Wargaming

1
Wargaming
Simulation methods pretend you're in the new
business, and try to imagine what your
competitors might do. Fully automated SimCity,
with business modeling software instead of
teenage games. Manual groups of people
role-playing. Value lets you think about what
your competitors might do. Risk somewhat
disconnected from reality.
2
Wargames for war
Invented by von Reisswitz (Prussia) in 1824
(Kriegspiel). Below is a picture of a US wargame
from 1895. Before the Battle of Midway, a
Japanese participant in one of their wargames
suggested a counterstrike Admiral Ugaki
overruled the result.
3
H. G. Wells
Strangely enough, H. G. Wells, later a pacifist,
invented a game he called "Little War" in 1913,
based on toy soldiers. This is, by some accounts,
the origin of all the shoot-em-up computer games
now popular. One of Wells' colleagues, F. W.
Lanchester, invented a set of mathematical rules
for deciding the winners of wargames. These
rules regulated the ratio of your soldiers to
enemy deaths, for example. (He also built the
first four-wheeled car in Britain and invented
disk brakes). In Japan Lanchester's equations
are used in marketing (doubling your effort
increases sales four times). Most wargames are
decided by umpires, not equations. In Germany
the Kaiser insisted that he would always win.
4
And we're no better
On February 7, 1932, the US Navy had a fleet
exercise in which two carriers commanded by
Admiral Harry Yarnell launched planes for a dawn
raid on Pearl Harbor. They dumped 20 tons of
dummy explosives on the naval base, without
seeing a single fighter in opposition. The
umpires ruled that 1/3 of the attacking planes
would have been shot down by antiaircraft fire
and concluded that "it is doubtful if air
attacks can be launched against Oahu". Over the
next few years, Yarnell was regularly ignored in
his calls for more airpower and more opposition
to Japan he retired, disappointed in 1939.

5
New York Times, Feb. 8, 1932, page 3 below is
December 8, 1941

6
So what about competitive intelligence?
Companies use wargames to plot out marketing
strategy. To quote the Fuld-Gilad-Herring
Academy of Competitive Intelligence, "War games,
the hottest strategic tool, enable management to
examine strategies and initiatives in light of
external competitive conditions. Realistically,
what they do, if run honestly, is to recognize
some of the risks of a new venture. In companies
where there is a tendency to be overly
optimistic, they can be a counterweight. (And,
unfortunately, if your company is too
conservative, they can add to the tendency to do
nothing).
7
What is a full-scale business wargame?
Two days, with a couple of weeks
preparation. Teams do data collection. They spend
one day working on the rules, and then one day
trying scenarios and writing the final
recommendations. (At least according to
Fuld-Gilad-Herring.) Whether you use a
simulation program is probably less important
than whether you look at the results
honestly. Some people take this much too
seriously About 40 mid level managers from the
Latin American division of SmithKline Beecham
PLC, the 9 billion pharmaceutical giant, are
gathered around an elegant mahogany table. They
are outfitted in combat fatigues, although more
than a few also sport tasseled shoes.
8
SCIP survey of CI methods
9
SCIP survey of CI methods
10
What do these things cost?
Cruz has spent more than 100,000 on a
computer-based simulation to help his executives
think through plans to introduce a new consumer
product in the rapidly growing Mexican market.
The product, which leverages off an existing
SmithKline brand, makes great strategic sense --
at least on paper. This two-day simulation,
developed and run by Advanced Competitive
Strategies (ACS) of Portland, Oregon, is designed
to help the team convert ideas into action -- and
hopefully create a sense of urgency and esprit de
corps. (Fast Company, http//blackbox.fastcompany
.com/online/06/sidebar.html)
11
How the full scale process works
A mathematical model is made of the market,
showing how much people will buy at what price,
how much advertising affects the market, how much
brand loyalty there is, and so on. A model is
also made of production costs, investment, and
the like. Then different teams present their
strategies, the consultants put all of them into
a modeling program, and the computer reports an
answer. The consultants may also present
surprises a new competitor enters, a currency is
devalued, interest rates change,
12
A sort of example
The parts for a bicycle cost 50. Our company
currently pays 50 to have them assembled and
sells the result to retailers at 150 who sell to
the public at 250. We have 10 of the market.
So our profit is 50/bicycle. A competitor is
assembling the bicycle parts in Mexico for 25
each, advertising the same amount and thus making
75/bicycle. What if we propose a 25 rebate for
each purchase? It would reduce our profit per
bicycle to 25, but if it doubled the sales, wed
still have just as much total profit. If you
wargame it, you ask what the competitor would do.
Assume they would cut their price, too. Then
well keep the 10 market share, and lose half
our profit theyll only lose 1/3 of their
profit. What if they try a 50 rebate? They
will still make 25/bicycle, but well be out of
business.
13
Message from the last example
You cant win at price-cutting against somebody
who has lower costs. (You really should have
known that ahead of time). You might win by
doubling the advertising budget and emphasizing
that your bicycles are higher quality or let you
ride faster or something. Or by investing in
machinery that will assemble the bicycles more
efficiently. But its not likely that youll get
this kind of information OUT of a wargame because
you have to know in going IN to the game. And the
assumptions in the simulation (about peoples
reaction to price changes, etc) make all the
difference.
14
Many questions in defining a game

• What products are covered?
• What is the time horizon? (how long into the
  future will you look)
• What range of environmental contingencies
  (interest rates, new competitors, etc) will you
  consider?
• For all of these, the larger the scope, the less
  precise the answer but possibly the more
  informative.
• Setting up the rules how do you decide the
  effect on market share of advertising, or a price
  cut, or starting a filthy rumor about the
  competitors product?

15
Too much math?
16
Elasticity

• Elasticity is the ratio of the change in sales as
  a result of some other change.
• Thus, if a 10 increase in advertising yields a
  5 increase in sales, advertising elasticity
  would be 0.5 typically it is only about 0.1
• If a 10 increase in price led to a 2 decrease
  in sales, that would be a price elasticity of
  -0.2 note the minus sign. This number is
  typical.

17
How do you guess these effects?

• Horrible problem. Theres no agreement on the
  effects of advertising, for example. You can
  find a few studies eg

http //www.fsicouncil.org/who/docs/execsummary030
2.doc
http//www.ex.ac.uk/Psychology/docs/courses/
m092/ does.advertising.work.tracy20towner.ppt
But really judging how much you can expect to
gain from running ads is very difficult.
18
Price elasticity
Price elasticity is easier to understand than
advertising elasticity. Inelastic commodities
have sales that arent affected much by price
elastic commodities have sales which are
affected by price. Inelastic commodities are
commodities that are basic necessities people
will not stop buying them if they get more
expensive (e.g. gas for your stove). Elastic
commodities are items where its easier to cut
back, such as restaurant meals. Price elasticity
can be short-run or long-run for example, in any
given week the amount of gasoline bought may not
change much, but over a few years people can buy
cars with better mileage or rearrange their
commuting.
19
Price elasticity for a variety of commodities,
sign reversed (increased price means decreased
sales). From http//www.mackinac.org/article.asp?
ID1247
20
Estimating price elasticity

• On January 1, 1973, the National Association of
  Broadcasters imposed a restriction of 12 minutes
  per hour of nonprogram material during children's
  TV shows effectively reducing the quantity of
  advertising allowed during children's viewing
  hours by 33. Within four months, the price of
  one minute of advertising on network TV increased
  from 6900 to 7900 or roughly 14. Why did a 33
  decline in available minutes lead to a 14 rise
  in the price per minute of advertisements?
• Elasticity 33 minute decline/14 price increase
  an elasticity of -2.3 for price elasticity (of
  advertising!)
• Note that the TV networks were worse off.

21
Public policy implications
I conducted an analysis of the RIAA's market
data from 1992 through 2001. After adjusting
their market figures for inflation using the
Consumer Price Index, I found that the industry
has experienced an average price elasticity of
6.3 (CDs taken alone have an average price
elasticity over the period of 2.8). 2001's price
elasticity was broadly in line with historical
norms. What is the real issue? Perhaps it's that
in 1998, the recording industry was able to eke
out both a small inflation-adjusted price
increase and an increase in unit shipments, and
desperately wants to believe that the return to
historic norms was due to illicit file sharing
rather than the market returning to historical
norms of the past decade. http//www.oreillynet.c
om/cs/user/view/wlg/2496 (ps again the sign of
price elasticity is reversed he says so.)
22
Econometrics of advertising
These studies were done to refute claims that
advertising had no effect on the level of
smoking. It clearly has some, but less than you
would think.
23
Advertising elasticity of beef in USA
For the U.S. beef market, one of the smaller
advertising elasticities is that of 0.0005
reported by Altson, Freebairn, and James. This
suggest that a 10 percent increase in beef
advertising expenditures increases retail beef
demand by only 0.005 percent (10.0 x 0.0005).
Coulibaly and Brorsen obtained even smaller
elasticities (i.e., such as 0.0003).
The relatively larger advertising elasticity is
that of 0.025 reported by Ward and Lambert.
From http//www.ampc.montana.edu/publications/
briefings/Briefing32.pdf
24
Godiva chocolate analysis, advertising in Belgium
Viability of the Advertising Expenditure ?? A
breakeven of 54,167 kg is a 15.8 increase over
the existing 343,563 kg volume ?? This is a 54.8
ton increase, requiring an increase in market
share from 4.3 to 4.9 in a relatively flat
Belgian market ?? Increase in advertising
expenditure is 42 ?? Therefore, the
advertising-sales elasticity is 0.376 (15.8 /
42) ?? If the actual advertising-sales
elasticity is less than 0.376, Godiva will not
recoup its incremental advertising
expenditure NOTE typical advertising
elasticity is 0.1
25
Elasticity of egg sales in Australia
Note some confusing things here the price
elasticity is positive, which means that although
egg prices were going DOWN, egg sales went DOWN
as well.
26
Brand loyalty
27
So how estimate future sales?

• Take current sales and add
• Effect of price elasticity
• Effect of advertising/marketing elasticity
• Effect of brand loyalty
• Random chance
• So
• Sales Current
• (PEprice change)(AEadvertising
  change)(available customers by
  loyalty)randomness
• Dont believe too much of this math
• (AE, PE are advertising and price elasticities)

28
Example

• If you currently have a 5 market share, and the
  advertising elasticity is 0.1 while the price
  elasticity is -0.5
• If you raise the ad budget 50 and cut the price
  10, you might see an increase of 10 (5 from
  the ad increase, 5 from the price decrease).
• But if you think that brand loyalty for your
  competitors is 50 then youll only see a 5
  increase.
• And if you think that random chance is deciding
  50 of purchases, then youll only see 2.5.

29
Realistically

• We simply dont know the numbers involved well
  enough to do simulations well. It is not
  credible that youre going to do business
  analysis with the precision suggested by these
  programs.
• But you can get a feeling for what might happen
  when your strategy hits the market.