[musingsole]

Metrics with hard numbers are always too precise to be useful. Even the well-meaning will micromanage based off the difference of a 5 or 6 on a scale of 50.

I find the key is metric obfuscation. Instead of "1023 likes", color code the post. A deep crimson for the most downvoted and a sky blue for the most upvoted on a site. The specific value is so useless anyway.

[jobigoud]

How do you handle the high range of values? In one part of the site 1000 likes may mean a very high number while in other, a high profile could regularly see millions.

[musingsole]

Good point, but the overall system could be adjusted any number of ways. The point is to communicate to the user, not throw numbers at them and let them decide what it means.

[ricklamers]

I actually particularly dislike this, transparent reporting of metrics gives users the ability to have their own interpretation. Instead of guessing why platform Y gave post Z a deep crimson color, which you'd probably not be able to do because of 'algorithms'.

[musingsole]

Most people using metrics aren't great at interpreting them on their own in the first place. Look at Chile's notorious misuse of waiter ranking (you fill out a survey after your meal). If a server gets less than 5 stars, their job is on the line. Nevermind the fact that each customer will use the 5 star system differently to communicate different feelings. Some people will never give a 5 star rating.

If it involves people other than you, your metrics likely don't mean what you think they mean.

[Deimorz]

Which significant sites use that method?