ñòð. 159 |

Low Demand

NPV $(1.5m)

16% probability

Fig. 12.7 Swimming pool cover project with â€˜80/20â€™ survey

Since the overall value of $4.9m is below the value of $5.0m if we do not carry

out the survey, we should conclude that the survey was not worthwhile. The

value of the survey is actually quite close to zero and so what we have done

here is, by chance, identify roughly what set of assumptions will result in a

zero value. We will return to this approach in the following section when we

consider the reverse engineering technique.

513 Second view: Valuing flexibility

I think that it is important to reflect on the exact assumptions which I

chose to make about the survey and how they have influenced the valuation. I

assumed that the survey would provide a specific answer about future demand,

in particular whether this was expected to be low, medium or high. This, in

turn, would allow us to make a clear decision concerning whether or not to go

ahead with market entry. Had I assumed that the survey would simply place

probabilities on each of the three possible outcomes then I would not have been

able to apply this technique. This is for two reasons. First, there is the generic

reason that I cannot know in advance what the outcome will be and so must

assume that the survey tells me what I already know. The second reason is spe-

cific to this case and is that that unless the survey comes up with an answer that

is very different from our initial assessment, we will approve the investment.

The message that I take from this is very clear. This is that if we want to

undertake a valuation of a study it would be vital to agree with the author of

the study exactly what type of answer they will be giving us. We would then

need to think carefully about this and whether, in principle, it would enable

us to take an improved decision and also whether, in reality, we would have

enough confidence in it to allow it to dictate our decision.15

I have illustrated the technique here for a market survey mainly because

it presents a fairly simple situation which can be explained in a few pages as

part of this book. A similar approach can be used to assess other situations

where there are multiple decision points between the present and the ultim-

ate realisation of value. Exploration and research and development are two

examples. In each case one is faced with a series of steps and one does not

need to justify the full spend at the outset; one just needs to justify the funds

which are going to be committed prior to the next decision point. As I have

illustrated, one can then use this to assess the value of a particular piece of

information and hence take a rational decision about how much money it is

worth spending in order to get it.

The usual warning must, however, be given about the quality of the deci-

sionâ€™s being limited by the quality of the assumptions. One particular concern

that I have is that studies are often not as clear-cut as one would like. So, for

example, my assumption in the above case study that the recommendation

The approach which I have outlined is what I would do if I had not initially thought that a survey was

15

necessary. If a new and major expense has come to light, it needs to be justified. There is an alternative

way of justifying the expense of a survey. This is simply to assert that an investment is too significant

to make based on the knowledge that your company has. Expert opinion is required in order to plug

a gap in the host companyâ€™s knowledge. This assertion should be made at the outset when the budget

for the development team is first put together. The survey is then considered to be just a part of the

overall cost of project development and this is what one must justify.

514 Three views of deeper and broader skills

of a study will always be followed is probably far too optimistic. Any ten-

dency to reject a study would greatly lower its value.

A further concern that I must raise with decision trees is that they can

quickly become overly complex. There can be a need to deal with so many

nodes that, in my view, one faces a great danger of becoming a slave to a

model and never really being able to trust the answer. Special software pack-

ages are available which facilitate the building of large decision tree mod-

els and, in theory, these can have advantages in allowing one to calculate

value and identify optimal courses of action in highly complex situations.

The problem with these is that the ease of computation for a skilled user of

the software16 may disguise the need to have quality assumptions. Unless suf-

ficient skilled people will be available to provide the necessary assumptions,

a large decision tree cannot be trusted.

Step 8G: Reverse engineering

Reverse engineering is the term which is used to describe a situation where,

instead of making assumptions and calculating the implied value, one speci-

fies the required value and then calculates the necessary assumption. The

typical calculation is done by specifying a zero NPV and then identifying

what assumption, or set of assumptions, would result in that answer. One

can then inspect the necessary assumptions, decide how credible they are,

and use this to guide a decision. The more difficult it is to make assumptions

concerning the impact and probability of a flexibility, the more appropriate it

will be to use the reverse engineering approach.

The results of reverse engineering can be presented as a single-point

assumption if just one variable is being considered; as a line of possible com-

binations if two assumptions are considered; or as a series of lines if three

assumptions are in doubt. If more than three assumptions were in doubt,

then my advice would be to try to identify the three which were of greatest

importance and to focus on these.

The decision-maker will be faced with one of three possible outcomes con-

cerning their judgement about the necessary assumptions or combination of

assumptions which are necessary to give a zero NPV:

1. They may appear highly conservative. In this case the decision can be

approved with some confidence even though its exact value is not known.

My experience is that there are o ly a f w such s l d users of the software and so t c i l assist-

e ien e s hat here re only few skilled ers he o re n o technical s s -

16

ance is typically required in order to deal with complex decision trees.

515 Second view: Valuing flexibility

2. They may appear to be roughly what one might think expected value

assumptions would be like. One would perhaps investigate a little further

to see if any additional insights might make the decision easier to take.

In any event, the decision-maker could be content that the decision was

marginal and that getting it â€˜wrongâ€™, if they were to do so, would be an

understandable error.

3. They may appear to be highly optimistic or perhaps just impossible to

believe. In this situation the decision can easily be turned down.

Examples of the reverse engineering approach abound throughout this book.

This is because it is not just of use in valuing flexibility. It is of use in con-

sidering any financial decision and is very simple to apply. The importance

of understanding the conditions which lead to a zero NPV lies behind why

I have suggested that a special category of sensitivity, the type 2 sensitivity

should be defined.

The more difficult it is to make good assumptions, the more useful the

reverse engineering approach becomes. This is why it is so useful in valuing

flexibility. It is difficult to make good assumptions in relation to high-impact

low-probability events but one often finds that one can be spared the neces-

sity of doing so because a clear-cut decision can be made without knowing

its exact value. So, for example, if I was considering whether to purchase

a diesel generator to protect against spikes in the electricity price I would

first simply ask what set of assumptions would justify a zero NPV and then

decide whether I thought these were conservative or optimistic. In either case

I would be spared the problem of making a difficult judgement regarding

exactly what a good assumption would be.

We can return to the swimming pool cover example that was developed

earlier and use this to illustrate how the reverse engineering technique could

be applied in practice. Let us recall the decision which we faced. We were

trying to decide whether to spend $1.5m on a survey in a situation where we

were fairly confident that we should go ahead and enter a new market. Our

initial analysis showed that if we had full trust in the survey it would add

$0.5m to value. We did not find the exact zero value set of assumptions but

we did show that if one thought that the survey would correctly identify the

low-demand world 80% of the time, the survey would have a small negative

value. Had I been a decision-maker I would have felt able to take a decision

based just on this additional piece of information and the reverse engineer-

ing approach. The information would have told me that I needed to believe

that the study would correctly identify the low-demand situation with a bet-

ter than 80% probability. I would stop the analysis at that point and say that,

ñòð. 159 |