Going to Hawaii
Selling one hundred million dollars worth of software was easy compared to the trouble we had fulfilling Alan’s promise to take everyone to Hawaii. Originally we had planned to ship 5.0 in the fall of 1987, hoping that by the first of 1988 our support calls would slow down enough that we could take the phones off the hook and all go to Hawaii for a week. If we all went together, we thought we could hold a few short meetings so the employees would not be taxed for the total value of the trip. This was a lot of wishful thinking, however. 5.0 was late, and by now 4.2 sales had risen to a level where we were taking 4,000 support calls a day.
Plan B was not to go as a group, but to let everyone go individually on their own schedules. For those who wanted to take the trip, we offered them an $1,800 credit with our travel agent, $300 or $400 cash to cover other expenses, and three extra days of vacation. Our travel agent negotiated special rates with the airlines and the hotels, so that the allowance was enough to cover the cost for two people. For those employees who did not want to go, we offered a $1,000 bonus instead.
What seemed to us to be a fairly generous travel bonus (which was given in addition to the normal year-end bonus), was not well-received by all of our employees. Many were grateful, but there were also many who were unhappy. Some felt cheated because they had to pay taxes on the value of the trip. Some were disappointed when there was only enough money to visit two islands in the seven day period, instead of the three they had their hearts set on. Some were angry because we asked them to use two of their vacation days. Some wanted to use their own travel agents. Some of those who decided not to go because of babies coming or the taxes felt the $1,000 alternative was unfair. Of course, most of the employees who were unhappy thought I was the cause of their frustration. Alan had offered a free trip, and any taxes or limitations had to be my fault.
When the ship date for 5.0 first slipped, the programmers set a schedule to ship the product in February and go to Hawaii in March. The ship date kept slipping, however. Our travel agent had to change their reservations to April, then to May, and finally to June. The rest of the employees also had trouble getting away. The marketing staff spent its time reassuring customers that 5.0 was coming and preparing for the 5.0 roll out. The support operators were busy trying to learn 5.0 while answering all the 4.2 questions. Others in the company were struggling to keep up with the growth in sales and employees.
As much as we wanted to ship the product right away, we could not hurry it out the door. It had too many bugs to finish it up with a few late night sessions. The programmers and testers were like a weary army on a forced march across a desert without food or water. We could not push them any faster for fear they would collapse. To make matters worse, their spouses were growing impatient with all the overtime and vacation postponements. We held meetings with the disgruntled family members to reassure them that the product would soon be finished and that the sacrifices being made were worthwhile. By now almost all of the WordPerfect for DOS developers had put their personal lives on hold for close to a year.
Our offices stayed closed on Sundays, but work continued around the clock during the remainder of the week. During the day and into the evening, the programmers would work to add the last few features and fix bugs. During the late afternoon and long into the night, the testers tried out the new fixes and looked for more bugs. Many of the support operators worked a second shift late at night to help the testers search for bugs. To keep them all going, we had pizzas and sandwiches delivered each evening for those who stayed late.
The pressure to ship 5.0 increased as each day passed. By announcing the product early in hopes of preventing Microsoft from stealing our customers, we had created a lot more interest in 5.0 than we had anticipated. We sent out literally thousands of copies early in an effort to make up for the delays. Our large customers wanted the product early so they could conduct their evaluations and get their support centers up to speed. The editors of trade publications demanded early review copies, some threatening bad press if we did not cooperate. Authors trying to write 5.0 books enjoyed the advance copies, but complained because the software was full of bugs and the interface was still changing.
The printer code was the biggest bottleneck to shipment. Until it was completed, we could not finish all the new printer drivers. By now we supported hundreds of printers in 4.2, including some which were very obscure and hard to find. Unless we decided to run the risk of having some very angry customers, we had to include all the drivers found in 4.2 in the new release. Once the drivers were written, we still needed time to test them.
We were all short-tempered. The programmers trying to finish the printing code had the hardest job. They were still finding flaws in parts of their design, and some of their fixes required changes to other parts of the program. The programmers working on the other areas hated to make changes so close to release, so the two sides did battle in the hallways and in front of white boards. Like the differences in the early design meetings, the arguments were hard to understand and the proposed solutions were very complex. If I asked a simple question like, “What is the difference between a typeface and a font?,” the programmers were likely to dispute each other’s definitions. Sometimes a problem had no good solution, and we were forced to go with a decision no one liked.
For example, if a user created a document at work that was designed to print on a laser printer and then took the document home and used a different computer and printer, should WordPerfect automatically reformat the document when it was retrieved? If the user did not want to reformat the document, what would happen if the printer driver from work was not available on the home machine? And if the document was reformatted automatically, would the user be notified and given a chance to cancel the reformatting? As I listened to the discussions, I told myself that all the difficulties we were experiencing would eventually give us a competitive advantage. If all of this had been easy, Microsoft would have done it first and left us in the dust.
To add to our pressure, we were offering free updates to anyone who purchased 4.2 after the official 5.0 announcement at COMDEX. We did this to prevent our 4.2 sales from falling in anticipation of the new release. The offer worked, and our February and March sales were more than ten million dollars each. Unfortunately, the offer was going to cost us $1,000,000 for each month we were late.
In March we went to MacWorld. We still did not have our Mac product ready to ship, but I decided we would sell it anyway. I used a new term, Betaware, to describe product offered for sale during the beta testing period. For $99 the customer received a buggy copy of the software and the promise of a good copy once it was ready. The Betaware customers actually received three copies of the software–the betaware copy, the first shipping version, and a third copy because our first shipping version had a few too many bugs for us not to release another copy.
On April 4, 1988 we finally sent our Macintosh product out the door after more than a year of testing. In its first month we shipped more than 20,000 copies of WordPerfect for the Macintosh, which amounted to sales of almost $3 million. The reviews for the product were less than glowing, however. Most reviewers felt the product was patterned too closely after our PC product. They were right, since our strategy had been to offer a compatible word processor across all platforms. Unfortunately, many in the Mac community who were fervently loyal to their machines did not appreciate the compatibility and were offended by the foreign DOS-like interface. We were more than a little disappointed at this lukewarm reception from what I liked to call the Macintosh bigots, who thought we were not sufficiently converted to the Mac way of doing things to merit their welcome. Even those Mac users who wanted compatibility were disappointed with the product, because its used the DOS 4.1 file format, rather than the 4.2 or 5.0 format. Eventually we would average sales of about $1 million per month and capture about ten percent of the Macintosh word processing market. This was not too bad, but we had hoped to give Microsoft more of a run for their money.
On Thursday, May 5, 1988, we finally shipped WordPerfect 5.0 for DOS. This time around we were determined to have plenty of product for all our distributors, so by Friday afternoon we had shipped an enormous number of copies, perhaps as many as 100,000. What should have been a big celebration quickly turned into a wake, however. Within hours of the release, we found a bug in the installation program.
At the last minute, just before delivering the final masters to our manufacturing company, the 5.0 programmers had discovered there was not enough room for all the files they had planned to include on one of the diskettes. As luck would have it, there were a couple of files which were not very necessary to the program, so they left them off the master diskettes in order to save space. Because it was such a minor change, no one bothered to test anything before delivering the masters. Unfortunately, the installation program had not been changed to reflect the missing files, so every person who ran it (which was everyone who bought the product) saw two error messages which warned that two files were missing. The program worked just fine, but the warning scared customers into calling us for help.
We tried to recall as much of the product as we could, but we had been too efficient at shipping it out. We quickly slipped in a new release to correct the problem, but it was too late. On Monday, May 9, the calls started to trickle in. By Thursday, the trickle of calls had turned into a flood. On Friday the 13th there were so many people trying to call us that our busy signals brought down the entire AT&T 800 system in the Mountain West. The phones in the Delta Airlines reservation center and the American Express customer service center, both in Salt Lake City, all went quiet. AT&T called around lunchtime to politely inquire how soon we could clear up our busy signals. Much to our embarrassment, we had no answer for them. We were in deep trouble.
As it turned out, the new release had a few other problems as well. Not all of the printer drivers were ready, and not all had been thoroughly tested, so we had a lot of irate people wondering how they could get their printers to work. We were also having trouble with print preview and the graphics editor, features which required a graphics mode. We had assumed, based on representations from manufacturers, that most of the graphics cards on the market would be compatible with those we had tested. After the release of 5.0, however, our customers were finding some incompatibilities which we had not planned for.
Looking back, it is hard for me to blame the programmers or the testers for the problems we had with 5.0. The developers had worked so hard for so long that if we had not shipped when we did, it would have taken us the rest of the summer to recuperate and finish the testing. They had marched across the desert until they dropped and then crawled as far as they could. When they could go no farther, we shipped. Prior to 5.0 our criteria for shipment had been to go a week without discovering any major new bugs. The last week before release the developers did not find any new bugs, but then everyone was so tired that no one wanted to find them. In their defense, except for the obvious installation problem, most of the problems had to do with little known graphics cards and printers. Perhaps if we had used more customer beta test sites, we could have found more of the incompatibilities and problems.
Though hard to believe, people were buying 5.0–bugs and all. Our customers were angry and disappointed and took many opportunities to chew us out, but they still bought the product. WordPerfect 5.0 delivered on almost all of the promises we had made. With a laser printer attached to a PC, 5.0 could print what looked very much like typeset quality work. In fact, my neighbor who sold Macintosh computers for a living would not believe an IBM PC could print like a Macintosh until he saw 5.0 print with his own eyes.
The reviewers were very kind to us. InfoWorld had changed their scoring criteria again, but, even with the “poor” rating for technical support, the 7.9 rating they gave us overall was a good score. The headline read, “New WordPerfect is Once Again King of the Hill.” PC Week, which by now had become an influential publication, was even more complimentary. “This is simply the best there is in word processing…well worth the wait…an all-too-rare software product: one that is genuinely worth getting excited about.”
As soon as the programmers came back from Hawaii in June, we started fixing our other bugs. Soon we were shipping much cleaner software. That summer we spent a lot of money sending out free diskettes to those customers who had received the flawed early copies. We also spent a lot of time visiting customers and speaking at user group meetings to reassure people that the 5.0 software was reliable and that we were doing all we could to answer the support calls.
I enjoyed visiting the user groups. Practically every city had at least one, where computer users would meet about once a month to help each other learn to use their computers. Most of the members were men, but women were welcome and many attended. Users of all ages came, including quite a few elderly people. Some members came in suits straight from work, but most people dressed casually. A normal meeting would consist of a few announcements, a question and answer period when people with problems could stand up and ask for help, one or two demonstrations or presentations from computer or software companies, and a drawing for a prizes provided by the presenters.
When I visited a group to make a presentation, I liked taking questions from the audience. I could count on a few hecklers who were unhappy about not getting through to customer support, and I learned to enjoy the exchanges. I did my best to explain our problems and what we were doing to fix them. I also dropped a few hints about upcoming products and explained the reasons behind many of our decisions. Almost always I found that customers were willing to give us time to correct our problems and mistakes as long as we were honest and fair with them and tried to do our best.
Fixing customer support was not as easy as fixing our buggy software. In the short term, we asked everyone in the company to take support calls. When customers could not get through on a support line, a lot of them started calling our other phone numbers. All of us, including Alan and me, tried our best to help people when they were on the line rather than transferring them to a busy signal on the support lines. To further relieve the strain on support in the short term, we installed a special 800 number just for installation questions, using trailers as temporary office space and new hires reading from scripts to help people get started with 5.0.
Coming up with a long term solution was not easy. Even if the product had been clean, we still would not have had enough people in customer support to handle the needs of our growing customer base. We simply had underestimated everything having to do with the release. First, we had hired and trained enough people to cover a 25% sales increase, but the number of copies we were shipping increased by more than 100%. Second, we had not planned on many calls coming in from our old customers buying updates. Our 4.0 customers had not needed help with 4.1, nor had our 4.1 customers called when they moved up to 4.2. With 5.0, however, a very high percentage of update customers asked for help. Third, we had expected the average number of minutes per call to remain about the same, but we were wrong again. Due to the increased complexity of the software, call times went from seven minutes per call to almost fifteen.
Before the release of 5.0, we had 200 operators taking about 5,000 calls a day. By August the support staff had grown to 340 operators, but, as a result of the long call times and the inexperience of the new hires, they could handle only 6,000 calls a day. Because our sales had doubled, we needed to be answering at least 10,000 calls a day, even if none of our update customers called us. Three months into the release of 5.0, we were still not close to solving the problem of how to answer all the calls.
Things were so bad that Dan Lunt actually asked me for help. After almost nine years, Dan and I still had trouble working comfortably together. He started with the company only three months after I did, and when Don left and Bruce and Alan asked him to help me with the marketing, he said he would on the condition that he would work “with” me, but not “for” me. At the time this condition was not very important in our all-for-one-and-one-for-all type organization. Over the years Dan and I developed a friendship of sorts, but he was never one to get very excited about my organizational philosophies. He liked to give his managers the freedom to run their departments the way they wanted to run them.
He picked me up in his black Porsche one weekend afternoon that summer, and as we drove out through the desert west of Utah Lake, he described the problems in customer support. He was worried most of all about the morale of the group, because though they were working their hearts out, they were not catching up. Some of the experienced operators were leaving, something which was unheard of for us, and the organization was having trouble finding new operators. When they did, they were having trouble getting them up to speed. He also told me that our completion rate, or the percentage of callers who got through to an operator instead of getting a busy signal, was a discouraging 10%.
I found out that the starting salary for a support operator was only $850 a month, which was probably the biggest reason why we were having trouble finding good people. Once hired and trained, the operator was assigned to a supervisor who was responsible for a group of about 75 people. The groups were large, because Stan Mackay, the director of the department, was trying to follow my advice and keep a flat structure. The groups were, however, too large for one person to manage, and an unofficial layer of administration had developed to fill in the management gaps. This layer included trainers, who were responsible for the ongoing training of the operators; resource people, who were there to answer tough questions; monitors, who listened in on calls and graded the operators from time to time; and ICE people, who answered correspondence and kept track of enhancement requests.
I learned that those in charge of the support department were busy designing the floor plan for their new building in the research park. In the center of both floors of the building were private offices for the director, his group leaders, and the elite corps of trainers, resource people, monitors, and ICE people. The operators, the unfortunate many, were destined to work in the outer fringes of the building in small cubicles far away from their leadership.
It was not hard to figure out why morale in the department was so low. Not only were the operators paid poorly, but they were treated as if they were second class citizens. There were so many of them in each group that they hardly knew their group leaders, and when their work was graded, it was done secretly by monitors who never talked to them face to face. Ironically, many of the people we were paying to answer phones viewed taking calls as punishment and were doing all they could to arrive at the day when they could leave their headphones behind and join the ranks of the elite who had made it off the phones.
Over the next month after our ride in the desert, I helped Dan and Stan reorganize the department. Of course, the group leaders did not want smaller groups. They wanted assistants to help them so their groups could grow. Instead we divided all of support into teams of no more than 23 people, each team being made up of one team leader and 22 operators. We took a few of the old trainers and formed one training group, putting the remaining trainers back on the phones. The ICE people and the monitors also went back on phones, except for a small group assigned to edit the database of support information. Instead of using what I fondly called the Secret Police, we turned the monitoring over to the team leaders.
Another step in the reorganization was to eliminate the private offices in the middle of each floor and replace them with more cubicles. We gave the team leaders two cubicles each and enclosed their space with glass. By positioning the leaders in the center of their groups and providing them with these semi-private offices, we hoped that team members would get the attention and help they needed.
We arrived at a team size of 23 people because the cubicles were originally laid out in groups of 12. Given the two cubicle space required for each team leader, we could have had teams of 11, 23, 35, or 47. I thought at the time that ten operators were too few for one team leader and any more than 22 would be too many. As it turned out, we learned that it was best to start new team leaders with only ten or twelve operators and let them work up to 22 as they gained experience.
Once we felt we had provided the operators with a means of getting the support they needed, we turned our attention to solving the problem of their low wages. We raised the starting salary for support operators to $1,350 a month and designed a salary schedule that would allow a good operator to move up to $1,800 a month after two years. We hoped the changes would help the operators feel as if their jobs were important and provide incentive for them to stay on the phones.
One of the best changes we made to support was to limit the number of phone lines going into each group. At the time operators were required to spend five and one-half hours of an eight hour shift on the phones. The rest of their day was spent either in training or finding and calling customers back with solutions to the problems they could not answer on the first try. Our support lines were open eleven hours a day, from 7:00 a.m to 6:00 p.m. mountain time, in order to cover the hours of 9:00 a.m. to 5:00 p.m. for all time zones in the continental United States. Each team of operators had a specific assignment, for example, answering installation questions or answering network questions, and was responsible for that assignment for the entire 11 hour period. Because each of the 22 members of a team was on the phones for 5.5 hours each day, or exactly half of the 11 hour period, it made sense to limit the number of lines for each group to 11. Dan added a twelfth line so that an extra caller could get into the system and be waiting for a line to come free. By limiting the number of lines into a group, the number of calls an operator had to face became a finite quantity and our hold times went down.
As we moved into the new building and worked into the new organization and higher salary schedule, morale improved. Gone were the hidden monitors with their point system for evaluating calls. Gone were all the elite positions and the tremendous incentives to get off the phones. Gone were the days when supervisors did not have time to meet the needs of their people.
Of course, not everything worked perfectly. For some reason, the trainers decided they should turn their initial operator training into something resembling a military boot camp. I never understood why they took so much pride in making a new operator’s first experience with the company a very difficult one. We also had trouble finding good team leaders. At least one third of the people we tried out were not strong enough to let their people know when they were doing a poor job. Another third were simply too tough and had trouble being nice and treating their team members with respect. They had difficulty understanding that their purpose was to help their team members, not to boss them around.
It took us until November to hire and train enough people to begin to handle all the support calls. It had taken us six months to get our completion rates back up to a more respectable 50%. Except for those times when we seemed to forget that we needed to keep hiring more operators as sales went up, customer support functioned much better over the next few years.
In spite of our focus on word processing, and specifically on WordPerfect for DOS, we still seemed determined to try to release as many unsuccessful products as possible. I started telling the old joke about the rancher who won one million dollars in a lottery. When asked what he would do with the money, the rancher said, “I’ll keep ranching until the last dollar is gone.” I felt like we would keep releasing products until our last dollar was gone. We were still headed on a course to put WordPerfect, PlanPerfect, DataPerfect, and WordPerfect Library, which was now called WordPerfect Office, on all important platforms. The fact that some of these products were not doing well did not slow us down.
Luckily, we had a few successes along with the failures. Surprisingly, WordPerfect 4.2 for DOS kept selling well, as a lot of large corporations decided to take their time about moving to 5.0. Our new WordPerfect Office with electronic mail for PC networks and WordPerfect 4.2 versions for UNIX and VAX computers also had good sales. WordPerfect for the Amiga and WordPerfect for the Atari ST started out well, but after only two or three months their sales started to slide. We went ahead and later tried to boost Amiga sales by releasing WordPerfect Library for the Amiga, but sales for the product were miserable. DataPerfect 2.0 sold only well enough to keep the product alive for one more version. Sales for WordPerfect for the IBM 370 (IBM’s mainframe) were very disappointing.
For COMDEX in 1988, which was a presidential election year, we used “Campaign 88” as our theme. Without any new version of WordPerfect to introduce or any plans for GUI which we felt we could announce, we chose to have fun. Our press conference was filled with buttons, balloons, speeches, a Dixieland band, and very little substance. We had won our first battles with Microsoft decisively, so we could enjoy our popularity. Our war was far from over, but it was clear that Word for DOS was not a potent weapon.
It was a great feeling to beat Microsoft, because they were a worthy competitor. If you play chess with a three year old and win, it is not as satisfying as beating a grown-up. Microsoft was, in effect, a grand master, which made the game very interesting and the victories gratifying. Bill Gates liked to tell reporters at Fall COMDEX every year that Word would surpass WordPerfect in the coming year. It was very satisfying to prove him wrong over the years.
We ended 1988 with sales of $178 million, up more than 75% from the previous year. Because so many of the packages we shipped were updates, and because we shipped a lot of free copies to correct the DOS and Macintosh bugs, our pre-tax profits were down to about 25%. That amounted to about $25 million after taxes, so we still had more than enough to pay for all the building projects, computers, and furnishings we needed for all our new employees. During the year, we had grown from 554 employees to 1,130, with about half of the new employees going into customer support.
Unfortunately, I ended the year on a sour note. I broke my ankle four days before the company Christmas party, and having left my temporary cast at home, I was experiencing a fair degree of pain as I hobbled up on the stage to give my short speech. During the previous few weeks I had received more than my fair share of crank calls and anonymous hate mail, so I was not in a very good mood. When I reached the microphone, I told the employees that I was tired of the complaints and that it was time for all those who were unhappy with their jobs to go somewhere else. It had been a tough year and I was out patience.
To my surprise, a lot of the people in the audience applauded my comments. Most of our employees were hard-working and very grateful for the good salaries, the good benefits, and the generous year-end bonuses we provided. They were as tired of the complaining as I was.
If we could have found a way to get rid of all the griping, we would have had the perfect place to work. I wish we had made it a requirement for all employees to work somewhere else before coming to WordPerfect Corporation, so they could have understood how lucky they all were. If I ever start another business, I will have all employees sign a contract requiring them to come to work with a good attitude. If they decide one day they do not like their jobs or they do not want to be cheerful, then they will agree in advance to an immediate termination without notice, without severance, and without receiving any accrued vacation pay. Life if too short to spend it with the dissatisfied.