[Short story]Days of reading the contents of the box. The Intersection of Data and the Mind

このnoteは、こちらの話を英翻訳したものです。
This NOTE is an English translation of the story here.

有料設定していますが、無料で最後まで読めます。
You can read to the end for free, although it is set up for a fee.

---

This is the Information Systems Department of a medium-sized toy manufacturer.

I was assigned to the Information Systems Department as soon as I graduated. I was not an expert in computers or systems. There were five new graduates who joined the company this year. The other four are all in their area of expertise.

A woman who was a licensed patent attorney and had already been assigned to the legal department.
Two men were strong in mechanical and electronic specialties and had already been assigned to the product development department. The other, although not a local hire, was the son of a subsidiary factory manager and a relative of the founder of the subsidiary factory. He was assigned to work at the factory from the beginning. I was the only new graduate with no special skills or areas of expertise.

They were putting a lot of effort into hiring this year, and they were supposed to hire 10 people, but 6 people who had been hired by a larger company declined. I was a further alternate to those who declined. It had been decided before I joined the company that I would be assigned to the Information Systems Department, which had been voicing its need for more people for several years. After a month of training, I joined the Information Systems Department and have been doing the same work for the past six months.

This is the verification process of MRP (Material Requirements Planning). MRP is a program that calculates the number of parts required for each process of production, using as input information such as order plans, parts composition, inventory information, and process route information. The program seems to have complex logic and is provided by a major system development company. It is said that the calculations are done using special logic.

If this calculation is not done properly, excessive inventory will accumulate in the process and the timing of the parts production request to be produced by another company will change. It is best to have parts arrive at the process at exactly the right time to produce them, without unnecessary inventory. The program itself has been studied and they are continuously researching and developing the logic to make it even less wasteful and with the right amount of inventory. From time to time, a version is released that reflects the results of that research and refines the defects found. But if you change to the new version as provided, something terrible happens.

It would be nice if all products could be produced earlier, but sometimes what used to take 3 days to produce is now calculated to take 7 days due to a change in logic. Sometimes we find irregularities in the data we have now in the new version because of the correct logic, even though we did not find any defects in the previous version.

The part information registers a part and its child parts; if the materials to make part A are B and C, then B and C are registered as child parts of A. If part C is further composed of D, E, and F, then C is registered as the parent part and D, E, and F are registered as child parts.

To briefly describe the case in question, three parts, D, E, and A, were registered in the C part. Until now, when calculating product A, no error occurred, but the same A in the grandchild parts caused an error in the new version.

When using the grandchild part information in the calculation logic (which was a great-grandchild part to be exact), a situation occurred where the calculation results were not calculated due to my presence there. You might think that there is no such part; A is a paint for toys, which is sometimes added during the process. In this case, I avoided the problem by renaming the paint A for the product and A´ for the paint used in the part in the registration.

When MRP is upgraded, not all good results happen. What may be the appropriate calculation for another company may not be the appropriate logic for us.

That is why we hear that some factories that have been making the same products have been using the same MRP for 10, 20, or even 30 years in some places. There may be a reason why they do not want to pay for the use of the program, but in fact, no matter what logic is used, there is no reason to change the MRP because the calculation results are not much different from those of the MRP they are using now.

Since we use MRP, which is also used by other companies, we submit production data to the new version and the current version of MRP before the new version is released and compare the results. Verification will continue all the way until the decision is made to use the new version or not (if not, the same version will be used until the next version). The main verification is done using production data in the middle of each month, when production volume and variety is high. Every month for six months, we get the production data for the day with the highest production volume and use that as the basis for verification. This is a job that no one wanted to do because it takes a lot of time and I can't do my job, but I like this job a lot.

The morning begins by checking the differences between the output results of the current and new versions that were run for verification the previous day. Checking for strange data, especially differences that have a large production volume and a large impact on the whole, we try to guess why the data was created. Is it due to incorrect data or a change in logic?　What kind of logic was incorporated, or did the logic conditions change, or was a different logic employed? Find features in the pathways and component configurations to get a good guess.

Create the simplest data and run it through the new MRP to check the results to see if it is correct for the reasons you have considered. It is less likely that what I have considered is correct. We repeat that many times a day.

We will continue that process until the evening. If we identify the cause of the problem, this won't work for us. Can you please provide a workaround or a workaround within the version?　and email the developer of the system development company.

Other than me, the people in the information systems department are very busy. Some are always responding to complaints from the factories, others are always making measures to make production work better than it is now or to improve quality, and some are attending meetings all day long to discuss system improvements. When a problem occurs and production does not go according to plan, it means that the original sales plan will not be achieved, so all these people are investigating the cause and thinking about how to deal with the problem, even if it is after regular hours.

On the other side of the room, I am just going through the process of verifying and guessing MRPs. I finish my work on time, plug in all the data that might be a problem, execute the MRP process, and go home.

I really like this process that people don't want to do. It's fun, like solving a puzzle, and I get to see a piece of logic that is very well thought out. I also enjoy dealing with engineers from great system development companies who have PhD's. And they never complain when I go home on time. And they don't complain when I leave on time. Thanks to this, I get to see my girlfriend, whom I met while job hunting, three days a week.

One day I noticed her in a bad mood without warning. At first it was nothing to be concerned about, but around summer, she started having extremely bad mood days. When I first noticed it, I asked, "What's wrong?" After that, I stopped asking her why.

The input parameters are different every day, but the principles of human behavior and thought patterns do not change that much. Every day, we are dealing with programs whose content logic is unknown.

I tried to keep a record of bad moods, good days, and what happened that day and the day before. Much more than work, there are many parameters. I tried to note the weather, her mood the day before, the number of times we exchanged lines, the news of the day, the previews of the dramas she was watching, and little things that came up in conversation. After three months of keeping notes, I finally found the cause of the problem.

He should definitely be in a good mood today.

The breaking news right now is that the Giants are winning by 5 runs in the top of the 8th inning.

Once, when we were talking about watching sports, I asked, "What about professional baseball?" After hearing him say, "My dad likes the Giants," I started taking notes on baseball results. I recorded the number of games played during the time we met, the score of both teams, and the game difference between the leaders.

Today, I'll ask the master of the bar I sometimes go to to prepare a special cocktail for me.

Finally, I'll be able to answer the question.

I had one drink first, at a standing bar near my office, and was about to call the bar when I saw the breaking news: a five-run lead had been overturned and tied in the bottom of the ninth inning.

No special cocktail. I would have to wait another time to see the verified results. Huh... I'm so depressed to see him today just because I'm confident I'm probably right.