Forum Discussion
Data from Google Analytics Doesn't Match!
I've got response from the product team.
This is expected behavior. The API we're using for Google Analytics is the one available to free users, and that API is known to return less accurate values for high volume sites due to the way it samples data. Unfortunately, the API for Google Analytics Premium customers is significantly different than that for the free users, so we'd need to implement a new connector specifically for Premium in order to get the more accurate values. This is not currently on our roadmap. Customers who would like to see this feature should request it or upvote it (if already there) at ideas.powerbi.com.
Best Regards,
Herbert
Thank you for the input. A question if I may...
I am working on a client with an average of 100 K monthly visitors (1.2 million a year). So if I wanted to include their GA data in a report, with historical trends etc, would that cause data discrepancies?
I understand the implications of aggregate data and the risks involved in using it in reports. But can't we simply replicate data (by avoiding re-aggregating in sums etc in PBI) as seen on GA? It defeats the prupose of PBI in terms of slicing etc..., but on the other hand allows to centralise reporting from different sources in PBI to make PBI the single reporting tool.
I guess what I find a bit disconcerting is the original poster's comment that the data as seen in the query editor in PBI si the same as seen on GA, but discrepancies occur in reports...
(BTW, I have looked at your product and it does sound very impressive, albeit I am not sure if this particular client wants to spend on extra tools when he can access data on GA for free).
Please excuse my obvious ignorance regaring GA, but I haven't needed to use GA data in PBI before. (btw, does the same problem happen with data extracted from Facebook, Instagram etc?)
Thanks for your help,
Paul.
EDIT: as an afterthought, does Google Data Studio show the same discrepancies as we might find in PBI?
DataStudio has GA natively built-in.
All queries are sent to GA backend and evaluated.
There is no local aggregation. But sampling still kicks in.
Yes, you'll get historical discrepancies.
When you cross the limit of 500K sessions, which will happen hitting the 5 months mark using your data set, the whole data set is be sampled.
What can you do?
You can make a M script that takes one day of data at the time. Build a joined table on e.g. 365 or have many days you need.
But you would still be limited by the number of filter dimensions and reporting dimensions. You would easily have multiple of these scripts pulling 1 day at a time. Which in turn will be really hard to maintain.
I see more and more running into this problem. All very confused by the amount of time they need to invest in the "free" GA data source. This is just one data source out of many of a reporting suite.
I agree with you - Checking data in GA is not fine - when you already have realized how hard it is to adopt employees/users of any given BI repoting platform. The platform should be where the employees build a better company.
I understand price matters.
Are you talking about the facebook and instagram connectors found on www.scitylana.com/connectors?
They have same issues.
Thank you for the explanation and warning regarding my particular scenario.
When I mentioned other conectors, I was referring to the connectors within Power BI. I take it from your comment that we run into the same issues using the inbuilt PBI connectors then.
Do the connectors on www.scitylana.com/connectors solve the issues?
Which kind of makes the PBI connectors pretty much pointless for medium to large traffic websites/social media sites. Unless I guess we keep the reporting to the very basic and avoiding aggregations? Or does sampling also have a big impact? I'm not sure what the impact sampling actually has (apologies again, I'm new to sources such as GA, Facebook, Instagram etc...but I will be facing these issues in the next two weeks)
It seems therefore that Power BI is not the tool to monitor web activity, unless you are prepared to fork out a relevant amount of €€€, which I guess is out of the question for most medium and small businesses (with free alternatives out there).
BTW, how should I contact you or whoever to understand the licensing for Scitylana? I have checked the page and I have a couple of doubts.
Thanks.
The connectors have their week spots.
This is nothing new, really. Classic ETL into a DW is still preferable in my opinion.
Power BI + connectors is entry level BI. But it's really great for entry level BI setups
Web activity just pushes it to the limit faster that the other data sources. But these kind of web activity data sources gets more and more normal. Power BI raises peoples expectations faster than other products out there.
MS has a whole suite of products you can buy when you grow out of the Pro licensing.
My free connectors has the same limits.
But Scitylana's Google Analytics solution is solid with built-in elasticity for growth.
As opposed to the connectors and PBI Import mode - DirectQuery in Power BI is much more relevant than one would initially think.
I think a combination of import and direct will be the future. To balance cost and performance.
That's good for the BI Analysts consultants. New types of complexity sees the light. :)
Sampling is mostly an issue when you want to
- integrate data sources (joining keys will be missing)
- calculate conversion rates (conversion rate of 5% could in reality be 7.5% or 2.5%)
- drill into user behavior
- rely on the real values
Write on our chat - or send mail to support at scitylana.com