If you look at Google Analytics sessions to check for high level metrics such as sessions day over day, then you are probably fine and you don’t really experience any issues. The things start to get a little complicated and frustrating when you try to get more complex data involving multiple dimensions and metrics over a longer period of time.
This article is based on working experience with several big GA properties (the old setup, not the new GA4s which are currently in beta). If you run a small site with a couple hundred visits per day, you may have never experienced sampling for example. If you run GA in a big company with lots of hits and events, all these issue are probably your daily bread even though you pay significant GA fees to Google.
There are 2 goals of this article:
- Summarize all the known GA issues – I am frequently googling for GA issues and I am quite surprised how few people write about the below mentioned problems. Also, I have never seen a nice “GA issues” list so I’ve decided to summarize all my important findings.
- Just to share my frustrations. You can take it as a useless rant. 🙂
I want to make one thing very clear at the beginning – if these issues were happening just in free Google Analytics then fine… It’s a free product so I expect nothing. However, all these issues are happening in paid GA 360 properties.
1) Sampling Everywhere
Sampling is a big issue in big GA properties. Sampled data is useless 90% of the time so the only choice you have is to wait for an Unsampled Report which may never come! More on that later.
I’ve seen a great deal of sampling happening in properties which cost thousands of dollars in GA fees just to run every month. And yet, you still see sampling.
Sampling is present in default reports, in custom reports, in attribution reports. It’s just everywhere in GA.
It’s enough for you a to pick a few dimensions, set up a slightly more complicated filter in custom report and you can bet the data will be sampled. Or you can just select “last 8 days” instead of the preset “last 7 days” and again…there is a big chance your data will be sampled.
There is not much you can do about it. Make your reports less and complex – the favorite solution of Google support.
2) Limited Date Filters
So we just turned into 2021, yet the default date filters in GA look like from 1985:
So we have: Today, Yesterday, Last week, Last month, Last 7 days, Last 30 days.
What about This week, This month, This Quarter, This year, Last 14 days, Last 60 days, Last 90 days…? I guess that would be too difficult to add…
3) No Pivoting Options in the UI
If you have ever worked with Adobe (aka Omniture or aka SiteCatalyst) products, you know you can basically create pivot tables directly in the Web UI. One can tell that Adobe is an expensive product but so can be GA360.
We want to create unsampled pivot tables in GA web UI!
4) Unreliable Unsampled Reports – The Report Never Comes
I especially love this one. Imagine this situation:
You prepare a custom report in GA UI and “by coincidence” (irony here), the data is sampled. So what do you do? You request an Unsampled Report, right? Well, that still does not mean anything. Unsampled Reports can fail and never come. And as a “cherry on top”, there is no notification going out when the Unsampled Report fails. So you just keep starring into your mailbox waiting for the report like an idiot.
How to tell whether an Unsampled Report has failed or not? Go to Unsampled Reports > Completed and look for reports with “Failed” status.
You will see that you should “Try creating multiple smaller requests” – that’s the universal answer to all issues with Unsampled Reports. You see it in the UI and you’ll hear it from GA support. Don’t waste your time talking to GA support. They won’t tell you anything helpful when it comes to Unsampled Reports. It’s just a lottery.
Generally speaking, it’s quite safe to pull the unsampled data by month. However, it quite sucks when you have to pull e. g. last 2 years of data = 24 reports.
5) Unreliable Unsampled Reports – Randomly Missing Rows
So when I am lucky and the Unsampled Report does not fail and finally comes, I still have not won. I’ve seen reports where part of the data was simply missing. I have not found a cause of the issue. I don’t know how to get around it. I did not waste my time talking to GA support about this. I just re-request the report and the next one is usually OK.
6) Only 1,000,000 Cells in Unsampled Report Exports
When I am lucky again, and the Unsampled Reports comes, it’s always cut off at 1,000,000 rows. So if the final outcome was supposed to be 1,200,000 rows, you’ll still get only 1,000,000 rows in the exported file.
And of course, you will NOT be notified anyhow. So make sure you check your CSVs. If the file stops at the 1Mth row, you are missing some data.
7) Poor Scheduling Options for Unsampled Reports
So say you are aware of all the issues with Unsampled Reports and you still want to try your luck and schedule a daily Unsampled Report. These are the options you’ll see:
You cannot really tell when the report comes tomorrow. 8 am or 8 pm or 6:32 am? Only God, errr Google knows. It just comes when it wants to (of course, if the report does not fail).
Another slight annoyance is the fact that it will keep coming only for the next 6 months.
8) Only 5 Dimensions for Custom Reports in the UI
When you are setting up your custom reports (which you can be later requested as Unsampled Reports) you will sooner or later hit another major limitation – you can only select 5 dimension in the GA UI. Absolutely subpar. I frequently need to select more than that. 10 dimensions as the limit would do it for me.
You can partially get around the 5 dimension limit by using Supermetrics for Google Sheets which allow you to select up to 7 dimensions. Basically, any API based query will let you use 7 dimensions instead of just 5.
9) Multi-Channel Funnel Reports Cannot Be Scheduled via Email
Only a few people probably work with these attribution reports:
The sad fact is that you cannot even schedule these reports so they come say every Monday to your email. You have to generate them in the UI and wait.
10) Direct Channel in Default Channel Grouping
I see this every day – if you are using Default Channel Grouping in your organization, many people think “Direct” channel is people typing your URLs into their browsers or accessing your site via bookmarks. The sad truth is that Direct is basically trash which cannot be identified. It’s the channel of the last resort in GA. If all other channels do not match, the traffic will fall under Direct.
You should probably define something like “True Direct” which would be only people landing via Direct channel to your homepage (or a few main URLs with simple URL structure). It’s foolish to believe that Direct sessions to someurl.com/level1/level2/level3/url would be typed by human beings. A direct visit to this kind of URL could be just a bot or a visitor from previously expired session.
If I had the power to change this naming it would really call Direct as “Trash” or “Other”.
Conclusion
While I expect zero positive changes future in the old GA properties, we can hope the new GA4s will take care some of these issues. And let me restate again – if these issues were happening in free GA properties then there is not much to complain about a free product. The problem is it’s also happening in the paid GA360 setups.
There are no easy solutions to the mentioned issues. You can make your reports smaller or you can start exporting GA raw data to BigQuery but then you need to know great deal of SQL to be able to mirror the data what you see in GA UI. And BigQuery is not free either. It can get really costly to store your GA data in BigQuery in the long run.
I hope that I have described some of the GA issues you are coping with – you are certainly not the only one. Feel free to add your issues into the comments below!