Decision Making Under Uncertainty

According to statistician George Chacko,
decision making is the "commitment of resources today for results
tomorrow."
As such, decisions are usually made in a situation of
some uncertainty, because we can never be completely sure what
tomorrow will bring.
For example, imagine you were trying to decide between two
candidates for a new sales job. One has considerable experience of
selling in the field in which you operate, but has only an average
performance history. The other has never worked with your type of
product, but she's got a superb track record in another type of
selling. You're effectively "comparing apples with oranges".
How do
you pick the one who will generate the best future sales?
Alternatively, imagine that you're deciding whether you'll invest
in a new project. Given an uncertain future (and therefore
uncertain future sales) how will you decide if the additional
sales you'll generate will justify the additional costs?
This is where you need to manage the level of uncertainty you're
working with, so that you can make a decision based on rational,
disciplined thought.
In both cases, the solution is to quantify the problem, although
each involves a different approach. In the first, you need to turn
qualities like "experience" and "sales ability" into numbers, so
that you can compare them. In the second, you need to understand
the ways that things may change in the future, and factor these
into your decision.
We'll start by looking at how you can quantify your
decision-making. We'll then move on to show you how you can factor
different possible futures into your decisions.
Quantifying Non-Numerical Features
When the uncertainty you're working with arises from having to
choose between unlike options, you'll need to work out how to
quantify the elements of each option, so that you can make a
direct, numerical comparison.
There are many tools that you can use to do this , ranging from
simple plus/minus lists, through to sophisticated grid-based
analyses.
One of the most common approaches that people use when making a
serious decision is to look at the pros and cons (pluses and
minuses) of two alternatives. A good way of structuring this is to
use the Plus/Minus/Interesting technique, which involves assigning
numerical weights to each pro and con. Read our article for more
details on how to do this.
If you need to compare more than one different option, and, in
particular, if you need to rank options in order, then paired
comparison analysis is a useful tool. This approach works by
comparing each option with each other option, each time choosing the best option, and rating how much better it is; and then consolidating
results to give a balanced answer. This type of analysis relies on
a certain amount of intuition, so it's most useful where decisions
are highly subjective, or where it's really difficult to identify
or weigh the importance of the decision criteria.
Stepping up from the simple paired comparison approach, you can
use a more sophisticated grid to weigh your alternatives. Also
known as a decision matrix, the grid analysis approach asks you to score your
choices according to a set of weighted decision factors. It's
particularly useful when you need to compare several different
options using many different decision criteria.

One further option is to use the analytical hierarchy process .
This is most useful when you have a lot of competing factors as
well as different priorities and perspectives to consider. What's
more, making decisions on your own can be difficult – but when you
add the interests and viewpoints of other people, it's easy to
spend so much time discussing the problem and negotiating a
solution that you never make a decision.
The analytical hierarchy process (AHP) was developed to try to
quantify the different needs and values of the various
stakeholders and alternatives, helping you compare them in a
rational manner. You take the intuitive part of a paired
comparison, and then use that to assign weights to each decision
factor. From there, it's easier to quantify the factors, and you
can see the best alternative.
AHP is very complex and time-consuming. For a detailed explanation
and a step-by-step example of analytical hierarchy process , see
our comprehensive article. And remember, before using this
approach, to make sure that a simpler grid method doesn't give you
the answer!
Looking at Different Futures
Our second group of techniques involves looking at the different
ways in which the future may turn out, and making decisions based
on these.
The simplest and most purely numerical way of doing this involves
creating a decision tree . Tree diagrams provide a useful
way of organizing your options visually, and of thinking about
the consequences of each of these. By constructing a decision
tree, you can calculate the risks and potential rewards of the alternatives
in a way that makes it easy to interpret the results.
See our article on decision tree analysis to work through a
detailed example of how to construct one.
Although decision trees allow you to factor in the likelihood of
various futures actually occurring, they can't really cope with
the unpredictable variations or even randomness that occurs in
everyday life. For example, your profits could be affected by more
than just sales levels. What if raw materials prices went up, or
demand was affected by unseasonal weather?
A good way of considering factors like these is to use two
techniques – scenario analysis and Monte Carlo analysis –
together.
With scenario analysis , you think about all of the ways that
things may change in the future, and from this list, identify the
changes that are most likely to occur, and which could have the greatest impact on your decision. For each
of these, you develop scenarios that explore these different
futures. Using our earlier example, you might look at what the
future looks like if raw materials become very expensive, what it
looks like if they stay the same, or what happens if raw material
prices drop significantly.
You then repeat the exercise for one or more alternative futures.
You can then use Monte Carlo analysis to model your decision
across all of these scenarios. (The name "Monte Carlo Analysis"
refers to the casinos at Monte Carlo in Monaco, where hundreds of
chance events happen every day.)
The idea behind the technique is that you set up probability
distributions – representing your scenarios – within the forecasting
model you're using to make your decision, and then feed random
numbers generated by these probability distributions into this
model. After hundreds of sets of random numbers, the consolidated probability distribution that comes out at
the other end shows the most likely consolidated outcome when
everything has been taken into consideration.
This is far from straightforward to carry out, but it's
considerably more sophisticated – and comprehensive – than a basic
tree diagram. What's more, it gives you a very good understanding
of your decision, and of the way that the future may turn out.

Key Points
Unless a decision is entirely a matter of personal choice (would
you prefer to have the team meeting in the conference room or in
your office?), most decisions involve some level of uncertainty.
That doesn't mean that you have to resort to guesswork to make the
decision.
One approach is to quantify the non-numerical aspects of the
options between which you are choosing. Another is to consider the
most likely alternative futures in which your chosen option may
exist, and look at what the outcomes of your decision are likely
to be in these futures.
We've listed many different tools that help you make decisions
under these types of uncertainty. Practice using these to evaluate
your options: while you'll never be able to eliminate all
uncertainty in your decisions, they'll give you the skills and
confidence you need to find your best alternative, given the
information you have available.