Predicting the Unpredictable

copyright 1997, Rex Ballard

Re: Gartner Group nails coffin on linux

(Maybe Not)

In article , wrote:
> I don't like Linux, but the Gartner group is, and always has been way
> off course.
> They are like the Physic hotline. 
> Has anyone actually kept statistics
> of their predictions.

> On Thu, 14 Oct 1999 19:36:46 -0400, "Drestin Black"
>  wrote:
> >A damning report from Gartner has 
> >all but put the kiss of death on Linux,
> >but devotees of the OS maintain it's a
> > force to be reckoned with.
> >"Linux is the 'hype du jour' that is
> > thought by some to have the potential
> >to upset Microsoft's dominance on the
> > mainstream desktop," the Gartner Group
> >report says.
"We believe Linux deployments for desktops will not usurp OS dominance from Microsoft. "The lack of standards in the Linux community, coupled with a lack of key productivity applications and with Unix complexity, will continue to make Linux a poor choice for the mainstream business productivity user.

Gartner has made many interesting predictions:

   In 1993, they predicted that Windows NT (3.5) would dominate
            the Desktop market with 80% within 2 years of it's release.
            (NT 3.5 became 3.51 and failed miserably).

   In 1994, they predicted that the Internet would grow to 6 million
            users by 2005.  (later that year IDC estimated the user
            base was already nearly 16 million)

   In 1995, they predicted that Windows 95 would have 80% of the
            market by 1997.  (Windows wasn't even close to being
            released yet, and Windows 95 didn't install on legacy
            hardware very well, if at all).

   In 1996, they predicted that Windows NT 4.0 would have 80% of
            the internet server market by 1998.  Today, NT has
            less than 30% of the server market.

   In 1997, they predicted that Linux would never be a dominant 
            player in the server market.  Today, Linux has 30% 
            of the server market with BSD (freeBSD, NetBSD, 
            and BSDi) holding another

How could they have such a lousy record?

Actually, Gartner bases it's predictions on predictible events.

For example:
If 95% of all businesses are growing at average rate of 10%/year over 5 years, then there is a 95% chance that Linux will grow at 10% year too. In 1995, the predictable rate of growth for the Internet would be for it grow at 20% per YEAR. It actually grew at 20% per MONTH for nearly 36 months.

Gartner uses the most conservative numbers available when estimating market size. At the time Gartner made their Internet prediction, there were about 1 million registered host IP addresses. Gartner had no proof or accurate statistical sample of how many users were attached to each address, so Gartner assumed 1 user per address.

The later IDC report based it's estimate on the number of e-mail addresses registered with nic servers (which was still under the mark). It wasn't until the following year that accurate counting mechanisms were available. By then, the user base had grown to about 40 million users.

Gartner analysis does not attempt to factor in innovations and breakdowns. If Microsoft released a product (such as Office) in 1992 and that product took 40% of the market per year in the first two years, the assumption is that their next product would have the same penetration. Until there is tangible and verifiable evidence that there will not be such acceptance, Gartner can only go an the basis of past performance.

In 1979, any technical analyst would have told you that by 1999, interest rates would be 40% or higher, gasoline would cost $10/gallon (if you could get it), and unemployment would be 30% or higher. Because that was the economic trend from 1976 to 1979. Prices rose, inerest rose, unemployment rose, demand of fuel increased, and unemployment increased.

But here we are, in 1999, with the the economy in the best shape it has been in this century.

What Gartner couldn't predict was that someone at Dow Jones would start passing out copies of Trumpet and Mosaic, would give 8000 publishers detailed instructions on how to create their own server that would be accessible through a shared online services being developed by MCI that would be shared with AT&T, Sprint, British Telecom, and every other major telecommunications company.

When you think about it, the probability of any one of those things happening would be nil. Publishers do not give away trade secrets. Online services do not provide unlimited access to any server wishing to attach. No one would use share-ware for anything other than games, and MCI would never allow interconnectivity directly with it's competitors.

But here it is, 1999. Nearly every telecommunications carrier has the capacity to route TCP/IP traffic. The adoption of these standards and the interconnectivity has resulted in mergers that would have been science fiction 20 years ago.

These are breakthroughs, events or shifts in thinking, policies, and relationships that were "never going to happen".

To Gartner's credit, once they could identify a breakthough thay immediately and radically altered their predictions. It's not that Gartner is incompetent, they are just responsible.

Tracking a breakthrough industry like the Internet or Linux is incredibly hard. The industry is densely populated with people who produce breakthroughs on almost a daily basis. In addition, there are breakdowns. There was no indicator that thousands of Windows 3.1 applications would fail under NT 3.5. There was no indicator that Windows 95 would be nearly impossible to install on a VLB machine (which comprised 60% of the market and 90% of those most likely to upgrade).

Linux is even more difficult to track. There are 5 major Linux distributor companies who generate their own distributions. What makes this very unusual is that they compete primarily on the basis of service. Each uses almost exactly the same software packages, but each competes by asking "What would make it easier for a user to use this system".

Even the "user" has shifted. Up until 14 months ago, most Linux systems were focused on the "user" who was using Linux as a server, or the administrator "users" of that server. In reality, it has only been since the releases in May and June that releases were directed toward a "desktop user". Many prior users used Linux as a desktop machine, but they were willing to deal with some configuration chores to have a workable machine.

Linux can effectively appeal to three different markets:

The power user. These are the people who bought the first copies of Windows 95 and tried to install it themselves. These are the people who used NT on their desktop. These are the people who created their own web pages using HTML. If they could afford it, they'd have a Sun ULTRA workstation or an Indy, but Linux on a 500 Mhz Pentium III would be a pretty close second choice. Maybe next year they'll get the ULTRA.

The low-income user. These are the people who couldn't afford to pay the $1000-$2000 for a machine preloaded with Windows NT and Office 97. Some of them may have been the software pirates that Microsoft despises anyway. In this case, Microsoft isn't even loosing the market, but Linux is gaining an invisible one.

The appliance user. These are people who really don't know or care what the operating system is. They want something as reliable as their television or their telephone, that they can use for days at a time without even thinking about CTL-ALT-DELETE.

Appliance users aren't just the "set-top-box" crowd either. They are the clerks at millions of retail and franchise stores (many of whom have used ANSI terminals connected to SCO servers for years. They are customer services clerks who have to try and keep an irate customer from hanging up while their MS-Windows console reboots.

Gartner has some sense of the size of the "power user" market (about 10 million), and one could estimate the appliance user market by comparing it to the number of cash registers in the world, adding the number of televisions, and then adding the number of cell-phones.

The last market, on which the Gartner author was focusing, was the existing Microsoft PC Windows user market, specifically the Windows 95 and Windows 98 markets. Based on Gartner's requirements to base prognostications on predictable events, it's easy to see why this Gartner analyst would make this prediction. In 20 years, no one, not even IBM, with OS/2, has displaced Microsoft in it's establish market. DR-DOS tried and failed, OS/2 tried and failed, GEM tried and failed, Novell Tried and failed, Coherant tried and failed, and Sun tried and failed. What is predictible, based on statistical evidence, is that Linux will try and fail. Statistically, if they use the same strategy that everybody else has used, they have, a best, a 4% chance of gaining any market share in the first year, and would die on the vine in the second year. The most penetration they could get, based on the past failures, would be 1%. OS/2 had 4 million users in it's best year after the break with Microsoft. That represented about 1.2% of the total market at that time.

What this author doesn't take into account is that Linux is not using the same tactics as previous contenders. In much the same way that the U.S. minute-men were able to win independence from the British by shooting from behind trees, Linux has already gleaned a 4-5% market share by offering dual-boot options, coresident operation, and forming alliances between the top 5 distributors and the top 10 computer manufacturers.

Finally there's the Anti-trust factor. The DOJ has already successfully staged a very successful campaign against Microsoft. Even if there are no consequences (Microsoft wins in appeal), Microsoft suffered heavily in the press, and once the vendors began breaking ranks, many of Microsoft's carefully woven nondisclosure agreements and secret business practices became unraveled.

Microsoft is already watching it's back, and the DOJ is already collecting the evidence for other charges. Microsoft courtroom antics have been getting almost as much press as their press events.

The analyst assumes that all 600 million Microsoft Windows users (3.1 through 9x and NT) are happy, satisfied customers, with no unfulfilled expectations, and no unpleasant experiences. But if an analyst were to take customer satisfaction surveys which asked questions such as:

Analysts could see that this is a market that is ripe for picking. When one correlates such satisfaction numbers with other companies facing similar levels of dissatisfaction, when confronted by a competitive market, you see a predictable outcome.

When AT&T failed to respond to customers, and MCI was given even the most basic opportunites to compete in that market, MCI grew at an unprecidented rate, and AT&T had to make radical improvements in it's customer service program.

When IBM failed to meet customer expectations for MVS 4.0, it opened the market to competition from UNIX systems with powerful relational databases. Oracle, Sybase, Informix, Sun, and HP grew at an unprecedented rate and, under Lou Gerstner, IBM radically improved it's treatment of the customer. Today, even the Mainframe OS/390 systems feature UNIX compatible interfaces.

And then we have Microsoft. While Microsoft has been very successful at protecting it's monopoly on the PC industry, retaining a 95% share of the desktop and a 99% share of the "PC" (Intel 8086, 80286, 80386, 80486, Pentium, and compatibles attached to a desktop monitor and primary keyboard), it really hasn't been that successful at meeting the needs of their customers.

Windows 286 was released, but only served to invalidate CP/M and DR-DOS.

Windows 386 was also a marketing failure. Quarterdeck took largest share of the "Windows PC" market in 1989.

Windows 3.0 nearly lost the desktop to Sun, who gained as much as 5% of the market in 1990.

Windows 3.1 was also dissappointing, many IT managers began implementing SunOS and UNIX workstations in critical situations such as trading floors, air traffic control, and engineering. NT was announced to stop the loss of market share.

NT was not a smashing success. Sales of NT 3.51 workstation was under 2 million units. Most of those were demo copies included with the Microsoft Developer Network subscription. Backward compatibility with MS-DOS and Windows 3.1 was a big problem, and 3rd party developers had been driven out of the market. All Microsoft had to offer was it's own product line.

Windows 95 was released on the last saturday of the fiscal year, and Microsoft pre-sold 1 million copies to retailers, many of whom did not reorder for months. Users of 80486 VLB machines, and Pentium 66 or Pentium 90 machines not equipped with Plug-n-Play ROM and Logical Block Addressing were having a very difficult time getting the "Windows 95 Upgrade" to install. Many customers paid as much as $1000 for additional memory, hard drive, and CPU upgrade, only to find out that Plug-n-Play didn't work on their "obsolete" machines.

Windows NT 4.0 also had a very rough start. The first service pack seemed to be more about getting IE into the system than fixing any bugs. The second service pack resulted in failures on machines equipped with Cyrix chips, and ultimately became embarassing in court. The third service pack yielded a somewhat stable machine, but increased the minimum RAM required for a stable system from 16 megabytes to nearly 64 megabytes. Most corporate workstations were installed with 32 megabytes. Again, 3rd party software for NT 4.0 was slow in coming, and many desktop users reverted back to Windows 95. In it's second year of production, NT 4.0 had only sold about 4 million desktop machines, less than 1 percent of the overall market. Many of these weren't even fully licensed machines, but rather "educational" and "MSDN" releases.

The following year, Microsoft offered NT to students for less than $100/copy, but business users still paid up to $400/copy. Often, corporations chose to put as many people as they could into MSDN so that they could get NT at a reasonable rate. Many fortune 500 companies purchased license packs in quantities of 25,000 to keep the costs down (less than the price of 5,000 at retail). In some cases, less than 60% of the licenses were installed, but all were included in Microsoft reports on their web sites and SEC filings.

Windows 98 was the latest fiasco. Many users who installed Windows 98 upgrades found that 3rd party software didn't work, and their hardware didn't work. Toshiba, HP, and Dell, among others issued a number of warnings to their customers NOT to try to upgrade certain of their most popular models.

And just when you thought it was safe to go into the water, Microsoft has issued a number of mandatory Y2K patches, many released as late as mid-july. Unfortunately, these patches adversely affect other 3rd party software and some hardware.

Unfortunately, Gartner analysts are not allowed to speculate on the impact of such customer service. All they are allowed to do is observe that 99.9% of all machines sold after January 1st of last year were sold with Windows 98 preinstalled. Based on this, they can only observe that Windows 98 has captured 99.9% of the desktop PC market.

But then again, there is the success of that IMac. And the sudden surge of user feedback containing the comments of very satisfied Linux customers. And there's those surveys in DejaNews which rate Linux number 1 in the Server market, and rate NT last. Or the survey that ranks Linux number 4, just behind OS/2 (Dell, Micron, Gateway, and Compaq want to give IBM total control of the Desktop OS market - NOT!). Mac (very cool if you make a MacIntosh). And BeOS (I'd like to see BEOS available as a Linux application) a Microsoft wannabe.

Then we have Linux, which is supported by 5-7 corporate entities, cannot be controlled by any single company, and is available in Source format to all PC makers.

If you were a PC maker, looking for an alternative to Microsoft, would you rather have IBM (who makes the same stuff you do), Mac (who hasn't been particularly eager to license anything to anybody), BEOS (whom Microsoft could buy with cab-fare) or Linux - where the only thing you pay for is the support program.

And down there at the bottom of the heap BELOW MS-DOS, is NT, 98, 95, and 3.1, in that order. But when we drill down, this survey comes with feedback from the users...Lots of it. Most of the people giving NT high ratings, are comparing to Windows 3.1, Windows 95, and perhaps Novell Netware (even though this survey is for desktops). Those with an informed opinion, comparing NT with anything other than Microsoft seem to prefer anything other than Microsoft.

Now, for that unpredictable breakthrough. What would happen, if a PC, running Linux, were to show up and be displayed and demonstrated at every major PC store in the country?

What would happen, if a PC, running Linux, was available in every classroom in the world?

What would happen, if a PC, running Linux, was running at a Kiosk in every international airport?

What would happen if 500 million users, with absolutely no knowledge of Linux, other than what they have been told by Microsoft, were to have a "hands on" experience with a fully installed, fully configured, fully loaded Linux system, in the next 3 months?

What would this do to the gartner prediction?

Remember, Gartner has hundreds of analysts, and they often don't agree with each other - because they are operating from different sets of statistical facts)

> tek

Rex Ballard - Open Source Advocate, Internet 
I/T Architect, MIS Director
Linux - 50 million and growing at 3%/week!