I develop an open-source, system monitoring tool, ServiceMon, and I'd really appreciate some help choosing the right words to use for a feature.

The state of the system that is being monitored is represented using a named severity level. The level changes when the number of errors, recorded within a given time period, breaches a particular threshold.

ServiceMon screenshots showing deteriorating error state
(source: rightcalc.com)

I'd like to give each level a name, which will be used when configuring custom colours and threshold values. I'd ideally like to use meaningful adjectives - rather than cold, generic names like ErrorLevel1, ErrorLevel2, etc.

This is what I have so far:

Default Colour | Number of errors | Adjective
Green          | 0                | Operational
Yellow         | 1                | Temperamental
Orange         | 2-9              | Unstable
Light Red      | 10-99            | Critical
Dark Red       | 100 or more      | Dire

Do the words I've chosen seem appropriate? Can you think of any better alternatives?

Updated to elaborate on the primary use case

I envisage ServiceMon being primarily used as an "information radiator" - a permanent presence in a room that provides immediate visual notification of a system's state. Each ServiceMon instance is designed to be lightweight so that many can be tiled on one screen, with each one tracking a different system, or different aspect of the same system.  For example, where a system is being monitored for availability and responsiveness, I would envisage a separate instance running for each aspect, each with its own specific thresholds.

  • 1
    What will users be doing with this information? Are they looking to make long-term observations about the states of services, or are they looking for errors which require their attention? Will users have English as a first language and will 'critical level' always be tied to the number of errors, or will weighted errors always create an 'urgently broken' status (if I was writing webtests, I could easily imagine a situation where one or two specific failures could be more important than twenty or thirty performance warnings). Sep 16, 2012 at 18:46
  • 1
    Great questions, I've edited the answer to provide more additional information. You make a good point regarding localisation of names and this, together with the excellent answer below, is pushing me towards using the colour names alone. Thanks
    – Wheelie
    Sep 16, 2012 at 21:27

2 Answers 2


My first thought is that most people will abandon the secondary naming scheme you assign each colour, and will probably call each state by its colour.

The first real-world example of this behaviour that comes to mind is the American Department of Homeland Security's (discontinued) colour-coded threat level warning system.

Thread Levels

You will note that, even in the linked article in Wired, the threat levels are called by their colours, with the names added almost as a parenthetic afterthought:

Today’s threat advisory color is “Yellow,” or “Elevated.” For domestic and international flights, the level is “Orange,” or “High.” (Duct tape your windows at your own discretion.)

(From an article on Wired.com)

My gut hears more sysadmins looking at their monitor and calling their supervisor saying "Yeah, we're in the red on example.com" rather than "We have a dire-level warning on example.com".

No street racer anywhere says that "engine speed is invading the critical band"; they say they're redlining.

  • 2
    Fantastic answer. Thank you very much for your help!
    – Wheelie
    Sep 16, 2012 at 21:28

Colors alone are not enough. For example the image included by msanford in his excellent answer would be meaningless if printed in a monochrome laser printer.
The meaning of colors might be affected by culture. I recall having read that in Japan mourning people dress in white. In China the red color is so common in decoration that a red alert might rest unseen.
The sense of growing levels can be conveyed by numbers and size. Numbers combined with other elements.
ErrorLevel1, ErrorLevel2 are not good because they lack identity: almost the same text with only one different character at the end.
Colors alone do not convey much information for people who are colorblind, like for example me and a significant portion of the male population.
IMO the strongest effect would be a scale made of numbers, names, colors and sizes. Like for example:

  1. low: small on green background
  2. guarded: big on blue background
  3. elevated: bigger on yellow background
  4. high: huge on orange background
  5. severe: humongous on red background.

This way the non English speaking user can rely on the numbers without having to fully understand the words, which might be quite difficult to accurately translate to every language.
This way the words choice will not be that critical.
Like in the Homeland Security levels, whether you assigned a name or not, you can always add a detailed description for the users to read if they are unsure of the meaning of the error level.
IMO people would end up calling the levels by their numbers. This is so for earthquakes and it works. As of the Homeland scale, people name the levels by their colors.
I suggest that you do some user testing with the Homeland scale, to check if most people remember the colors but not the names. If so, then stop worrying too much about the names.

  • Great points also. I like the idea of incorporating secondary and ternary status indicators to backup the colour. Thank you!
    – Wheelie
    Sep 17, 2012 at 19:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.