I am developing a web app where a prominent feature is the ability to adjust the signal/noise ratio of the app's content on the fly. The value needs to be represented along a spectrum which includes minimum, maximum, and default values. The user will have already defined what type of content is considered signal or noise by giving it a 'score'.

After using the app myself for some months, a range slider seems to be most intuitive way to display and adjust this value. As a bonus the same control works equally well with both mouse and touch input.

The idea is that sliding to the left would gradually increase the noise and lessen the signal, changing the visible content; and sliding to the right would gradually increase the signal and lessen the noise.

What are the best min, max, and default values to use?

Technically, the scale for S/N goes from 0 up to infinity, and 1 is the default (source: Wikipedia). But it's not very intuitive for the average person to see or use a scale like this: enter image description here

I'd rather have it be intuitive than technically correct.

Here are some examples I've been toying with:

enter image description here

  1. From -5 to +5, with 0 as the default

    • Pro: By using positive and negative values, the user is clearly rating content as good or bad.
    • Con: It feels odd to be looking at content with a 'score' or 'rating' of zero.
  2. From 0 to 10, with 5 as the default

    • Pro: In many ways it feels more comfortable, as a user, to work with strictly positive integers.
    • Con: Is the 1-10 scale universally understood to the point that when people see '5' they immediately think 'average'?
  3. Non-numeric values, or no values at all

    • Pro: This makes it a bit cleaner, and frees the user from having to understand the numbers.
    • Con: Numbers are good for comparison: "I rated that content a 6, so I should rate this content an 8 because I like it more." With no discrete values, things could feel loose and disorganized.
  • Is this scoring based on crowd rating, or personal rating? Are scores of unrated items estimated (via crowd or a netflix-like algorithm)? May 5, 2012 at 17:32
  • The ratings are personal. Unrated items get the default rating (right in the middle). Rating content is optional, but obviously the more you rate the more you can fine-tune what you see.
    – craigpatik
    May 6, 2012 at 19:14
  • One question... why would people filter content they've already seen and rated? Isn't filtering more effective for new content? Can you describe the use case where users want to look at content they've already reviewed so often that they need to filter it by quality? May 6, 2012 at 20:07
  • Myrddin, it's a Twitter client, so you would tend to see some of the same content repeated (a word or hashtag; a person you follow for sports chat who occasionally delves into politics; etc). After you see it once or twice you can designate it as 'noise' and see less of it from then on. And you could put in phrases of interest to make sure you don't miss them.
    – craigpatik
    May 7, 2012 at 10:51

2 Answers 2


When rating items, 'stars' are the most common and accepted method. Simply displaying a list of 5 faded stars is enough to get many people to realize that they can click the stars to rank the item they're looking at. How you store the stars internally is up to you.

However, when filtering do not filter from 0 stars to 5 stars. Instead, filter from 100% of the content shown to 1% shown.

Say one person rates most of their content in the 2 star to 4 star range, and you do Netflix style estimating of new content with 'fractional' stars (so a movie might be scored at 3.2 stars, even though manual rating can only do integral scores). At full noise, all movies should be displayed. When the slider is dialed down to 80% noise, then 1 in 5 movies should be hidden based on that user's personal scale. This may mean the cutoff was moved, not to 1, but to 2.1 (because the user almost never scores things at a 1).

So calculate the distribution of scores for a particular user, personalized, and have the S/N slider (which is a fine control to use) be a linear scale of the % of items that will be hidden, based on their individual rating distribution.

  • Users understand exactly how to rate with stars; it's better than numbers, and eliminates confusing as well as reducing the explanatory text required for controls.
  • Users don't need to worry about their personal score distribution. The filter control will cleanly and linearly hide content based on their personal rating distribution without their having to worry at all about what the slider represents numerically.

Downside: without estimated or fractional ratings, then the slider will not be a smooth distribution, but 'chunks' of content will vanish at the integral cutoffs, which can be jarring for the user as they drag the slider... nothing... drag it further... nothing... drag it a bit more... and ALL the '3' star items (their most common) fall off. This method requires allowing fractional star ratings, or needs fractional estimates based off of the user's integral choices (Zagat is a good example... users rate 0 to 3, but published ratings from 0 to 30).

  • Very interesting. I'm not sure if I can get stars (in particular) to fit with the signal/noise meme, but I like your idea of a 5-point rating system. Also, good point about grading on a curve (one man's 2-star is another man's 1-star, etc). Perhaps a scale with 10 points is too fine-grained and a 5-point scale is easier, but I will research that UX discussion separately.
    – craigpatik
    May 7, 2012 at 10:54
  • A possible answer would be to allow fractional star ratings... I have not seen it done in practice (which probably means there's something wrong with it!) but maybe a 5 star ranking where you can slide anywhere from 0.1 star up to 4.9 stars? May 7, 2012 at 13:35
  • I think fractions (even half-stars) is over-doing it and detracting from the whole point of the slider itself. I will play with both 5-point and 10-point scales to see which feels right and pick one or the other.
    – craigpatik
    May 7, 2012 at 17:43

Why not just label them signal and noise? Or use the words that describes it. For me as a user any number will be arbitrary. What's the connection between 5,+,10000 with more noise?

So if calling it signal when scoring I would have a slider with "more signal" and "less signal". Although depending on you user finding another word to user rather than signal and noise could be preferable. Maybe just "More scored content" and "Less scored content".

  • Excellent idea. I think I may combine this with Myrddin Emrys' 5-star, sliding scale system.
    – craigpatik
    May 7, 2012 at 10:56

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.