The technical articles on this site are not published as soon as they are written. Every one of them goes through an internal review, and being sent back is a normal part of that. Looking back at the files, the reasons fall into three groups, and none of them is about the writing style. This article sets out those three groups, first because they are the most common, and second because a reader outside the company usually cannot see them.
The first group: the data has no source
The most common case is a figure that is very specific but cannot be traced. It happens when the wording of one model is used for another, or when a single value of a series is written as if it were the value of the whole series. The way to handle it is to take every number back to its source: the parameter table, the test record or the text of the standard, check that the wording matches and only then write it into the article.
There is one easy mistake in the way a parameter table is used: a range must not be turned into a single value. The working vacuum of the ZYD-100 is -0.08 ~ -0.0999 MPa, the water content is ≤3 ppm (GB/T 7600), the gas content is ≤0.01% (GB/T 423), the breakdown voltage is ≥65 kV (GB/T 507), the cleanliness corresponds to NAS 1638 class 5 and the filtration accuracy is ≤1 μm. These figures are right only when they keep the limiting symbol and the standard number that belong to them; written as one particular reading they say something else.
The second group: a standard number added in passing
When an article deals with oil testing, the easiest thing to happen is that a standard number is added in passing while a certain index is mentioned. The habit looks harmless, but a standard number carries an obligation: it means the test was performed by that method and judged by it. Before a standard number is written, two things have to be confirmed: that the test really was carried out to that standard, and that the threshold really comes from it.
The related problem is a standard number that is right while the threshold comes from somewhere else. GB/T 7600, GB/T 423, GB/T 507 and NAS 1638 each govern their own index, and the standard number and the threshold have to appear together and must not be cited apart. Checking the two together is more reliable than checking the standard number on its own.
The third group: a conclusion written as a recommendation
Stating the conditions under which something applies is what a technical article should do, but while writing, a conclusion sometimes turns into a recommendation. The test is simple: the article says under what conditions to choose what, what phenomenon appears and how to judge it, not which approach is better or which machine to buy.
When a recommendation is turned back into a condition, it often takes only a few words, but the nature of the conclusion changes completely. Writing the performance of a machine in a specific condition as a general conclusion has gone past the boundary.
What the review actually checks
- Every number can be traced back to its source, the wording matches the model, and a range has not been written as a single value.
- The standard number and the threshold appear together, and the method named is the one that was really used.
- The conclusions are conditions and not recommendations, and there is no ranking, no market share, no customer name and no list of awards.
- The connection between a parameter and a model has not been crossed over, and a series value has not been attributed to the wrong machine.
- The pictures show what the article is about, and the image paths match the files that exist.
Why the process stays
Once a technical article is published it gets quoted, screenshotted and used as a basis for decisions. The cost of correcting a wrong number that has travelled is far higher than checking it once while writing. The value of these articles rests on every number being right, and the review process exists for that value.
An article is not finished when it is written, it is finished when every number can be traced back to its source.
