KQL Query Date filter on non timeseries data

[Mark_Rodman]

Hi,
I'm tying to use Lens to plot data changing over a period of time. We generate the data once a week at single point in time, and so it doesn't look good as a time series. We also want to plot the count of a data point known as version which can have around 15 different values. I'm therefore plotting count of version as a series (x axis) within a Bar vertical.

I have my initial x series count plot in Lens using the following formula;

count(kql='general.lastContactTime >= "now-30d" and date_iso >= "now-7d"'

date_iso defines when the data capture occurs,
lastContactTime is used to filter the data and drop stale records.
The above is also configured with a Last 30 days date range.

I'm trying to get a second series on the x axis which picks out the previous weeks data as a comparison.

Something along the lines of

count(kql='general.lastContactTime >= "now-30d" and date_iso >= "(now-7d) -14d"'

But that doesn't appear to be a legal query.
Any chance of some guidance on how to structure the second series?
Thanks

[Marco_Liberati]

Hi @Mark_Rodman

have you tried already the shift option?
It would be something like: count(kql='...', shift="2w")

[Mark_Rodman]

I believe shift requires a time series dataset and also can't be applied to differing series.

Thanks for the suggestion though.

[Mark_Rodman]

I'm still struggling with this, and I hate to be cheeky, but I don't suppose anyone else has some suggestions.
Thanks

[Marco_Liberati]

The shift feature requires the usage of at least one time field in the configuration, not necessarily a dataView with a time field configured.
You could define a Break down dimension with a date histogram based on your time field, collapsed by Sum to hide the actual break down. That should make the shift feature work in the formula.