A. Decision Making
The Risky Ventures Finance Company (RVFC)
Designed by Bud Banis, September, 2001. (Hints: I've founded it most
useful and flexible to set up simple trees like this laid out on an excel
sheet to do the calculations. The drawing toolbar--view/toolbars/drawing--
gives you arrows and other forms to visualize relationships and it's easy
to move things around and do recalculations for sensitivity studies. I'll
put up some examples at the website)
RVFC: Decision Tree Problem on the value of Imperfect information
The Risky Ventures Finance Company (RVFC) takes chances on hundreds of
projects per year, and typically works with projects that have only a 30
percent probability of success. Since most of these projects are
short term fads requiring rapid rollout, there is a lot of money committed
to national rollouts before they know if a product will be successful.
Successful products have an expected net present value of $10 Million
each. On the other hand, failures are also spectacular, resulting
in a loss of $6 million each. Assume shutdown would have a net cost of
A. Suppose RVFC didn't have market research capabilities? What should
they do with the business?
B. Risky Ventures manages to be successful at this business because it
has an extraordinary market research function.
For a cost of $20 thousand per project, they are able to do a preliminary
survey that makes it possible to discriminate somewhat between successful
and unsuccessful products.
Half of the projects have a favorable result on the survey, and of
those 50% will give a successful program. Products with bad results on
the survey have only a 10% probability of success if they are taken to
market. The benefit of the survey information is that RVFC can decide not
to pursue a project after spending only $20K and finding it gives a bad
result on the survey.
Alternatively, test marketing costs $600K but gives better discrimination.
Only 25% of the products give a good result on test marketing, but 90%
of those will be successful when rolled out. Only 10% of those (the other
75%) that look bad on test marketing will be successful if rolled out nationally,
and, of course, RVFC can decide to flush the project if it has a bad result
on test marketing, and suffer only the $600K cost.
Draw a tree comparing the four options:
1) Shut it down.
2) Take every project to market without testing. $10
M NPV for the 30% that succeed, $6M loss for each of the 70% that fail.
3) Survey each project before deciding whether to take
it to market. Cost of test $20K per project. 50% give a good test result.
50% of those succeed when rolled out. 50% give a bad test and only 10%
of those would succeed if rolled out. Remember you have a decision after
the test results to go/nogo.
4) Test Market before deciding. Cost of test is $600K
per project. 25% give a good result, and 90% of those succeed when rolled
out. The other 75% are mostly losers and 90% would fail if rolled out.
Use squares for nodes where you are to decide, and circles for nodes
where you have no control. At the terminal nodes of each of the paths,
calculate the NPV, ignoring adjustments for time value of money. Analyze
the tree by working backwards from the future, assigning values to the
nodes as you trim back and summarize branch points. Where you have
control, pick one, else use probabilities to calculate an emv for the node
where you don't make a decision.
Eventually, you will be back to the branch corresponding to the original
Is it better to follow option 1, 2, or 3?
What is the value of the survey information?
What is the value of the test marketing information?
B. Effect of future options:
Suppose some of the losses from unsuccessful projects could be recovered
by either diversifying the resources to other projects or selling them
off to other entrepreneurs. Then, if a project were unsuccessful,
you could either continue and lose the whole $6M, or else diversify and
lose only $5M, or sell the assets and lose only $2M. Make another copy
of the tree, adding these options to the ends of branches with unsuccessful
What effect do these new options have on the initial decision?
With these new options, is it best to go without testing, to do surveys,
or to test market the candidates?