(from Derivatives
Strategy, August 1998)
My customers
are all geniuses. After all, they were smart enough to buy my
product, TOPS, which allows them to price and hedge all manner
of exotic options and derivatives. I'll even go so far as to admit
there are some pretty smart people who have bought software from
other vendors (though I wouldn't exactly call them geniuses…).
But
there are still people out there who don't buy from any of us.
Over the years, they've given me a variety of reasons why they
want to go it alone. Naturally, I don't think any of the reasons
are any good. Despite the fact that they say "The customer is
always right," until you've bought a third party analytics package,
you're not a customer, and therefore, in my humble opinion, you
can be wrong. In that spirit, I'm going to share some of the reasons
potential customers have given me for not wanting to buy my product,
along with some unassailable rebuttals. I call them "The Top Ten
Reasons I Don't Need Derivatives Software."
10.
I already have free software from _________.
Fill in the
blank: it can be a dealer, like Goldman Sachs or Morgan Stanley,
who will sometimes give their clients copies of some simple routines
to price options. Or it could be something downloaded from the
Internet as freeware, or something that was written by the guy's
brother in law because he saw the formula in a magazine spreadsheet
column. In any case, the old adage "You get what you pay for"
certainly applies. For one thing, the product will almost certainly
not be supported. For another, if it's freeware, it might have
bugs in it. Even if it comes from a big dealer, it's probably
not going to have the user interface features you're looking for
to save time, or the ability to calculate needed sensitivities
like delta and gamma. And if the program was put together by someone
who lifted the formula from a magazine article or a book, it could
have typos in it. All of which means the software isn't really
free anyway, if it costs you time and aggravation in the end.
9.
What if there's a derivatives blowup?
OK, what if?
If the blowup occurred because a trader was sticking tickets in
a drawer, it doesn't matter whose software was used. The same
thing is true if bad volatilities are being used to mark options
to market, or there are data entry errors, or a trader predicted
the market was going to go up and it went down instead. On the
other hand, if the blowup occurred because a model was misspecified,
or because the coding of a good model was done poorly (isn't that
nicer than saying "There's a bug?"), you'd have a legitimate gripe
if it was a third party program that caused the blowup. But if
you look around at the derivatives debacles of the last few years,
I can't name a single one that was caused by an external software
package. Every instance of model error came from an internal program.
This doesn't mean every third party package is perfect. It just
means that keeping your analytics in-house is no guarantee of
safety. Besides, if there is a blowup in the future, and the cause
was an outside package, wouldn't you want someone outside the
organization to blame it on?
8.
You guys can't be as good as us. After all, we work for a major
bank.
So did we.
That's not to say every software vendor has a team of people from
major Wall Street firms. But those that are should be just as
good at modeling derivatives, if not better. Speaking only for
myself, I can say that the modeling I'm doing now for Savvysoft
is better than what I did at Bankers Trust. It has to be, because
I know more now than I did then. Which isn't to say I was stupid
when I was at BT. After all, we never had a problem with our pricing
models, well publicized or otherwise, while I was there. It's
just that the models I sell now are even better.
7.
The NIH syndrome.
NIH is Not
Invented Here, and it is used to describe companies that want
to be totally self-reliant. Like a hermit that builds his own
log cabin, grows his own food, makes his own bombs, etc., those
with NIH effectively worry about the day something vital in their
world will no longer be available, and they'll die because of
it. It's just as true for an organization as it is for an individual.
The irony is, corporate NIH can actually lead to the event it
is meant to avoid, as witnessed by Apple's near death experience
these last few years. And as one of our client/geniuses pointed
out to management while fighting NIH to buy our software, where
do you draw the line? Do you build your own pricing models to
run inside Excel, and then also build your own Excel compatible
spreadsheet, and then build your own version of Windows, and then
build your own BIOS, and then design and build your own C compiler
and computer? At some point, you have to trust someone else to
be a supplier. And there's no real reason why the line should
be drawn at option modeling.
6.
Our structures are too complex and proprietary to be handled by
an off the shelf package.
That may in
fact be true. But take a good look at a high end derivatives package,
and see if it can't really price your trades. We've worked with
lots of companies, and if Will Rogers never met a man he didn't
like, I've never met an option I couldn't price. When we come
across something new, we add it to our package. That's how we
stay current, and keep our product up to date. But even if we
can't handle the odd deal, or you don't want to tell us your secrets,
that's OK too. Because you probably want your top research people
working on the really hard problems, not doing date routines and
valuing more common derivatives. So farm that work out to a third
party vendor, and concentrate on the places where you can add
value. But also consider the fact that we vendors often come across
a wide spectrum of trades, as we are asked to handle new deals.
What you may feel is proprietary may be something we've already
come across, and it may be a problem we've already solved. It's
worthwhile to at least ask the question "Can you price this thing?"
The only cost is the time spent asking.
5.
We're already making money.
If the amount
of money you made last year is enough, then please do me a favor:
If you make more money this year than you did last year, give
me the excess. Or do yourself a favor, and stop working such long
hours. However, if you would like to make even more money, then
consider how you can improve your operations. And buying a third
party package is a great place to start. It will allow you to
increase revenues by getting into new markets, and save time and
costs by getting into them faster and more cheaply. Besides, even
if you're making money now, that's no guarantee of success in
the future. Markets change, and you'll need to stay on top of
change to stay on top. Don't be left behind by technological obsolescence.
Think of a third party analytics supplier as your partner in a
joint venture, but a partner that doesn't get their grubby hands
on any of your equity.
4.
We have an in-house quant group.
That's OK,
because if you're doing enough business, you should have research
people on staff. But let's say you have only one PhD doing modeling.
Do you want to trust them with your whole business? Probably not.
So for a second opinion, maybe you hire another doctor. But it
would be much more efficient to buy a third party package for
that. That would save you the expense and trouble of hiring a
second person, or, if you want the second person anyway, it would
let them model other products, so you'll get twice the work out
of your staff. The argument holds no matter how big the research
group is: You want a double check on people's work, and it will
always be better to free up existing staff to do what they do
best. Another
less obvious reason you may prefer to go outside for much of your
modeling is that a commercial package is commercial quality, which
means it comes with a proven interface, documentation, is stable,
and market tested by others in your mission critical situation.
In house research groups often concentrate just on the math, without
considering a host of other variables, all of which can lead to
unnecessary operations risk.
3.
I don't need a model. I just call around for quotes.
It seems
like a Catch-22: An option model is no good if it can't give you
prices that match the market, and if you have market prices, you
don't need a model. This argument can be used to avoid both third
party purchases, and also hiring in-house researchers. But the
argument falls apart as soon as you need to hedge the trade, or
measure its risk, or compare two different trades, or exploit
a mispricing, or set target levels, or do any kind of what-if
analysis. Of course, if you don't fall into any of these groups,
why are you wasting your time reading this magazine?
2.
I'll lose my job.
I know about
this one, because it was the same argument I gave at BT whenever
a vendor came in to show us a package. Hey, I was young and insecure.
So I spent a disproportionate amount of time reinventing the wheel
and not doing fun stuff like modeling even more esoteric options,
or going home before midnight. After I left BT, and became a vendor
myself, I went on a sales call to a major bank, and made a presentation
to the head of the desk, and his chief quant. To my surprise,
the quant admitted in front of his boss that he felt threatened
by third party software, and wouldn't want to buy our package
purely for that reason. Well, we didn't get the sale. And where
is the quant now? He decided to become an analytics vendor himself!
1.
It's too expensive.
It's a little
embarrassing sometimes to be selling software that costs more
than the computer. But it's really just because Microsoft can
amortize their development costs over millions of customers, and
those in our segment of the industry cannot. In fact, if you think
about how much quantitative researchers, systems analysts and
programmers earn, and how long it takes to derive option models,
develop user interfaces, code them, test them and write documentation,
and then support and upgrade the software, you'd realize it can
cost millions of dollars and take years to build the same system
you can buy off the shelf for a few thousand and get right away,
saving you money and allowing you to exploit new markets before
the window of opportunity closes. Granted, what you build in-house
will be customized for your needs. But the right third party package
can be used as a toolkit for a system that is just as customized,
and can be delivered in a few months for a fraction of the cost.
Ultimately, buying a third party package has to be cheaper than
doing it yourself, because even though we're amortizing our development
costs over fewer places than Microsoft, we're still amortizing
them over more than in-house groups, who spread their costs over
just one place. In
summary, there is an eleventh reason I'm sometimes given, which
is "Your package looks great, but we already have something built."
To which I reply "There's always going to be new derivatives invented,
and new models developed." So going forward, it's worthwhile considering
third party analytics. Even if you haven't bought a package yet,
it's never too late to become a genius.