Heuristic Methods
Monday, November 25, 2013
Most of us use "rules of thumb" in all sorts
of areas of our daily lives.
"When the needle on the fuel level
indicator gets to the red, I know it'll last at least another 20
miles", for example, or "Software with a user interface that's
anything other than grey or blue hardly ever gets through the user
acceptance testing."
These rough rules, based on experience, are
invaluable because they help us to make decisions without further
detailed fact-finding: Drivers know they don't need to divert
immediately to find a fuel station, and the software designer
can save him or herself rework later by making the interface blue
or grey in the first place.
But no one expects these rules to be 100%
accurate: If the car ran out of fuel in 18 miles, the driver
wouldn't be particularly surprised, and if testing showed that
users actually liked an application that had a tasteful green
interface, the developer might raise an eyebrow, but that would be
all.
This is because rules of thumb are only used
in situations where the risks associated with using a "good
enough" approximation are acceptable. No one would use a
rule of thumb such as this fuel usage one for a F1 racing car:
They'd do a detailed analysis of the exact amount required to
cover the race distance.
The formal term for these rules of thumb is
"heuristics". Heuristics are a topic of interest in various fields
from computer science to psychology and philosophy, but the
principle is the same in all of these.
When to Use the Tool
Because a heuristic is a model which offers
only a limited representation of reality, it should only be used
when the speed or low cost of using it makes it better value than
using a detailed set of data. So heuristics are often used in
calculating estimates, for proof-of-concept prototypes, or for
filtering a set of options so that you are left with a shortlist
of a manageable size.
Constructing a Heuristic
Heuristics need to be formalized if they are
to be most useful. This raises them above the level of "gut
instinct", and it also means that they can be shared with
other people.
When you find yourself using your experience to make a judgment,
try to work out the rule of thumb that you've just invoked. If
possible, add this to your procedure manual, or talk about it with
a colleague so that your experience is validated and passed on.
A typical example in a car breakdown call centre might be, "25% of
all reported breakdowns are caused by running out of fuel, so
always start by asking the driver to check this before you
schedule a mechanic."
In this case, using the heuristic allows a certain proportion of
breakdowns to be "fixed" immediately, which means that the limited
number of mechanics can focus on the real breakdowns.
Heuristics can also be created specifically
to solve a problem. For example, creating a diagram or straw
man showing a possible solution will help you identify where
your basic idea doesn't work, or is lacking some important feature,
so that you can refine it more easily.
Another heuristic approach is to take a solution to another
problem, and adapt it to solve yours. TRIZ is a systematic and
highly developed approach for doing this.
Heuristic Checklists
Whether heuristics are used to filter a large number of options
down to a manageable number, or whether you want to make an early
"go/no go" decision, you will usually need to draw up a checklist
of heuristics, rather than simply use an individual rule of thumb.
For example, in the food industry, the following heuristics might
help the product development team decide whether it's worth
investing in taking a new pie to the test marketing stage:
Does the pie look appetizing in its packaging?
Can it be packaged so that it won't be damaged in transit?
Can it be cooked in under 20 minutes, so that busy people will
buy it?
Is its shelf life at least 5 days from manufacture to "use by"
date?
Does it maintain its integrity when sliced (or will the pastry
topping fall apart when being transferred to a plate)?
This type of list will have been developed based on experience of
previous products, and from market research. Of course, there's no
guarantee that a pie that meets all these criteria will prove to
be a success: it might not taste as good as rival products, and so
not attract repeat sales, or the mix of filling ingredients might
make it too expensive. However, the heuristic checklist will help
the development team quickly and cheaply weed out "non-starters".
Key Points
Heuristics or "rules of thumb" are models based on experience.
When you use heuristic, it's important to remember that these
models have their limitations, and so heuristics should not be
applied in situations where there is a high degree of risk
associated with making the wrong choice.
The term "heuristic methods" can be used to describe any problem
solving or creativity technique that involves creating a basic
model as a starting point for further experimentation or
refinement. Heuristic methods are trial-and-error approaches.
Apply This to Your Life
Think of situations where you regularly
apply your experience, in which you could formalize the
rules of thumb that you use, and so save yourself or others
time on future occasions. Possible areas could include:
Filtering resumes for shortlisting,
when recruiting for an in-demand job.
Choosing suppliers.
Homing in on the sources of a problem
referred from customer services.
Next time you have a problem to solve,
consider building a heuristic model of a potential solution.
See if this helps you evaluate whether your idea is worth
working up into a complete solution, or whether it contains
fundamental flaws that mean you need to start again.
Tags:
Problem Solving, Skills
of areas of our daily lives.
"When the needle on the fuel level
indicator gets to the red, I know it'll last at least another 20
miles", for example, or "Software with a user interface that's
anything other than grey or blue hardly ever gets through the user
acceptance testing."
These rough rules, based on experience, are
invaluable because they help us to make decisions without further
detailed fact-finding: Drivers know they don't need to divert
immediately to find a fuel station, and the software designer
can save him or herself rework later by making the interface blue
or grey in the first place.
But no one expects these rules to be 100%
accurate: If the car ran out of fuel in 18 miles, the driver
wouldn't be particularly surprised, and if testing showed that
users actually liked an application that had a tasteful green
interface, the developer might raise an eyebrow, but that would be
all.
This is because rules of thumb are only used
in situations where the risks associated with using a "good
enough" approximation are acceptable. No one would use a
rule of thumb such as this fuel usage one for a F1 racing car:
They'd do a detailed analysis of the exact amount required to
cover the race distance.
The formal term for these rules of thumb is
"heuristics". Heuristics are a topic of interest in various fields
from computer science to psychology and philosophy, but the
principle is the same in all of these.
When to Use the Tool
Because a heuristic is a model which offers
only a limited representation of reality, it should only be used
when the speed or low cost of using it makes it better value than
using a detailed set of data. So heuristics are often used in
calculating estimates, for proof-of-concept prototypes, or for
filtering a set of options so that you are left with a shortlist
of a manageable size.
Constructing a Heuristic
Heuristics need to be formalized if they are
to be most useful. This raises them above the level of "gut
instinct", and it also means that they can be shared with
other people.
When you find yourself using your experience to make a judgment,
try to work out the rule of thumb that you've just invoked. If
possible, add this to your procedure manual, or talk about it with
a colleague so that your experience is validated and passed on.
A typical example in a car breakdown call centre might be, "25% of
all reported breakdowns are caused by running out of fuel, so
always start by asking the driver to check this before you
schedule a mechanic."
In this case, using the heuristic allows a certain proportion of
breakdowns to be "fixed" immediately, which means that the limited
number of mechanics can focus on the real breakdowns.
Heuristics can also be created specifically
to solve a problem. For example, creating a diagram or straw
man showing a possible solution will help you identify where
your basic idea doesn't work, or is lacking some important feature,
so that you can refine it more easily.
Another heuristic approach is to take a solution to another
problem, and adapt it to solve yours. TRIZ is a systematic and
highly developed approach for doing this.
Heuristic Checklists
Whether heuristics are used to filter a large number of options
down to a manageable number, or whether you want to make an early
"go/no go" decision, you will usually need to draw up a checklist
of heuristics, rather than simply use an individual rule of thumb.
For example, in the food industry, the following heuristics might
help the product development team decide whether it's worth
investing in taking a new pie to the test marketing stage:
Does the pie look appetizing in its packaging?
Can it be packaged so that it won't be damaged in transit?
Can it be cooked in under 20 minutes, so that busy people will
buy it?
Is its shelf life at least 5 days from manufacture to "use by"
date?
Does it maintain its integrity when sliced (or will the pastry
topping fall apart when being transferred to a plate)?
This type of list will have been developed based on experience of
previous products, and from market research. Of course, there's no
guarantee that a pie that meets all these criteria will prove to
be a success: it might not taste as good as rival products, and so
not attract repeat sales, or the mix of filling ingredients might
make it too expensive. However, the heuristic checklist will help
the development team quickly and cheaply weed out "non-starters".
Key Points
Heuristics or "rules of thumb" are models based on experience.
When you use heuristic, it's important to remember that these
models have their limitations, and so heuristics should not be
applied in situations where there is a high degree of risk
associated with making the wrong choice.
The term "heuristic methods" can be used to describe any problem
solving or creativity technique that involves creating a basic
model as a starting point for further experimentation or
refinement. Heuristic methods are trial-and-error approaches.
Apply This to Your Life
Think of situations where you regularly
apply your experience, in which you could formalize the
rules of thumb that you use, and so save yourself or others
time on future occasions. Possible areas could include:
Filtering resumes for shortlisting,
when recruiting for an in-demand job.
Choosing suppliers.
Homing in on the sources of a problem
referred from customer services.
Next time you have a problem to solve,
consider building a heuristic model of a potential solution.
See if this helps you evaluate whether your idea is worth
working up into a complete solution, or whether it contains
fundamental flaws that mean you need to start again.
