Locked lesson.

Upgrade

About this lesson

The OFAAT method is often considered the best scientific method for creating a plan of experiments. It is very controlled, and the design performance often grows in capability over time. But it is also the most timing consuming and expensive approach when conducting a set of experiments.

Exercise files

Download this lesson’s related exercise files.

One Factor At A Time.docx
64.3 KB One Factor At A Time - Solution.docx
63.3 KB

Quick reference

One Factor At A Time

One Factor at a Time (OFAAT) is an experimental methodology where many experiments are conducted.  The experiments are designed so that all factors are held constant except one that is varied throughout its normal range.  The factor is set at the optimal setting and the next factor is selected and varied throughout its range to determine the optimal setting.  The process continues by testing one factor at a time, holding all other factors constant.

When to use

OFAAT is best suited for basic research projects and for the characterization of new technologies or inventions.  This technique allows the researchers to define the relationships between the factors and the system performance.

Instructions

The OFAAT method is a disciplined methodology for characterizing how the selected factors impact the system (product, service, or process) performance. The process starts with the identification of the factors to be used in the study.  For each factor, settings are determined that will vary the factor throughout the typical or allowable range of possible settings. 

Once the factor and settings are determined, the sequence of factors experimentation must be determined.  Normally the factors are prioritized based upon which factors are estimated to have the most significant impact.  Or the factors are prioritized based upon which factors are the easiest to manage.

The experiments are then done that vary the first factor through the range of settings.  All other factors are held at the same or constant value for those tests. Once the experiments for that factor are complete the factor setting that has the best performance is locked in as the setting for that factor.

The experiments for the next factor are then conducted.  The first factor is held at the optimal value and all other factors are held constant for these tests.

The process is continued until all factors have been analyzed or until the performance of the system is well-above the desired level.  The target values for each factor are based upon the optimal levels for that factor.

Hints & tips

  • The order of factors can impact the final setting and performance.  If the impact of a factor is linear, the order will not matter.  If the effect is non-linear or curvi-linear, the order will likely be important.
  • Once acceptable performance has been achieved, the experiments can be terminated, saving money.
  • If all the factors must be tested and they have many possible settings, the total program can be very long and expensive.
  • This approach does not test interaction effects between factors.
Login to download
  • 00:04 Hi, I am Ray Sheen.
  • 00:06 Let's consider one of the experimental approaches that you can use to test
  • 00:10 a design or design theory, and that's the one-factor-at-a-time.
  • 00:15 One-factor-at-a-time, or it is sometimes called OFAAT,
  • 00:19 is often referred to as the scientific method of experimentation.
  • 00:24 Unlike trial and error, where a solution or design is hypothesized and
  • 00:28 immediately tested, this approach starts by determining all the factors in
  • 00:32 the design that could contribute to the desired design performance.
  • 00:36 Then holding all but one of these factors as constant,
  • 00:40 the design is tested many times as that one remaining factor is varied from
  • 00:44 the minimum to the maximum level.
  • 00:46 The tests are evaluated to determine what level gave the best overall performance.
  • 00:52 Now that first factor is held constant at this optimal performance point,
  • 00:56 and all of the remaining factors but one are held constant.
  • 00:59 This same set of tests are now done as this second factor is varied from
  • 01:04 its minimum to maximum level.
  • 01:06 The process is repeated until all factors have been tested through out their range,
  • 01:11 conducting a separate series of tests for each factor as it is varied.
  • 01:15 This approach is often used when basic research is characterizing new materials
  • 01:20 products or brand new inventions.
  • 01:23 Like every experimental approach, this one has some benefits or advantages and
  • 01:27 there are some keys to success when using this approach.
  • 01:31 First let's discuss the benefits.
  • 01:33 For starters, it's easy to plan.
  • 01:35 List all the factors and then start testing them one at a time.
  • 01:39 Another benefit and the one that I think is most useful for
  • 01:42 this approach, is that the performance of the new product is improved over time.
  • 01:47 As you find the optimal setting for each factor, the overall system or
  • 01:51 product is often getting better and better.
  • 01:53 If you have prioritized the factors based upon how easy they are to use or
  • 01:58 control you can often stop the analysis before every factor is fully tested
  • 02:02 provided you've achieved acceptable levels of performance with a particular set of
  • 02:07 factors settings.
  • 02:08 There are a few keys to successful use of this approach.
  • 02:11 First, you need to follow the methodology.
  • 02:14 Fully test one factor before you switch to the next factor.
  • 02:18 And the second is one that I have already eluded to.
  • 02:21 You should prioritize the factors.
  • 02:23 You should prioritize them either based upon the ones that you know I strongly
  • 02:28 suspect to be major factors or
  • 02:30 prioritize them based upon your ability to control those factors.
  • 02:34 But like with any methodology, there are some limitations,
  • 02:38 some potential problems, traps, or pitfalls.
  • 02:40 One is that if there are many factors with many levels,
  • 02:44 this approach can take a long time.
  • 02:46 For instance, if there are 20 potential factors, and each one could have 20 or
  • 02:51 more settings, you're looking at 400 experiments.
  • 02:54 A second is that the order of the factors can influence the final performance.
  • 02:59 The optimal level of one factor by itself may prohibit the ability to achieve
  • 03:04 the ultimate performance of other factors.
  • 03:07 If the limiting factor is tested first,
  • 03:09 the ultimate possible performance level may never be achieved.
  • 03:13 Another problem is that you need to make a trade off between the intervals
  • 03:17 between factor levels and the time and money available for running experiments.
  • 03:22 Obviously, if you can use small intervals, you will get a much better sense of
  • 03:26 the factor's relationship to overall system performance.
  • 03:30 But small intervals means a lot more testing.
  • 03:32 For example, I could do just two factor levels, low and high.
  • 03:37 And that gives me a little bit of information, but
  • 03:39 it doesn't really characterize the factor and the sensitivity of that factor.
  • 03:44 However, it is only two tests.
  • 03:46 By the same token, I can divide the factor interval into 20 small steps between
  • 03:51 the low and the high level.
  • 03:53 And I will have a much better characterization of the factor effect but
  • 03:57 I've increased the number of tests tenfold.
  • 04:00 The last point I wanna make is that since we are only testing one factor at a time,
  • 04:05 we don't see the impact of combination effects between factors.
  • 04:09 We might get lucky, and stumble into a best case.
  • 04:12 But let's say that there are three factors that have a combination effect between
  • 04:16 them where the interaction is what really determines the overall system performance.
  • 04:21 This approach will not identify that effect.
  • 04:24 OFAAT is often used in research and
  • 04:26 scientific organizations to characterize new discoveries.
  • 04:31 That is the best application but it's often not a good approach when
  • 04:36 creating a design solution for existing system problems.

Lesson notes are only available for subscribers.

PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.

Gift this course