BLOG
BLOG
2021/09/28 | Time to read: 4 min
デイブ·バーグはブランド構築から開発、買収まで、技術系企業の変革に20年以上の経験を持つベテラン経営者。共同設計と革新的な取り組みに基づき、洗練されたロードマップで製品ビジョンと戦略を構築することに秀でている。製品戦略担当副社長として、ドリシュティの製品ビジョンの舵取りを行い、ドリシュティ全体のこれらの哲学に基づき製品ロードマップを洗練させています。シカゴに拠点を置いています。
The Toyota Production System (TPS), the gold standard of manufacturing, is at the forefront of a digital renaissance. The manufacturing world is agog with all kinds of new Industry 4.0 innovations that promise to elevate the industry to new, digitally transformed heights. Some of these innovations are legitimate; others just have a really good hype man.
If you’re striving to make your operations look more like Toyota’s, it’s worth noting that TPS is built on scientific thinking. An imperative of the system is to help workers make that kind of logical, methodical, data-based problem-solving innate. But thinking like a scientist doesn’t come naturally to most people, and it can be particularly hard for workers without specific scientific training — which is often the case with assembly line factory workers.
So what role does manufacturing technology play in encouraging scientific thinking?
For manufacturing technology that augments a TPS-based plant operation, there are three ways to integrate software that will extend scientific thinking on the factory floor:
1. The software creates a lot of data that manufacturers use to problem solve
When following the scientific method, a plant worker will form hypotheses based on what the data tells him. For example, if a manufacturing line isn’t producing enough units, the worker will turn to data created from line operations to find the bottleneck, then dig further into the data to understand why the bottleneck exists.
Conversely, it’s easy for a worker to fall into the habit of relying on his expertise and past experience to hypothesize the cause, then look at selective data to validate that assumption. And if you already know what conclusion you expect to draw, it’s often possible to find enough data - taken out of context - to give the impression that a given hypothesis is correct, which renders the data useless and can lead the worker down the wrong path. Further, when a worker is limited in the data available to him because he’s relying on manual data-gathering methods like time and motion studies, it’s even more likely that the data collected will be in pursuit of a pre-formed hypothesis.
For example, if Amit thinks Pete is working slowly on station 2, he may gather time and motion data for a few cycles behind Pete and the line associates at stations 3 and 4, Bill and Monique. If Amit is looking for the performance data to support his hypothesis of Pete being the bottleneck, then he may not collect enough data, or analyze it in an unbiased way, to realize that Pete has been exactly following the standardized work instructions, while Bill and Monique, his “faster” colleagues, have been skipping steps and creating defects, so those units will need to be reworked or scrapped.
In the absence of data, expertise and past experience are good starting points. But as more companies are using AI to create and process data where none previously existed, it is prudent for workers to develop the habit of reviewing and understanding more data before forming opinions.
2. The software presents data in a digestible, actionable format
Data without action does not add value. Good manufacturing software presents data in a simple, practical, actionable way and creates workflows that guide the user through the data analysis process in the context of his or her day-to-day operations.
In the example with Amit, Pete, Bill and Monique, a line balance chart from Drishti will show that the line is unbalanced and indicated where each station ranks in relation to takt time. This view would show Amit that Pete is regularly close to takt time, while Bill and Monique are generally below it. Amit could then use Drishti’s line variability chart to click the shortest cycles for Bill and Monique and watch video from those cycles. He’d immediately notice the standardized work deviations and quality issues with those units.
Amit could then take the video out to the floor and use those examples to spot train Bill and Monique, showing them exactly where steps were missed and reinforcing standardized work instructions. Just as importantly, he could recognize and praise Pete for his excellent performance.
3. The software provides a pleasant and effective user experience
Scientific thinking does not mean interpreting loads of data all at once; people using this process start at a high level and drill down deeper as they learn more. Software built on scientific thinking principles will mimic this behavior, providing clear information rather than overwhelming the user with hundreds of data points.
Take Drishti as an example. The system records video and creates data from every single cycle on the line. A human processing that data would quickly get overwhelmed, and the data would not add value. So Drishti flags the data points that require attention, simplifying the discovery process for the manufacturer. And in some cases, it’s not even critical for the user to understand the data, as long as he can see the impact of what the data shows.
Note: Simple does not mean simplistic! While it may be easy for the user to interpret what’s happening on the line, the backend software is incredibly complicated. It’s kind of like Google: a three-year-old child knows how to use the ubiquitous search engine, but the algorithms that make up its functionality are a mystery to all but a very few.
Scientific thinking is a key component of TPS that doesn’t come naturally to most people, and providing workers with tools and software programs that help drive home its core principles will reinforce those thought processes in other areas of the plant, as well. Ultimately, a data-based, unbiased, methodical approach to problem-solving will ensure better plant operations.
