Expand my Community achievements bar.

Workspace: Set Graph/Chart Scale Max & Min

Avatar

Level 6

1/26/18

Right now in Workspace, the range of the scale for the line & bar charts is set by the tool automatically and is based on the data set being presented.  While most of the time this is fine however when it comes to supporting metrics that have high occurrences of major outliers (like page load time) or high amount of volatility , the scale can get 'messed up' if a single outlier data point comes in. 

it would be preferred if we were able to 'overwrite' the automatic scale to whatever range we as business analyst feel would be best for our purposes.  In the event that we set a scale and outlier data point comes in the 'normal trend' doesn't get obscured for by a single point.  this applies a lot to the minute and hour charts as daily averages seems to be more normalized than micro time period data points.

this would help my organization a lot because we regularly do nightly code releases that can momentarily impact the performance of the user experience

7 Comments

Avatar

Employee

2/8/18

Great suggestion.  One follow-up question - do you find yourself wanting to just exclude spectific outliers, rather than adjusting the axis?

Avatar

Level 6

2/8/18

Outlier suppression is the main motivation and is what sparked my idea

But we need this to also work if we turn off ‘anomaly detection’, so how would we detect outliers if the anomaly detection feature is not visible on the graph?

Hope that makes sense

Avatar

Employee

2/8/18

Thanks for the extra detail - I might be misunderstanding, but if you want anomaly detection to detect anomalies, you will need to leave that turned on.  A few months ago we added a setting (that is default=off) that will auto-rescale the axis based on the confidence bands of the anomaly detection algorithm.  So by default, it won't expand the y-axis to include the confidence bands, but it will still try to include the outliers (actual anomaly points) - which is what we want to surface.  What I was originally thinking is that you could deselect cells in the corresponding table that you didn't want to show in the visualization, and then the y-axis would adjust accordingly.  But using that in conjunction with anomaly detection may be difficult.

Avatar

Level 6

2/8/18

Brandon

I think the confusion came by your original email asking if I wanted to ‘exclude outliers’. In my mind, it made me wonder - how would Adobe define an outlier?

My original ideas was to allow me to set my own range for the y-axis, implying that I as analyst wanted to have more control of my graph than Adobe’s AI could provide automatically.

My concern about your comment below regarding ‘deselect cells’, is many times I don’t select cells at all I just graph the entire data table. So moving this direction would a) require me to graph selected cells so b) I could deselect cells to adjust my y-axis scale.

To me it seems more user friendly if like ‘back ground colorization’ we enable a ‘set y-axis range’ option where the ‘automatic’ is selected by default but if unchecked would allow me as an analyst to set a min and max.

And to go a step further, allow me to select my own scale markers (quartile, deciles, single units, etc). Meaning on the y-axis scale right now most of the time I only see 1 number (50% for example) in the middle of the y-axis, but what if I wanted 4 y-axis numbers to show on the left side of my graph (25%, 50%, 75% & 100%). The labeling of the y-axis options would also be very helpful.

That’s a different idea sure, but that could also be controlled in the same UI experience where I can select my own y-axis range.

Hope that makes sense

Jeremy

Avatar

Employee

2/8/18

Good stuff - yeah, I just wanted to make sure I was understanding your use case.  The deselecting cells idea was hopefully just an interim solution that could help you address your problem immediately.  But there may be a better solution around directly interacting with the chart and removing specific points, etc.  We have discussed this in the past.  I also agree it would be helpful to have more control over the axes, and something we are definitely considering (along with other visualization configuration settings).