I am able to compare the transactions in Google Analytics with offline data of the transactions actually placed. There is about a 2.5% discrepancy between the two, where Google Analytics is either:
a) missing some transactions altogether
b) reporting duplicate transactions, or
c) reporting "delayed" transactions
I have researched how to address these issues to no avail. Server-side code has been implemented to correct duplicate transactions (by not resending any orders already placed), but are still seeing some duplicates.
"Delayed" transactions are those that were placed days or weeks ago, but does not show up in the GA transactions report for what seems a random amount of time. For example, one order was placed on 2/3, but did not show up in GA until 2/10. I have been unable to replicate this issue, and not really sure what has caused it.
Is these common problems? Is a 2.5% discrepancy (we are using the free version of GA) somewhat expected, or considered typical? Any resolutions I could try?
Arnold Majlath, Google Analytics Top Contributor
Was my response helpful? If yes, please accept it as solution.
Circle Me On Google+ | Visit me @ Redfly Digital
Just to add to Arnold's great answer - I've worked with a lot of properties with ecommerce implemented and agree that a 2.5% discrepancy is absolutely fine. You need to look at the discrepancy trends over time - if it spikes, rather than being more evenly distributed, then this would point more towards a tracking issue than just the issues that Arnold describes. The fact you're seeing still duplicates though, needs to be fixed. Delayed transactions are also odd - GA cannot magically cause hits to appear, something has to trigger the code. I'd suggest you focus your troubleshooting on thank-you pages and whether a user a can revisit these and fire the transaction code. I've seen similar issues when users returned to the site to check the status of an order when prompted to do so by an email, for example.
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.