SR&ED Article - Forensic Observation a Precursor to Experimental Development
Forensic observation is a necessary precursor to experimental development. It assists in the formation of an abstract model that describes a 'technological cause' and 'effect' system and a hypothesis we wish to investigate. The 'effect' is a response of interest and a measureable quantity. It is tied to the technical objective we seek to achieve. The cause is made up of factors we wish to manipulate and measure their effect on some response of interest. Such factors are apart of the technological base level we hope to improve.
I'll use an ordinary example I trust the reader can relate to in an effort to illustrate the importance of forensic observation in experimental development.
I started the day by getting into my car to run some errands. I backed out of my driveway and drove off. As I accelerated a knocking sound grabbed my attention. I listened intently to determine where it was coming from. The sound was coming from the front driver's side wheel. I stopped the car at the end of the street and got out thinking I might have a flat tire. In technical terms I defined a hypothesis, whereby the null hypothesis Ho = the tire is not flat versus the alternative hypothesis Ha = the tire is flat. Once I place my gaze upon the tire, I failed to prove Ha - the tire was not flat.
The previous hypothesis is an interesting example. Let's examine it in greater detail. The knocking sound is what we call a response variable; lets call it 'y'. The hypothesis is an assumption of the technological cause system that is creating the sound 'y'. In this case, the tire is the technological region whereby the sound is believed to be coming from; let's define this technological cause system as 'x'. As air is added or removed from the tire its' level changes in relation to the road surface. We refer to the change in air as the 'level of x'. When the tire is inflated I assumed no knocking sound as opposed to the assumption when the tire is deflated resulting in a knocking sound. The abstract model that defines the cause and effect system is shown below.
Since there was no observed change in the tire level, this hypothesis was quickly disproved. How could I justify a hypothesis that the level of air in the tire had anything to do with the knocking sound when the air in the tire did not change? Since the cause model I hypothesised was incorrect, I needed to define another hypothesis. I broadened the scope of my forensic observations to include a larger technological cause system that being the 'wheel' system. Unfortunately I could not see anything unusual with the 'wheel' system and decided to take the car to a local service centre. Perhaps an automobile technician could provide deeper insight that would lead to another hypothesis. Once the car was in the service centre I agreed to pay $40 to have the wheel inspected. After 30 minutes I was told the wheel bearings were shot and needed to be changed at a cost of $400. My curiosity was peaked and I asked if the technician could show me the problem. Once I was escorted to the service bay, the technician began to explain how the sound was coming from the wheel. That was gratifying; we had at least localized the nature of the problem to the wheel system itself and eliminated other potential technological systems. However, there was no direct confirmation the wheel bearings were an issue. The technician didn't take apart the assembly to inspect the bearings. The technician also added the steering arm may need to be replaced. At this moment, I couldn't understand why the technician was bringing up another cause system? He continued to say that he would need to replace the hub with new bearings and then inspect the system to know if the steering arm needed to be replaced. By this time I was getting frustrated. I was being told the knocking sound was related to a couple of technological systems with no observable evidence to suggest anything was wrong with either of them. Over my 20 years working as an engineer and R&D researcher this picture sounded similar to what I have often observed in an industrial setting. People would often assume what the problem was, point to a technological cause system assumed be the root cause with no direct evidence, spend time and money implementing changes and run the risk of not solving the problem! Unfortunately I see a lot of experimental development efforts run the same way. It starts with a meeting far removed from the shop floor, followed by a boardroom decision to run some trails of which the analysis is inconclusive. This is the worst kind of approach to experimentation! It is not based on a hypothesis derived from a technological cause system that was the result of forensic observation.
Trying not to let my frustration get the better of me, I asked the technician if he observed anything else that was unusual. To my surprise, the technician said yes! He explained 3 out of 5 wheel bolts were loose. So much so that he could remove them by hand! The two remaining bolts were firmly in place. Wow! Just imagine what went through my mind? Could it be that the knocking sound 'y' was some function of the wheel bolts 'x' being loose or tight? The following abstract equation flashed through my mind along with the following hypothesis.
Ho: 2 bolts tight → knocking
Ha: 5 bolts tight→ no knocking
I told the technician to put the wheel back on, tighten up the bolts and let's take the car for a drive. The knocking sound was eliminated. Imagine how I felt? I saved myself $400! If the technician had replaced my existing hub with a new one he would have mounted the wheel by securing all the bolts and I would have incorrectly believed the root cause were the bearings when in fact it was the tightness of the bolts.
Forensic observation is a required exercise. It defines the scope of an investigation by identifying the technological cause system to investigate and helps define the factors that are the basis of a systematic investigation.
I often see projects classified as experimental development without forensic observation to support it. If forensic observation had been conducted, it could have potentially resulted in a solution by means of routine development. However, when it points to one or more technological systems having many factors the likelihood of technical uncertainty increases. The reason for this is a result of a multitude of two or more dimensional interactive effects that often occur in many technological systems. Such effects are often difficult to comprehend through experience and must be derived mathematically to understand. When such knowledge in missing, statistically designed experiments are the only way to undercover such interactions.
When forensic observation is missing, experimental efforts tend to be broad in scope and require extreme effort to investigate leading to a depletion of resources thereby minimizing the opportunity to identify a solution. If a solution is found, it is often unclear what caused it! Within the context of SR&ED is the concept of a systematic investigation. In the absence of forensic observation, postulating a cause system is difficult and tends to be nothing more then guesswork; meaning the hypothesis itself is a guess. In my opinion, this is one reason why experimental development efforts are often 'inefficient searches' rather than a 'systematic investigation'.
I have argued that the start of experimental development occurs when engineering effort employs 'forensic observation' to define a technical obstacle in measureable terms and employs prior knowledge to postulate a model that is thought to create it. Think about the importance of completing this task and the justification it provides for substantiating your SR&ED claim. If the hypothesis is validated, a solution may be implemented through routine development efforts. If however, the model is disproved, it could be argued a reasonable search was conducted and it is found that no prior knowledge exists and technical uncertainty is likely. To resolve such uncertainty requires a new model and a continued search. As this continues, the size of the experimental space increases and so does the uncertainty as to which experimental treatment will reveal a path to a solution if at all.