Visualize count of unique identities over time

[Matjo]

Hi, each document in my index has an author id and a date. I would like to create a bar chart in kibana where I for each month can see how many new authors have contributed. I imagine it would look something like this:

I tried using a histogram with unique counts but then I get the unique count for that particular bucket. I want the unique count in the bucket that has not appeared earlier (by date).

Is this possible and how would I go about it?

Best regards,
Mattias

[LeeDr (Lee Drengenberg)]

I've seen this question before and unfortunately I don't think there's a good answer.

One possible way, although it looks quite ugly, is to do your Unique Count metric, and then to Split Chart instead of Split rows. I think it's doing what you want but looks terrible;

You could click the little arrow under the visualization to open the "spy panel" and look at the results to see if it looks like it's doing the right thing;

Here's another long post where we tried to get the same thing (and failed);

Cumulative Sum regardless of time interval

Hi Chris I feel it's up to how people are defining cusum. For me it could be a cumulative number based on the whole time series. It equals to sum an auto date histogram of a certain metric from begging of the specified timefield until now and when date range filter is applied, it's just slice a piece of data from timeline instead of re-calculate it based on the date range filter. There might be needs for both cases, I just hope if kibana team could add a variable in cusum function to make both …

Regards,
Lee

[Matjo]

Ok, thanks for your answer. In my randomized quest for better solutions I might have stumbled upon something.

I did the following:

Y-Axis
Min bucket aggregation.
Bucket: Date Histogram
Field: Date
Inteval: Custom 5y

Metric:
Aggregation: Unique Count
Field: Author Id

X-Axis
Aggregation: Date Histogram
Field: Date
Inteval: Monthly

The reason I tried this is that it seemed similar to the a query I knew would solve the problem. If I could group the documents by Author Id, select the min date and plot a date histogram I would have the solution.

I do not know exactly what I have done in my visualization but the results seems very similar to what I want. I have not yet verified it.

What are your thoughts on this?

Best regards,
Mattias

[Matjo]

Pretty sure it was only due to randomness and wishful thinking that I got the result I was looking for.