’s where we host StarText.”

At first, I was sure I must have been looking at the wrong thing. I wasn’t.

Parked on a microwave cart was a solitary Tandy Model II, a maze of 16 phone lines pouring out the back. I don’t recall my exact reaction, but my expression must have approximated that of Richard Dreyfuss in Jaws when he informs his colleagues, “We’re going to need a bigger boat.”

Fortunately, this was something Joe Donth had already realized. As the person in charge of the technology, he was acutely aware of the problems, both internal and external. Our new customers weren’t “getting their newspaper delivered” when lines were busy or the host was down. The editors were struggling to keep the database fresh in the face of more frequent crashes. And every new customer we added, while precious for our survival, only exacerbated the issue.

It seemed StarText had traded one crisis for another.

While we continued to press Tandy hard for performance improvements, Joe was quietly taking matters in his own hands. One night, after another “dog and pony” for a local user group, Joe dropped a bombshell: He confided he had already started work on a new program that might replace our dependence on the Tandy system. The news was both a shock and a surprise, and after it sank it, very exciting.

But I had to wonder: Could one person working part time replicate and better a system developed by the team at Tandy?

Soon, more details began to emerge. Joe’s software would take us from the Tandy Model II to a powerful VAX 750 (the Star-Telegram was heavily invested in DEC computers and the VAX was among their newest models at the time). Gone would be our reliance on Tandy; we would take StarText in-house and solve the technical issues holding us back. It seemed too good to be true.

Working nights, weekends and any spare time in-between, Joe cranked out line after line of code, the thousands of instructions which would orchestrate innumerable zeros and ones inside the VAX to operate StarText. As an editor, programming seemed at best a cross between calculus and Greek. But I quickly developed an immense admiration and respect for the people who understood and spoke its language. And among them, Joe was a true maestro.

By mid-February, Joe was far enough along to break the news to Tandy. In an exchange of letters between Donth and Carroll Reeves, Tandy’s director of contract marketing, an agreement was hammered out. Among its main points:

-- StarText would be moved to the Star-Telegram “on or before July 1, 1983.”

-- The Star-Telegram will utilize their own computer equipment and software to provide a successor to the StarText system.

At the same time, Tandy was working on its own “next generation” videotext product. In an interesting twist, they sought our support to test it at the same time we were about to test our own solution. We agreed to continue providing them content with a few caveats. Key among them was the test period would be limited to 90 days duration and only open to 100 total StarText subscribers on a voluntary basis.

The result was StarText customers had their choice of using Joe’s new VAX-based service or trying the service supported by Tandy. In effect, for a period of 90 days there would be two versions of StarText in the market in a likely winner-take-all competition.

Another initiative gaining momentum at that time was a growing demand for metro service. Essentially, a way to make StarText a local call for customers in the Dallas area. This would increase our potential market by a factor of three. We already had a small number of subscribers who willingly paid long distance rates for the privilege. No one doubted opening Dallas represented a potentially target-rich environment. The question was how to pay for it.

Central to that equation was answering, “How many subscribers can one line support?” I quickly learned the subscriber/modem ratio is a key factor in any online business plan. The idea is to accommodate the maximum number of customers with the least number of lines. While that sounds simple, arriving at the number is anything but. By looking at the usage reports of any given day or week, there is an obvious pattern to usage “peaks and valleys.” For us, daytime usage was mostly slow. Our peak usage was between 7 pm and 10 pm, with another peak around midnight. Ideally, you have enough lines so when the peaks come, no one experiences a busy signal. (This is of course the exact reverse for Internet usage. Typically the peak time is 8 am to 5 pm weekdays, when there is massive usage from workers with PCs at their desk.)

The other factor that comes into play is “average time spent online.” If this number gets higher, the chances of a busy signal increase and thus drives the need to add more lines. So while you wanted to have great content and features that prompted people to buy and use the service, you never wanted it to be so compelling they never hung up. (This is one major reason we never featured chat rooms or online games.)

Taken together, the online service provider walks a tightrope between providing the best possible service and trying not to go broke doing it. While the phone lines themselves weren’t a big expense, each line required its own modem. Behind the modems were multiplexers. The multiplexers required computer ports. And before you know it, your newspaper has quietly become a telecommunications company.

To launch metro service, we couldn’t have just one line (it would be busy all the time) and the cost of a metro line was many times the cost of a local line (remember – it’s 1983 and competition for local service is just beginning). Still, we started planning for a metro service that would be part of a post-Tandy rollout.

A New Era Begins

By late March we were ready to turn on the VAX-based version of StarText. Tandy was ready to serve up its new and improved “StarText Classic.” Letters were mailed to existing customers (a little over 500 at the time) advising them they could choose which StarText they wanted to use. Then, on April 1, we “threw the switch” that brought up StarText as an independent service running on in-house equipment at the Star-Telegram.

Expectations were high. The VAX was a real, state-of-the-art computer. We expected data – even at 300 baud -- to speed along an Information Autobaun compared to what subscribers had been experiencing. Partly because of that, we decided to raise the monthly rate from $5 to $7.95. The extra revenue would help underwrite the cost of the VAX, telephone lines and modems.

While testing of the VAX-based service had gone well, nothing compares to placing any new software under full load in “real combat conditions.” As subscribers began lighting up the modems it didn’t take long for the verdict to come in, and it wasn’t at all what we expected: There was a problem – a big problem. As incredible as it sounded, the VAX was performing worse than the Tandy Model II.

As editors fought back the urge to panic, Joe quickly went into troubleshooting mode to isolate the problem. What he discovered was that large sections of his code needed to be rewritten. In the weeks that followed, we manned the phones and urged patience while Joe set about making the changes. In what I sometimes call the “Miracle of ’83,” Joe rewrote the software in a matter of weeks. The new service was everything we promised and more.

In the 90-day period that followed, it became literally, no contest. The vast majority of subscribers migrated to the VAX service. By June, it was over. Tandy pulled the plug on its system and the two companies parted friends.

Looking back, it’s fairly obvious without Tandy’s initial involvement there would have been no StarText. At the same time, it also taught that regardless of how attractive online businesses looked on paper, launching, maintaining and growing such services was far more complex and challenging than most companies imagined. While the technology existed to provide services, the market for those services was limited at best – and would be for some time to come.

Over the years I came to relate this to an item that appeared in the Star-Telegram around that same time. It was called a “filler.” For those who don’t know, “fillers” were those tidbits of mostly useless information used by newspapers in earlier days to fill news columns when stories ran short. This particular filler was about a certain species of mole that had the amazing ability to burrow up to 100 miles in a single night. The headline – which won an award, by the way was this:

“But he seldom wants to”

It struck me this was applicable to technology and technological innovations. Just because we can do it, should we? Is it something customers really want, or just something the marketers, researchers and engineers think they should want? That little industrious mole has became my yardstick for evaluating service “improvements” ever since.

It was now just over one year since StarText was launched. Our first birthday was followed in short order by our “independence day.” It had been a tumultuous 13 months but we had a lot to feel good about. And a lot more work in front of us.

While it felt good to have the entire operation housed at the Star-Telegram and under our control, it also meant new responsibilities, like new billing, marketing and technical support. We could expect some help from the newspaper but it still meant new programs to track revenue and labor to stuff and mail monthly bills. For the latter, Joe’s administrative assistant became our part time accountant. And it wasn’t unusual to see Joe, myself or one of the editors stuffing statements in between other duties.

Because we were a small group, and expense control was crucial, we each wore whatever hat necessary to get the job done. One job we all shared was customer service. Through the magic of email, we quickly developed “one on one” relationships with many of our readers. The result of personally answering a question, complaint or suggestion was amazing. Readers couldn’t believe someone would take the time to actually respond. They were conditioned to large, “faceless” corporations who either sent form letters or just didn’t acknowledge them at all.

While time-consuming and labor intensive, this was important in many ways. For one, it made our customers feel a certain pride and ownership in the product. We didn’t have subscribers as much as stakeholders. For another, it encouraged ideas and feedback. At one point we estimated more than half of every improvement we made came from a subscriber suggestion. There were only five of us. Imagine tapping into the brainpower of 500. Or 5,000. Or someday, like the big dailies, 500,000 and more. What might we accomplish?

Among the new content we offered was a simple column we called “Comments.” Updated several times weekly, it published selected email from subscribers along with our responses. Almost immediately it went to the Top 10 most requested keywords and remained there for most of the next 10 years.

In July of 1983 we took another bold step forward when we officially launched our StarText Metro service. For the first time, Dallas-area customers could dial us up as a local call. At the time, we weren’t sure how The Dallas Morning News might react. Even though they had shuttered their online service, BISON, the year before and showed no signs of re-entering the fray, this was the first time the Star-Telegram, albeit electronically, had entered the Dallas market in a direct way. As it turned out, they ended up running several stories about us. We used to joke that StarText got more coverage in Dallas than we could get in the Star-Telegram.

While as a journalist I understood why they didn’t cover us (it’s best to avoid possible conflict of interest areas by not blowing your own horn), it was a continuing source of frustration that we had to read about progress in other publications.

For the record, Rich Casey, an avid Ham radio enthusiast, was our first Dallas metro line customer. His subscriber ID number – 1081 – was among the lowest on the system, something Rich was always very proud of.

“I was on StarText the morning the metro number and the VAX came on-line in
1982,” said Casey in a 1994 interview. “My original mail code of 91 converted to 1081 sometime in 1984. At the StarText 10th anniversary, I was surprised to find I had the lowest subscriber mail code.”

Why did he choose StarText? “I've always liked the open exchange of ideas, the willingness to experiment with new concepts, and the local feel of the service. On StarText, I feel more like a ‘member’ than a ‘subscriber.’ I have subscribed to CompuServe since 1979 (one evening there were four users on the entire system); I sampled and quickly left Prodigy (I couldn't get past the smell of Sears popcorn), and am now wandering the Internet. Still, StarText feels most like home.”

To be continued ...