Hello, everyone
I have channels and number of messages inside them for the last hour (with 1 minute interval). The chart is meant to show channels activity, so user can quickly understand what channels are active and what doesn’t have a lot of messages.
I’ve thought about multi-line chart, but the issue is - I might have up to 20 channels, so it looks messy
How would you visualize such kind of data?
Its too spiky (high frequency noise). The common way to remove high frequency noise is to apply a smoothing function (e.g. average over a window i.e. moving average) Smoothing - Wikipedia
Personally I tend to prefer bars over lines. The interpolation in line charts often suggests intermediate values that aren’t there (consider e.g. spikes in between intervals that aren’t captured).
I’d then either choose a stacked bar chart or (if the design allows it) facets where each series has its own chart, which imo makes it much easier to spot correlations.
Another benefit of facets is that they give you the space to display different bin sizes simultaneously, e.g. to show high-level trends while also allowing you to surface spikes:
I don’t think that calculated correctly.
A 1h wide window contains 60 one minute samples.
When no messages are posted the moving average is 0
If a message enters the window, 1 sample is 1, and 59 are 0, so the average is 1/60. The swings on your graph look more than that. The window remains at 1/60 for 1 hour. So the moving average goes up a little and stays up for 1h. This is what a moving average does, its replaces every single tall spike with a short wide one, the width being the time base.
I do not see this pattern in your chart, its still high frequency, but definitely less so some smoothing has occurred. More smoothing!
