Thomas Edison conducted more than 100 different material trials before landing upon a viable solution for the
light bulb. There were many more trials
involved in identifying the need for a vacuum
within the bulb, and a process for mass-producing the filaments. Clearly, with persistence, and the right resources, trial and error
can be an effective means of developing a
That said; given what we know today and
our access to the Internet, any of us with
a resourceful mindset and an engineering
background could have eliminated a great
many of Edison’s trial-and-error experiments with a little research. With access
to powerful computing tools that enable us
to run simulations and calculate outcomes,
trial-and-error is obsolete – or so we are
expected to believe.
Let us first examine some reasons why we
might choose to use simulations and probabilistic analyses instead of trial-and-error as a
means of testing and improving our designs.
1. Trial-and-error is time consuming
Conducting experiments can rapidly chew
through a development schedule. It’s rare that
we are developing something that facilitates
a simple declaration of, “Turn it on, let’s see
what happens,” as a means of testing our
design. We must execute changes to the system or manufacture the materials and devices
to test. We must carefully plan our experiment so that we might learn what we need.
Experiments can be very time-costly.
2. Trial-and-error is expensive
Many of the same elements that cost us
time, also cost us money. System changes,
manufactured elements, personnel, services,
expertise, and transportation of such things
all cost money. The more we demand those
things, the more money we spend.
3. Trial-and-error isn’t always informative
If we get too excited to test and break
our latest iteration, and we don’t carefully
plan how we intend for that experiment to
produce information, data, or knowledge, we
can very easily find ourselves with another
broken prototype and nothing to show for
it. Alternatively, we can run a test and get a
result that is precisely what we expect. If that
is our final product test prior to launch, or our
final system test prior to turn-on, that’s great.
However, if we are in development and only
proving what we already know, we are wasting time and money.
4. Trial-and-error is a never-
ending, unsatisfying method
It is easy to get carried away
experiments. It is fascinating to
run out and test each and every
adjustment to see what impact it has on the
design and the developing solution. If we
allow that behavior, without carefully controlling how much experimentation we conduct,
all of the problems mentioned above become
When we can acquire much the same
information concerning our solution’s probable performance from models, simulations,
analyses, calculations, and probabilistic estimations, we can often save ourselves much
of the cost and time demanded by trial-and-error methods.
However, there are two fundamental
limitations to substituting simulations and
calculations for trial-and-error. First, our
simulations and calculations are based
entirely on what we already know. As soon
as we broach an inference space outside of
information and data already available, we
can’t build a model for our solution. Second,
we must either have a model or a calculation already on-hand, or we must be able to
Sometimes, the development of a simulation to model our solution’s behavior can take
more time and money than a trial-and-error
experiment. Often that challenge depends
upon the information and data available.
While there are several things that can
make trial-and-error methods undesirable, we
should keep two common sense axioms in
1. We learn more from our mistakes than
2. There is nothing more informative than
watching a phenomenon take place
before your eyes.
Do not just assume that analyses, models,
and simulations are as good as or better than
trial-and-error. Instead, consider for each scenario you encounter which will be better.
Trial-end-error should, in my opinion, be
preferred any time that it can be accom-
plished in less time than a simulation for
comparable cost. It is rare that an experiment
would cost less than a simulation, but look
for that opportunity anyway. Absolutely con-
duct the experiment if there is doubt that the
model is rep-
what will actu-
ally occur. Use the information to update the
model when you are done.
If time is not a limited resource and the
cost is not prohibitive, choose trial-and-error
over simulations. Simulations are based on
what we already know. Sometimes they don’t
show us what unfolds; they only show us the
end result. A physical experiment, occurring
before your eyes can be much more informative to a developer. Seeing what actually
happens can inspire valuable and creative
enhancements. In other words, if you have
the luxury of choosing between a simulation
and an experiment, and resource limitations
don’t dictate one over the other, choose the
Don’t let the stigma of antiquated practice
deter you from considering trial-and-error as
on option. Just because it’s been around longer doesn’t automatically mean it’s obsolete.
Instead, make it a habit to consider if it might
be a better option.
Stay wise, friends.
If you like what you just read, find more of
Alan’s thoughts at www.bizwizwithin.com.
By Alan Nicol
• Tim Balz, Founder & President, Freedom Chairs
• Marty Boykin, Ph.D., Director of Consumer
Durables & Tritan, Eastman Chemical Company
• Robin Gray, Chief Operating Officer & General
Counsel, Electronic Components Industry
• Ron Jr. “Reg” Gustafson, Vice President of
Business Development, Clinkenbeard
• Mike Littrel, President & Founder, C.ideas
• Harry Moser, President, Reshoring Initiative
• Alan Nicol, Executive Member,
• Mike Rainone, Co-Founder, PCDworks
• Drew Rink, Senior Manager of Manufacturing
• Paul Scheidt, Product Marketing Manager, LED
• Lanny Vincent, General Partner, Vincent &
• Anna Zevelyov, Director of Business
Development, Artec Group
EDITORIAL ADVISORY BOARD
Is Trial & Error