Problem solving

When faced with a problem, you need to be able to identify what type of problem it is. This enables you to adapt the right problem-solving method. Broadly speaking, there are two types of problem:

Type 1: return to an initial state

The desired performance of the system has already been achieved by this same system but is no longer achieved today, or is only achieved from time to time. In this case, we find ourselves in a "return to an initial state" type problem. In this case, we need to identify the parameters that prevent the system from being in its ideal state.

Type 2: improving performance

The desired system performance has never been achieved. In this case, we find ourselves in a "System performance improvement" type problem. This type of problem is generally more complex than type 1, where the DMAIC methodology will be used. In this section, we will focus more specifically on problems of the "Return to an initial state" type, and we will see which tools will enable us to solve the problem as efficiently as possible.

Different types of problem (Vinck Iceberg)

Solving a problem means getting people and processes to talk, but you have to know when. D.Vinck summarised the state of our knowledge through the notion of an iceberg:

The tip of the iceberg represents our consciousness, what we are aware of and can quickly either give an answer or say that we don't know the answer but that we know how to find it.

Then this iceberg is separated into two parts, representing knowledge. On the right, what we know, consciously or unconsciously, and on the left, what we don't know.
Obviously, the tools will not be the same depending on the type of problem you are facing.

Type 1: We know the answer to the problem

In this case, simply apply it

Type 2: We know how to find the answer to the problem

In that case, just look for it.

These two cases generally represent 80% of the problems we face. We don't realise it because they are solved very quickly. The difficulty arises when we are not aware of the answer to the problem. The remaining problems represent only 20% of the problems, but unfortunately 80% of the time, because they are more complicated to solve.

Type 3: We don't know the answer, but it's buried in our collective unconscious.

In other words, at least one person knows why it's not working properly.

In this case, we need to get people talking, using tools that will allow us to delve into our subconscious to find the root cause of the problem. The tools we use in this case are the QQOQCP, the 5 Whys, brainstorming, etc. Tools that get people talking.
The reasoning used in this case is of the "Forward" type:

Hypotheses are put forward
Hypotheses are tested using data

Type 4: We don't know the answer and even if we look hard enough we don't know the answer to the problem.

In this case, there's no point in getting people to talk, and brainstorming is of no use because we don't know the answer.

In this case, we need to get the processes talking, and use tools that will enable us to dig deeper into the way the system works to understand the origin of the problem. The tools we'll be using in this case are multi-variate testing, inversion testing, experimental design, etc.... Tools that make processes talk.

The reasoning used in this case is of the "Backward" type:

We generate data that will highlight the origin of the problem
Hypotheses are put forward
The hypotheses are verified

This reasoning is quite different, and we will show in the next chapter that it is much more powerful.

Conclusion

Most companies waste an enormous amount of time believing that all the problems they face are Type 3 problems, when in fact half of them are Type 4 problems. In this case, the tools used are absolutely the wrong ones, because they try to get people to talk when they should be getting processes to talk, and the time lost is considerable. When faced with a Type 4 problem, you need to change your thinking and use the statistical tools available to get the data to talk. From training can help you choose the right problem-solving method.

Backward vs Forward reasoning

As we have just seen, two types of reasoning can be used to solve a problem: forward and backward reasoning.

While the first line of reasoning is very useful for solving "simple" problems where we know the answer, it does not apply when the answer is unknown. In this case, like Sherlock Holmes, we need to generate the clues that will enable us to find the causes of the problem. The objective is as follows:

"Once you've eliminated the impossible, it doesn't matter what's left, however improbable it may seem, it has to be the truth."
*Reasoning Backward, Gregg Young

The aim is therefore to generate clues that will highlight the main source of the problem. There are three types of contrast:

Intra-unit: Within the same product, there are contrasts: one part of the product is good while another part is bad.
Unit for unit: One product is good, while the next is bad
Period to period: Process performance varies over time, sometimes 100% of products are good while at other times only 80% of products are good.

The multi-vari is generally the right tool for highlighting the main cause of variation in performance. To use multi-vari, you need to measure 3 to 5 consecutive parts, usually every hour, several times a day for a few days. This generally provides enough data to highlight the main contrast, even if more parts are sometimes needed.
Once the main contrast has been identified, the study can be continued using the following tools:

Main contrast = intra-unit and defect concentration diagram
Main contrast = unit to unit and inversion test or pairwise comparison
Main contrast = period to period and pairwise comparison

Here are the modules that allow you to analyse your data using Backward or Forward reasoning:

Data Analysis

Statistics in 1 click. Use the power of statistics to find out what's behind your production data from the SPC, APC or IQC modules. Thanks to its machine learning algorithms, the Data Analysis module can be used to understand the origin of machine drift or to differentiate suppliers statistically.