[ptsefton.com] | [CV & Bio]

How to write and format a technical procedure


About this document


The first job I had after doing a PhD in linguistics, back in 1995 was as a technical writer for TAFE NSW, working exciting stuff like how to set up networking in the MicroVax servers they had running the networks in the computing labs (TAFE does vocational education). The tech-writing team leader was called David Drinkwater, and he taught me some simple rules that have proven to be very reliable*.

The most important rule is to use the imperative voice. That is, tell the reader exactly what to do in each and every step without ‘you should’ or ‘the following needs to be done’.  Every job I’ve had since I have written up some kind of variant of this quick primer for my team, so this time I’m putting in on my own site so I don’t have to rewrite it every five or so years. Suggestions are welcome, via the comments.


This is about reproducible procedures that need to be unambiguous. Think training rather than education. If something can be expressed as a kind of recipe then these guidelines will help to make it unambiguous and easy to digest. This approach is not what you want for discussing the history of food or reviewing a restaurant.

There is a bit of a bias in this document towards command-line installation guides for server software because that’s what I have been working with lately, any suggestions of tips and tricks for other kinds of procedures will be gratefully received and incorporated.

I pay some attention to document structure and formatting here. I’m assuming most documentation these days is destined for the web. If it’s going to be housed on a project site then it probably makes sense to draft and manage it in the project site. For many projects this will mean using a wiki format in GitHub, Google Code Jira, Trac or the like. If there’s interest I could do a Markdown template or similar. Me, I happen to use Word, because it’s a comfortable writing environment and I happen to have a tool for creating clean HTML from Word docs.

This document is intended to be an example of the procedural style it is expounding, even though it’s not strictly a procedural document.


This document is for people writing technical procedures, particularly to do with software, such as installation or user guides. No special technical skills are required apart from the ability to format a document using headings and numbered lists.

The process

Provide background

  1. Make sure that it is clear who the document is for by putting a heading near the top like “Audience” or “Who is this for”. Then, as I repeat below, write for the identified audience all the way through.

  2. As with a recipe, at the top of the document, list all the ingredients and any special tools, software, permissions, licenses or URLs for name authority services as well as where to find them (if you don’t know what that is then you understand that’s why it needs to be up the top of the document).

    Your reader doesn’t want to find out they need a second computer, an account on iTunes, two litres of ethanol, a sorbetière with the frozen bit already chilled, a special proprietary wrench and an 8” floppy disc half way through.

Document the steps

  1. Use imperative statements formatted as numbered lists.

    That is, command-sentences like the one above. Yes this is the way we talk to dogs. Think “Sit”, “Stay” and “Shake”. Yes, it might be a bit rude, but it is unambiguously something the reader is to do.

  2. If context is required, then add it to the beginning of the command, as I did in this sentence, which is a still a command (the verb is “add”).

  3. Write for your identified audience.

    For example a Linux systems administrator can be told to “Edit the file” while if you find yourself explaining this to a non-expert then a lot more detail will be needed about how to get in and out of the editor and how to make the change.

  4. Never, ever, use the passive voice in the actual procedural part of procedures.

    For example:

    The file is now loaded into the database.

    The machine is shut down.

    Both of these are ambiguous because the agency is missing. Who loaded the file? Who shut down or is about to shut down the machine? (This is not general advice for all documents, the passive is useful and appropriate in many genres, just not in the server room).

  5. Remembering to issue commands, in context, use forms like this:

    To shut down the machine:

    1. Push the big blue button for ten seconds.

      The machine will emit an audible sigh and the lights will go out. When the lights are out that means it is shut down.

    2. If the machine did not sigh and shut down, repeat (i).

    3. If the machine is still not shut down after three tries, hit it with the hammer you prepared earlier.

  6. Don’t be relentlessly positive, even if they taught you that in business school. If there’s something that the reader shouldn’t do, warn them explicitly.

Factor the procedure into blocks

Yes, sometimes you need some non-procedural background to your procedure. Put it below a heading like this text. This kind of background is where you can put options such as variations people might like to try, but if it’s a technical procedure you’re writing don’t be tempted to turn it into a training manual about all the configuration options that might be possible or some kind of narrative. Pick one unambiguous pathway and tell the reader how to follow it. Most of us need to learn a few scales before we improvise.

  1. Don’t ask the reader to perform a lot of computational thinking such as adding alternatives such as “If you’re on a Mac”.

  2. If more than a few decisions need to be made along the way, make one or more separate procedures.

    Yes that might be a maintenance problem. Solve that problem yourself and spare the user.

  3. Avoid the use of variables like <$wherever-you-decided-to-install-this> and pick one best-practice way to do things. For example, on a Linux system tell people to install code in /opt/ or in the desktop world tell people to drag a mac application into Applications. If they disagree they can do it differently.

  4. Break procedures up into manageable blocks of, say, no more than seven steps which perform some logical sub task, use a particular user account or take place in a particular part of an application. There a couple of ways to do this:

    1. Start a new list of steps with a paragraph like “To shut down the machine”.

    2. Use an imperative heading. That is, a command.

NOTE: when I was drafting this document the above “Break procedures up into manageable blocks of, say, no more than seven steps” was step 8, that was funny, but it was also wrong, so I re-organized the document. No, seven is not a magic number.

Test it!

When you have finished:

  1. Get someone who fits the profile for the identified audience to test your procedure.

  2. Before they start, get the tester to read these or any other guidelines you have so they can check the procedure for compliance as well as tell you if it works.

  3. Fix any problems and retest.

NOTE: If you’re working on a development project, and the documentation isn’t working then there’s probably a problem with the process, which might mean the software has been built badly or the system isn’t working.

If it’s hard to document it’s likely impossible to use.

Tell someone!

Avoid writing documentation

No, “Avoid writing documentation” does not mean “don’t document things that need to be documented”. It means, try to minimise the need for documentation.

  1. For example, work towards turning multi-step installation procedures into installation packages or scripts.

  2. Go back to the business analysts, programmers, managers or whoever invented the process you’re documenting and work with them to automate or eliminate everything you can.

Tips and tricks

For command line documentation

  1. Format  command-line commands in <pre> blocks for the web or in monospace fonts for print.

  2. Make it easy to copy and paste commands, and not just one at a time:

    1. Do not include the prompt unless you can let the user copy and paste commands without having to edit it out.

    2. Don’t let line-breaks creep into commands in a way that makes them hard to test (this is a problem with PDF).

    3. If you really wanted the prompt because it showed the working directory or username then make that explicit with commands. For example:

      #Become the user this app will run as su dc21cd ~

    4. Use comments, as seen in the previous example as you would in code.

For desktop or web software

  1. Use screenshots where they’re genuinely better than using text, at least the first time you refer to that particular visual interface device.

  2. Even if you have used a screenshot refer to things in text, such as “From the File menu choose Close”, this will help readers to find things via searching in the documents and in search engines.

  3. If there are a lot of configuration items, summarise them in a table in text so they can be searched; even if there are screenshots as well.

  4. Format menu items and using <code> in HTML or a monospace font in print.

Copyright  Peter Sefton, 2012. Licensed under Creative Commons Attribution-Share Alike 2.5 Australia. <http://creativecommons.org/licenses/by-sa/2.5/au/>

*: The only thing I’ve changed from David’s original advice on writing about software is to recommend formatting mentions of menu items using <code> in HTML or a monospace font in print, but as David said, it is sufficient to use the exact capitalisation and punctuation of the menu: “From the File menu choose Save as…” Also we didn’t have to deal with stuff like the Microsoft Word Ribbon in those days, so fewer screenshots were required.