The Taguchi method is a tool used in engineering and manufacturing to improve product and process quality and reduce defects at an early stage. It is a statistical analysis approach used early in the product development process that reduces your costs and improves product quality and reliability, so is a popular way to develop products that are less risky to manufacture. Let’s dive into it in detail here.
Prefer listening to reading?
What is the Taguchi method? (Summary)
The Taguchi method (also known as ‘robust design’) takes variables for manufacturing a product and helps you to understand the ideal operating window to produce them to your expected standard and reduce the effect of any external ‘noise’ on the output of a product (or a manufacturing process). For example, the output of a lightbulb is the lumens that it emits, whereas the noise that might affect that output is moisture, static, variations in electricity, etc, and these noises can’t necessarily be controlled. We need to find a systematic way to produce lightbulbs consistently that perform as expected, despite external noise.
The method systematically compares numerous factors and their interactions at the same time with as few experiments as possible. Statistical analysis is done on the data, to find the best operating window that improves performance and is more resilient to external factors. Benefits include better product reliability, improved customer satisfaction, lower manufacturing costs, and a shorter product development process that gets products to market sooner because it identifies the best operating window.
The way we would use this at Sofeast is if a customer comes to us with a prototype that looks good and works well, but we need to confirm if it is actually possible to manufacture consistently in larger volumes at their required standards. Some prototype products are excellent, but the design is such that they end up not being able to be mass-produced. A good example of this is the TIKO portable 3D printer which included a revolutionary design that proved to be impossible to replicate in mass production. (01:07)
Would a Taguchi method analysis point out risks and provide solutions about how to change the product design and/or manufacturing processes so the product could be made consistently well?
Testing a new product using trial and error to get to the right output could take a long time. The Taguchi method is systematic and you add CTQ and noise (things you can’t control yourself) factors to the system, and it shows a grid with variations and minimal impact. You’ll get a clear answer quickly about which combination of design elements (materials, tolerances, speed of the process, etc) provides you with the least risk when it comes to consistently mass-producing a product. (04:01)
Example: Plastic injection molding.
With the plastic injection molding manufacturing process some product or component designs make it hard to get to the exact result you want and plastic suppliers usually go through a lot of trial and error getting there. They adjust variables that they can control like polymers, cycle times, temperatures, pressure, etc, and there will also be variables like the environment and weather that they can’t control easily. So they make changes to controllable variables in tests using the tooling and keep checking the output of the process until they get to where they need to be with the plastic part or product. Eventually, they fall upon the correct method to manufacture the plastic parts at your required quality standard, or they simply have to admit defeat.
The Taguchi method provides a systematic way to add the high and low limits for the different variables and the noise (such as moisture in the plastic pellets) and, when the data is run using the methodology, the result will be an ‘operating window’ where the product output reaches your standards. If it’s a wide window it’s a positive as you know you have some margin for error, whereas if the window is very tight it’s a red flag as manufacturing will need to be done with very little room for error and therefore is riskier. In fact, in the case of a narrow window, manufacturers may often choose to alter the product design, materials used, etc, in order to reduce risks.
Another issue with trial-and-error testing is that your manufacturer may stumble upon a wide operating window where they can manufacture the plastic parts consistently, but if they don’t really understand the product or material well they may unwittingly be right at the edge of the window without knowing it and then some of the parts may be within and some outside of your quality standard and they will not be able to explain why or even fix the issue easily. (06:36)
Taguchi method is not a recent innovation, it’s a well-established methodology.
This method was developed in the 80s and has been widely used since then. A good example of this is when GM became the first company to develop a method of painting car bodies with water-based paint. Their patented system was developed using the Taguchi method and has proven to be very important for environmental conservation, as GM and other automakers were able to use less harmful paints.
The Taguchi method was developed as a tool for manufacturers to improve products and processes, but it also has a societal and sustainability effect, too, because better quality and more reliable products that don’t fail put less of a burden on consumers and the environment. (12:25)
What do you need for good finished products?
The Taguchi method makes it clear that for good finished products coming off the production line, you need the right product design coupled with the right production process. Injection molding, for example, is not always the best process for making every part consistently and the many variables can result in a lot of different defects if you’re unlucky.
A scenario where a Taguchi analysis would be helpful is where you move your manufacturing to a new supplier and they take your existing plastic injection mold tooling. Your former manufacturer managed to be within the operating window to mold your parts, but the new manufacturer’s environment may be slightly different and this could result in them doing exactly the same thing as the old manufacturer but failing to produce parts up to your standards. Without doing a Taguchi analysis it will be hard for them to determine why that is or to show how the design is the issue (as perhaps the operating window is too narrow for comfort).
Doing such an analysis as early as possible in the product development process is helpful, as making changes to a product design later is increasingly costly and troublesome. (18:03)
Benefits of doing early Taguchi analysis.
The benefits of doing a Taguchi analysis early during the product design and development process, or NPI process, are as follows:
- Reducing product development costs – as this avoids going through a lengthy and costly ‘trial-and-error’ testing process saving time and pinpointing what is actually being tested.
- Improved product quality as early as possible – the further along that we are through product development, the cost of making changes to the product design goes up by 10%, then 100%, then 1000%.
- Better product performance – so where smaller, larger, or nominal (closer to the mean of a process) is better, that level of performance can be pursued more easily, for example, the least possible amount of friction in an engine or more friction for a vehicle’s brakes.
- Shorter time to market – as less development is required, so we can move into production sooner.
You’ll have more assurance that the product is reliable, has a good operating window, and is a robust design.
Famous companies use this methodology due to its benefits:
- Toyota used the robust design/Taguchi methods on some Lexus models to improve fuel efficiency.
- Sony also used it for the original Walkman in the 80s to improve its overall reliability.
- Apple also uses this methodology for the iPhone which is well-known as being a high-quality, highly reliable product. (21:05)
The concept of value loss.
Let’s take injection-molded LEGO bricks as an example. The pins must be quite precise otherwise the blocks won’t fit together properly. The pins have a target diameter of, for example, 0.3mm, but there has to be a tolerance, for example, +/- 0.1mm. Most manufacturers will consider that they have done a good job if they’re making parts within the given tolerance (of +/- 0.1mm in this case) even if some are towards the upper or lower end of the range.
However, Taguchi believes the situation should be approached differently. If everything is manufactured close to the perfect value, that’s exactly what the designer intended and the blocks would fit perfectly.
LEGO is one thing, and if some blocks are very slightly harder to fit than others, it’s likely no one will be badly affected. However, consider a car engine that costs thousands of dollars alone. Parts that are towards the upper or lower tolerance range will lead to shorter engine lifespans and increased risks of breakdowns which negatively impacts the value of the product, even if parts are technically within the tolerance range. Most of the Six Sigma literature, for example, doesn’t recognise this which is arguably a serious blindspot. (24:54)
How do teams who try to develop a new product without using the Taguchi methodology usually get on?
They would have to do a full factorial test following the trial-and-error approach and just hope they get a result that meets the specifications or client’s requirements. If they don’t understand any of the testing processes or methodologies, it’s the only way and is filled with risks. This is very common in China.
For example, with plastic injection molding, running numerous trial-and-error tests is probably pretty fast and cheap, so from a cost perspective it is possible. But even if they stumble upon a method of molding the part to specification today, doing it this way means that there is no guarantee the same method will work tomorrow or in a month’s time as variables and conditions may have changed. What if the weather becomes more humid, the pellets are drier, or the manufacturer adds 33% regrind to the mix? They won’t be able to replicate them following the same method, so they’ll have to go through the whole trial-and-error process again, and that’s where it starts to get costly.
For simple parts where only a little trial-and-error is required to get the product or process that you need, it’s fine, but for anything more complex, a systematic approach to testing is critical. (30:49)
When to follow a Taguchi testing approach?
Waiting until a part or product is in mass production is far too late. Taguchi testing should ideally be done at the feasibility study stage which is very early in the NPI process as its results feed into how feasible your product design is to be mass-produced, such as if the operating window is too fine, there is likely to be a lot of scrap or rework required, etc. If so, a redesign may be warranted before you go any further and at this stage that’s not a big issue. (34:13)
Statistical approaches are not a magic bullet.
Even if you adopt statistical methodologies like Taguchi, they’re not a guarantee. They’re a tool to help you improve results. Industries like the auto industry, still have around 20% rework required on paintwork even though they follow such methodologies. The same for some eyewear frames made with certain processes — major brands still have a 20-25% reject rate on certain frames. It may be argued that in cases like this, the manufacturing process is not yet as mature as it could be, so they accept the best possible result (including defects) that fits the design, sales, and manufacturing team. (36:50)