The technical articles on this website are not published the moment they are written. When a draft comes in it goes through an internal review, and the first thing the reviewer does is not to polish the wording but to go through the text sentence by sentence and find the statements that cannot be supported, replacing them with facts that can be checked. The procedure appears in no product manual, and it is still the reason these articles read a little flat.
First pass: do the parameters match
The first pass looks at numbers only. Every parameter that appears in the body has to be traceable to the product documentation: flow rate, working vacuum, ultimate vacuum, filtration accuracy, water content, gas content, breakdown voltage, cleanliness class, total power, power supply specification, overall dimensions and machine weight. A figure with no source is deleted outright, and it is neither estimated nor written as roughly equal to some value.
It matters just as much not to mix figures of different dimensions. Filtration accuracy and a cleanliness class are not the same thing, and a metric carrying a standard number is not the same as one without it, so the two have to be kept apart in the writing.
Second pass: delete the statements that cannot be supported
This pass deals with the adjectives. The principle is direct: any wording that is superlative in tone, sweeping in scope or sounds like a promise of service is deleted unless it can be supported by something verifiable, and it is replaced by a specific fact or by a clearly stated condition.
- Unsupported praise is replaced by a parameter that can be checked or by a step that can be repeated.
- Assumed openings are deleted, and the article starts from the problem that actually turns up on site.
- Open-ended service promises are replaced by statements with boundaries: the conditions it applies to and the conditions it does not apply to.
- Absolute conclusions are turned into conditional sentences, so that the reader judges whether it applies.
Third pass: turn vague figures into definite ones
Wherever a definite value can be written, no vague expression is used. Temperature ranges state both ends, pressures keep their symbols, dimensions state length, width and height, and standard numbers are kept exactly as they are. A range is never compressed into a single point, a unit is never rewritten, and the names of other manufacturers never appear. There is no exception to that last point.
Fourth pass: check for overlap with the other articles
The last pass looks at the relationship between the articles. Several published on the same day must not repeat a topic or an angle, and two different problems of one machine must not be approached from the same direction. The topic ledger is checked before writing and the body is read again afterwards, and this step exists so that a reader opening two articles finds two different things in them.
| Stage of the review | What it looks at | The usual problem |
|---|---|---|
| Parameter check | Whether every figure can be traced to the product documentation | An estimate written from memory is used as a parameter |
| Statement clean-up | Deleting superlatives and sweeping promises that cannot be supported | The evaluative wording of the draft is kept as it stands |
| Basis check | Keeping standard numbers, units, dimensions and range notation consistent | The same metric appears in two different notations in one article |
| Overlap check | Comparing the topic and the angle against the ledger and the other articles of the batch | Two articles reach the same conclusion from the same angle |
| Structure check | Number of headings, presence of the figure and the table or list, the quotation at the end | Only text, with no table and no figure |
Why the structure is fixed as well
Layout and retrieval both depend on a fixed structure: the body has to be long enough, it has to carry several headings, it has to hold a figure and either a table or a list, and it has to end with a quotation. During the review these are hard conditions, and a draft that does not meet them goes back for rewriting.
This procedure is not flattering, but what it does is let the people reading an article judge the parameters, the evidence and the boundaries in the same way each time. An article that leaves only a conclusion without the evidence behind it is of no use to anyone on site.
The hard part of writing a technical article is not making the sentences flow, it is deleting every statement that cannot be supported and seeing what is left.
