• Home
  • Ideas
  • Insert zeros when usage is missing on a weekly trending report

Insert zeros when usage is missing on a weekly trending report

Related products: None
I wasn't able to find a similar community post (apologies if this exists already). We've recently come across this issue with Cisco and a few other customers. When there is no data for a time period, that date period is removed from the report and it's connected to the next available date. We'd like the ability to display the missing dates as Zeros to provide an accurate view into the usage. This creates a more consistent experience for the end users. I've provided 2 customer scenarios below.





Cisco Use Case:


Trend view on R360 only displays months where data is processed, creating jumps in the timeline and it's not clear to CuSMs when there are months where there were not any scores for that particular month.











Customer #2 Use Case:


Customer2 is using our Segment connector - when there is no usage in a given week, the customer doesn't display any data (common use case). Given the volume of data they are sending us for various events, we'd want to avoid the workaround of inserting zeros for every data point that is missing for that week. Here is an example of a report where data is missing for a few weeks.











Does anyone else have use cases to share? Are there any workarounds that you've used that don't require inserting zero? 





Are there any plans to address this with product?
In GonG, we run a Rule to insert 0s every week before the Rules run that push page views to the Usage Data object
Thanks Will! We'd like to avoid inserting zeros as a workaround and are looking for a true product enhancement for MDA reports.
Where would you see this determination being most useful - should it be part of the metadata for a measure (i.e. treat no data as zero) or reporting specific (i.e. show no data time periods)?  Either way, reporting will need to respect but wondering if there are other places a meta data approach would be valuable.
Having this as an option in the metadata would certainly solve this use case, as well as a common problem in the rules engine with null data for MDA rules. I'd say that'd be more useful than just solving for the reporting gaps alone.
Kendra - Can you explain this a bit further "as well as a common problem in the rules engine with null data for MDA rules" . This can be solved in bionic rules or with Absence of data in Custom rules right?
On the "treat no data as zeroes" - the basics here are that if say a specific usage metric has no activity in a week, no data is sent to Gainsight for that specific metric.  What no data would mean for some metrics and the desired approach would then be to treat as zero.  Certainly in reporting this becomes an issue as we look at some trend line - dates are missing from the view.  Not sure what Kendra is referring to related to the "null data for MDA rules" piece.
Ah - we can do 'null data for MDA data' with Bionic rules - fantastic! Apologies I hadn't explored that prior to making the comment.





So in this case then with the initial post, we're really looking to see trends in reports where no data exists for a given week.
Chiming in on the "pivot by date" request - I wouldn't say the request is "treat no data as 0" - it's more a case of allowing me to pivot for 12 weeks, for instance, or 12 months, and see every month as a marker on the X axis whether or not there were rows with that date in the table.





If the purpose of creating a graph is for fast visual consumption of a data set, and we simply drop out the weeks on the X-axis where there was no usage, we're creating a misleading visual. Unless I look closely at a 24-month graph that had no usage during 2 months, I'll probably assume that usage has been steady for 2 years.





However we do it, we want to get away from writing 0's to fields with rules engine -- ultimately, there IS a difference between 0 and NULL -- and when looking at tables, we want to preserve that.  When graphing over time, though, we should treat it like any other incremented X-axis in graphing - retain each increment to show what happened during that month, or week, or day.
Agree Scott - we want to get away from writing zeroes.
Scott - In that scenario we need to tell the system there are 12 data points for which we need to use rules and write zero in the null scenarios directly right. Bionic rules and absence of data will give you an option to do that easily but still it has to be handled explicitly right? Is there any Drost methodology for doing this?
This goes back to our consideration of some kind of time dimension or a way to address time specifically in visualizations.  This would also be something we would need in considering time series calculations (no data = zero).  Time is a known thing (whether data exists for each time period or not) so it does seem possible to "tell the system that there are x number of data points".  Writing zeroes creates a lot of extra data and is a lot of extra work.  
Catching back up on this since this is one of the open Cisco requests that will be discussed Wed-Thurs (July 12-13) - I'm totally on board with Denise's comment above - treating all periods in the range as an x-axis point.





Guessing that this has fallen off the radar a little since we haven't been pushing it - is this something that's in our backlog? Something before end of calendar year?
Hey all - I'm reading through this and would love to understand a bit more on the thought process behind not using 0 as a proxy for null on a given x-axis (date) point. 





From a purely geometric standpoint, some value must be assigned if we create an x-axis point for date, correct? The only geometric equivalent for absence of data (null) is zero. Or is there something that I'm missing here? 





Or is it just that you want to know the difference between these two scenarios (using Kendra's original use case of a Segment connection):


* Segment sent no data for the specified date


* Segment sent data for the specified date and the data it sent was zero


?
Per Venky: we are evaluating this for aggregate report with day, week, month summarization. Not intended for a flat report where already aggregated numbers are brought into GS. Med sized effort as it needs to be supported across databases for all summarization options (day, week month etc) and performance also need to be considered. Do we replace missing data as 0, maintain trend, or add gap?
Terms & ConditionsCookie settings