Index Home About Blog
Subject: Re: SGI is a mess [really: high-tech company quarterly dynamics]
From: (John R. Mashey) 
Date: Jan 09 1996
Newsgroups: misc.invest.stocks

Receiving a Parametrics Technology annual report today reminded me that
I'd promised a couple of emailers a short essay to answer the
        Why do high-tech companies claim they do not know how a quarter
        will come out until fairly late in the quarter?

Don't bother to read this if you expect any immediate insider stock tips.


I have before me the 1995 annual report of Paramatric Technology Corporation,
a successful maker of CAD software.  [This mention is not an endorsement,
or recommendation, just a handy example of something you will see frequently,
and also to pick one outside Silicon Valley.]

As background, PTC's revenues have been:
1991    $ 49M
1992    $ 98M
1993    $179M
1994    $270M
1995    $394M
This of course is a substantial growth pattern. The annual report states:

        "Consistent with past experience, a high percentage of the Company's
        revenues are expected to be realized in the third month of each
        fiscal quarter and tend to be concentrated in the latter half of that
        month.  The Company's orders early in a quarter will not generally
        be large enough to assure that it will meet its revenue targets for
        any particular quarter.  Accordingly, the company's quarterly results
        may be difficult to predict until the end of the quarter, and a
        shortfall in shipments or contract orders at the end of any particular
        quarter may cause the results for that quarter to fall short of
        anticipated levels."

Some businesses have this characteristic; some don't.  SGI certainly does,
and I'll show SGI's comments a little later.

Without talking about any company's specific numbers, but relying on
various past experiences and random Silicon Valley discussions, it's not
unreasonable to guess that hockey sticks might look like:
                        2weeks+2weeks   OR, EVEN WORSE:
Month 1         20%     10+10           15%     5+10
Month 2         30%     10+20           20%     10+10
Month 3         50%     20+30           65%     20+45
The awful case at the right means that almost half the quarter's revenue was
booked, mostly likely, during weeks 7-10 ... and in fact, it is quite posssible
for many deals to have been up in the air until the last week.
Lest you say that is crazy to do business this way ... I agree! ... but
it's a fact of life for many companies, for a bunch of reasons.


Q: How likely is it that you will be having "a tough quarter" if you work
for a fast-growing, profitable, aggressive, high-tech company ...
with hockey-stick patterns?

A: Almost always.

Q: How can that be so in successful companies?

A: If a quarter seemed "easy", that is the orders were locked up in week 5 ...
everyone will decide the goals are too low, and that business is being left
on the table, and the company should be more aggressive ...
[Since 1983, i.e., ~50 quarters, I think I can remember a few that wer
considered "easy".  I can remember many more considered "tough";
sometimes the numbers were made, and sometimes they weren't.


Hardware systems companies have many of the same issues as software companies,
but have a stronger requirement of having to order and obtain, in advance,
all of the components to build their hardware.  [Software companies need
to duplicate software, documentation, etc ... but don't have to worry about
ordering N one-off special-purpose chips, perhaps 3-6 months in advance.
The market will savage a company if it guesses wrong.  A famous example
would be Apple, where once:
        a) Order too much inventory, fail to sell resulting products,
        have warehouse full of (rapidly-getting-out-of-date) items, have to
        sell them at fire-sale prices.
        ... Analysts and press stomp all over them ...
and then later:
        b) Forecast overly conservatively, order insufficent PowerPC
        CPUs and/or ASICs (some of which easily can have 6-month lead-times)
        Big demand, which cannot be met; customers angry.
        ... Analysts and press stomp all over them again ...
[This is not a criticism of Apple ... I have sympathy for them, as this
happens to lots of companies.]

Note, of course, that in heavily-integrated computers:
        a) Many chips, disks, and other items are used.
        b) Some items, like DRAM memory chips, some disks, etc, may be
        commodities, and if you guess low, maybe you can go to the spot
        market, or qualify a replacement.
        c) For some items, there is exactly one source, which sellers like.
        Buyers prefer "multiple sourcing" when they can get it.
        d) In many cases, missing even one chip means that the whole computer
        does not ship.  It is extremely unpleasant to have a bunch of
        $100K computers, each awaiting one of a single-sourced chip, and
        have the fabrication plant "lose the recipe" that month, with the
        result that you get zero chips that month.  Ugh.  Not fun.
        Likewise, suppose a big order gets delayed a week [weirdness in
        government, competitor announces new product and customer puts hold on
        order you thought was ready to ship, etc, etc.]
        Suppose you build $10M worth of computers ... but something happens on
        the last day to delay them one day.  If you have 50M shares, you
        might have just taken a $.20/share hit.

Of course, if you are a fast-growing company, you have to be buying inventory
ahead of the time for putting it together.  This ages fast; it is not like
wheat or oil; a top of the line microprocessor, kept a year, is no longer
top of the line.


At the end of each quarter, you may have a backlog of orders for the next
few quarters, or even years, i.e., orders scheduled for delivery later,
but which you know about earlier in the quarter.  More backlog means
people get to sleep easier, as it is more predictable.  Of course, it is
difficult to have very high backlog when you introduce new products very
rapidly, and your customers know that.   Since the business is moving
faster, it may well be that backlogs go down as a percentage of business ...
thus incurring more volatility.  A good example would be Cray Research,
whose largest machines have often been ordered 1-2 years in advance, since
only a limited number could be built, and a customer needed a "slot".
More recently, Cray's mix has shifted downward to lower-priced systems
ordered and delivered in much shorter times. You sometimes see the same
effect in large mainframes, where orders are placed fairly far in advance.
An explicit statement, in this case from Sun Microsystems 10Q:
   "As the Company has reduced delivery
   lead times, backlog levels have declined as a percentage of revenue.
   Increasingly, the Company must generate a higher percentage of revenue from
   new order bookings in the same fiscal period."

As a result, computer hardware companies, especially those making more
complex and/or higher-end systems, normally have annual reports with
thorough disclaimers.  I now break my normal rule of not offering advice:

As an example, I mention a few items from SGI's 1994 report, not to single
out SGI [as these comments are hardly atypical; see the previously-mentioned
Sun 10Q for similar comments], but because the report is handy.

This is in a section called "Factors That May Affect Future Results", which
is several pages long.

"The Company has experienced substantial revenue growth in recent years,
and it plans its operating expenses, many of which are fixed in the
short term, on the basis that its revenues will continue to grow.  As a
result, it may not be possible for management to quickly adjust expense
levels in response to revenue shortfalls.  Further, because of short
delivery cycles, the Company does not have a large order backlog, which
makes the forecasting of revenue less certain.  This uncertainty is
compounded because each quarter's revenue results predominantly from
orders received and shipments made during the last month of the
quarter, with a disproportionate amount occurring in the latter half of
that month."
        [That sounds familiar.  I think that is lawyer-speak for >=50%
        in last month, and more than 25% in the last 2 weeks.]

"This pattern sharply limits the Company's ability to react to revenue
shortfalls within a particular quarter.  Accordingly, even relatively minor
shipment disruptions, which result from factors such as supply constraints,
delays in the availability of new products, an unanticipated change in
product mix, transit interruptions, or natural disaster, may cause a
particular period's results to be substantially below expectations."

        [Then goes on for several pages on all the other risk factors, like
        fact that nearly half of the revenue comes from outside the U.S.A..]
"The Company's export sales to foreign customers are subject to export
regulations, with sales of some of the Company's high-end products requiring
clearance and export licenses from the U.S. Department of Commerce.  The
Company's export sales would be adversely affected if such regulations were
tightened, or if they are not modified over time to reflect the
increasing performance of the Company's products."
        [Or, of course, if the Department of Commerce is inexplicably shut :-)
        Fortunately, the US Government is becoming more sensible on the
        rules, of great happiness to Digital and SGI, both of whom ship
        64-bit computers, which score higher on the government's rules than
        32-bit systems, and hence get hassled more.]

        [Many more disclaimers, including all sorts of events outside the
        company's control.  The Gulf War is not called out explicitly: SGI
        once did get sued, basically, for not anticipating the Gulf War.
        All this ends with the obligatory generic Silicon Valley disclaimer:]

"The Company's corporate headquarters, including its research and
development operations and most of its manufacturing facilities, are
located in the Silicon Valley area of Northern California, a region
known for seismic activity."

        [SGI does not own a semiconductor fab, of course ... but a truly
        major earthquake could have interesting consequences, reminiscent of
        the plot in the movie "A View to A Kill" in which the villain wants
        to corner the world microchip market by creating earthquakes to
        generate a Valley-washing tidal wave.  Fortunately for us, he fails.]


Many high-tech companies have highly non-linear, hockey-stick
shipment patterns during each quarter, sometimes overlaid with
seasonal variation.  If a company's report says so, believe it.
It is more difficult to know what is happening early in a quarter than
most people imagine ... 

Q: This sounds terrifying.  Why on earth do people ever invest in high-tech
stocks, or work at such places?

A: Because aggressive ones do grow fast... and they are fun, and exciting,
if you go for this kind of thing.  

-john mashey    DISCLAIMER: <generic disclaimer, I speak for me only, etc>
DDD:    415-933-3090    FAX: 415-967-8496
USPS:   Silicon Graphics 6L-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311

Subject: Re: SGI *Totally* rules (was SGI Rules) [really: high-tech dynamics]
From: (John R. Mashey) 
Date: Jan 20 1996
Newsgroups: misc.invest.stocks

In article <>, (Steve Mayer)

|>  But seriously: why will SGI do well so long as it's held hostage by the 
|> government? If they couldn't make earnings this qtr, why will they next? 
|> (Okay, the shut-down is smaller, but it's still partially in effect. 
|> Further, appropriations and purchasing orders aren't exactly flowing like 
|> wine...) 

1) If you look at SGI's 1995 Annual Report (year ends in June), or,
you find the following application mix for FY1995:

30%     Manufacturing
15%     Animation
15%     Visual simulation
10%     Chemistry/medical
 6%     Earth Science
 6%     Database/finance
18%     Other

Also, one finds revenue by geography:

52%     North America
28%     Europe
20%     Pacific/other

And finally, one finds that there is the following rough mix of SGI
59%     Desktop
41%     High-end

A significant chunk (but hardly all) of Visual Simulation is
government-related, some of Earth Science might be, some of
chemistry/medical and "Other" might be through government-funded
research grants. From a previous posting, note that there are 2 effects
from the government mess:
1) Slowdowns in purchases by US Government or government-funded researchers.
2) Slowdowns in shipments of high-end machines to other countries, i.e.,
because export-control licenses not making it in the last month of the
quarter.  This only affects companies that:
        a) Sell high-end systems.
        b) Have relatively short book-to-ship times, i.e., if you'd ordered
        a supercomputer in June for delivery in December, you probably had
        the licenses done well before December. I can't recall the exact
        numbers, but it is possible that one of the SGI desktop models
	gets into the needs-export-license category, but usually, this
	would only affect some of the larger systems.  Thanks goodness
	they are relaxing the rules in a sensible fashion over time.

I don't know offhand how much is affected by government issues, and couldn't
say if I did ... but from what's posted above, you can estimate for bounds on
effects.  Note, of course, that as I explained in an earlier posting,
SGI ships a big fraction of its product in the last month/2 weeks of a quarter,
hence it is very easy for government gyrations to affect the overall
numbers in odd ways.  [That is, as explained before, when you're in a
fast-growing business like this, late-in-the-quarter surprises can lead to
gyrations, since you've *already* invested in people/infrastructure, etc,
and you may be able to do some jiggling for inventory ordering and things
like travel expenses ... but it's hard to move most of the expenses around.]

2) There's an article in Saturday's San Jose Mercury News called
"Silicon Graphics strikes back", which basically talks about this coming
Monday's announcements, i.e., a whole set of new products for this quarter
and next.  The following quote is included:
        "Gary Lauer, an executive vice president [he runs worldwide sales]
        for Mountain View-based SGI, acknowledged that the companies earnings
        suffered in the last quarter because customers were awaiting the
        product upgrades."
So: what is that about?
        The Osborne Effect happens when you announce great new products too far
        in advance of delivery, and all your current sales dry up.  This
        put Osborne Computer out of business, hence the name.  That's the
        extreme case.  On the other hand, reasonable companies who sell
        non-consumer products that cost a lot of money, work with their
        customers on future plans.  This is especially true for SGI, because
        the technology moves so fast, and some customers want to do things
        beyond the capacities of any current products ... but within the
        capacity of the next one, and they will actually arrange their
        project plans according to when something will come.  [It is obvious
        that this is fairly far away from consumer product marketing :-)]

3)  Competitive product dynamics
        When competitors announce new products, especially those
        with *major* changes, big deals tends to slow down while customers
        evaluate the new products.   That says nothing about how the deals
        finally come out: you win some, you lose some, but they do slow down.
        There's always a weird period when competitor A  knows everything
        about B's currently-shipping machines, can shoot at B with impunity,
        since B can't get its hands on A's newest machines.  A can pick
        the benchmarks it wants to talk about ... and avoid those it doesn't.
        A can demo against some elements of B's 2year-old technology, if the
        timing is right.  [For instance: a 200Mhz SGI Indigo2Extreme uses
        a CPU that's 1.5 years old, and a graphics unit that is 2.5 years
        old ... that was already being replaced starting in September with
        Impact graphics ... ]  At this point, it is difficult for B to
        shoot back very much, for instance, by buying A's systems, and 
        running *real* applications on them, and seeing what they really do.
        Of course, a few months later, the game changes around, as those
        things happen, and cusomers actually see whatever they see, for
        better or worse.

4) 4Q95: consider the following, all from public data
        a) In mid-November, both Sun and Intel announce new products ...
           and SGI is the #1 target of both; in addition, a whole raft of
           graphics board companies are running around saying "SGI graphics
           in a PC" [whether this is true or not, SGI is almost always the
           target of such things, that is, such vendors seldom say:
           graphics as good as XYZ :-)]
        b) As noted in the Merc, certain customers may have been holding off
           because they really want the new stuff.
        c) And finally, the hard-to-anticipate government mess, which interacts
           especially with SGI's selling cycles & export control, and probably
           produces a bigger effect than on most other computer co.
        a) Is competition that will get figured out the marketplace.  It is
        slightly unusual for 2 major competitors to announce major new stuff
        and go after SGI within a week, but that's life.
        b) Well, announcements are Monday.
        c) Hopefully the government will get its act together :-)

Note: despite *all* of this,  SGI *still* was profitable, and *still* gained
market share that quarter....

-john mashey    DISCLAIMER: <generic disclaimer, I speak for me only, etc>
DDD:    415-933-3090    FAX: 415-967-8496
USPS:   Silicon Graphics 6L-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311

Index Home About Blog