Google Analytics reports different results using the same filters

1
votes

We have created a couple of widgets on a Google Analytics dashboard. All widgets have filters for "Event Action: Fetched", which is a value that we have passed into Google Analytics via Measurement Protocol.

The first widget is a simple metric showing all events found. The second widget is a table displaying all events by Device Category.

We were expecting to see the same amount of events for the two widgets but this was not the case. Surprisingly, the second widget showed a larger amount of events than the first one did.

So, we tried to make another metric widget in which we also filtered for "Device Category: mobile|tablet|desktop". This resulted in a value larger than the total events from the first widget, exactly the same as in the second widget.

See screenshot: Screenshot of Google Analytics

Is this expected behaviour?

We are unable to see why a second filter would show a higher result. Could anyone give a hint as to why this is happening and how it works?

1
google-analytics
Hmmm. I was curious, so I tried recreating your dashboard widgets on a view of mine and couldn't reproduce this issue. For me, the device categories all added up to the same amount as when I wasn't using the device category dimension. – Philip Walton
Do you have the same problem when you filter by other dimensions? – Philip Walton

1 Answers

0
votes

After looking more closely at the screenshot again, I noticed that your dashboard widgets report a total of 53,450,217 events for the selected time period.

Now, I can't be 100% sure, but it seems likely to me that this is a sampling issue. According to Google Analytics:

Sampling occurs automatically when more than 500,000 sessions are collected for a report, allowing Google Analytics to generate reports more quickly for those large data sets

Unless you're sending more than 100 events per session (unlikely), these dashboard reports are probably based on sample data, which could explain why the results are different. Usually sampling is fairly accurate, but since events with an action of "fetched" represent only ~0.24% of your total events, it's not that surprising that using a sample of total sessions could generate such drastically different numbers.

To answer your bigger question: yes, the sum of the subtotals for each device category should add up to the the total reported without the device category dimension.