We have a critical situation. We have an external data warehouse where we track all our online ecommerce purchases with purchase ID directly from website and at the same time we track all purchases in adobe analytics as well with same Purchase ID. But when we compare total purchases of adobe analytics with the purchases captured in external data warehouse the adobe numbers are way less which mean many Purchase IDs are missing in adobe. We checked multiple test orders and the tracking is fine. What could be the issue?
How big of a gap are you seeing? In my experience, there is usually a 5-8% gap between Adobe data and other data sources. If the gap is more than that, then there is likely a problem. Unfortunately there are a lot of reasons why the data could not be present. Anything from implementation bugs to metrics/dimensions not being set up correctly could be the cause.
The first thing I would check is for patterns. Are there data gaps in other metrics too? Are the missing purchases happening in a particular time frame? Break down the number of orders by day or by hour, depending on what you're seeing, and compare that to your external data warehouse. This can tell you if maybe you're missing data from a specific time period.
If that isn't the case, I would then suggest looking at different customer types or order type (singed in vs guest, etc based on the options on your site). Being able to pinpoint specific variables that are present in all of the instances of missing data can help you determine what is wrong.
The way I usually tackle this is to eliminate one variable at a time. How did you check the test orders? Did you look at the data in a debugger or did you look at the data in workspace? Is it the total revenue amount that is off or the number of orders? If it is the total revenue, I would check for things like additional fees (eco fees or shipping/handling fees) that might be present in your external data but not in Adobe.
I tested by doing test purchase and used debugger. Also cheeked adobe report and the transaction was seen. The difference is in both no. of order IDs and revenue amount.
Sometimes just doing a single test purchase won't give you a definite result, especially if you're getting the majority of orders but not all of them. Have you checked for any other missing data? What does the number of orders and revenue look like compared to the external data? Are both revenue and order amount off?
Yeah, that actually sounds about right. Between ad-blockers and browsers blocking tracking, there are some customers that you won't get analytics for. You also then need to take into account any filters you have on your VRS (such as excluding bots, internal company traffic, etc that might accidentally be filtering out some revenue/orders). And finally, implementation/tracking bugs, something not firing, a customer moving through pages too quickly for a call to fire, etc. It's possible that the call might not get triggered and sent to analytics. Between all of these issues, having a 5-8% discrepancy is likely expected when comparing to an external source that doesn't rely on analytics tracking. If you start seeing more than 10%, that's usually when I would try to look for a systemic problem.
In your case the CRM you use. Is there any credits, cancellations occurring that could also help cause the gap?
Also do a check if there are orders missing from your CRM side is it a few customers only or many different? Can be some IT dept for them have settings that block some analytics tags from firing correctly or just blocked for their users.
Typically one part of delta is orders that are cancelled or credited(you see them in adobe but not the cancellations or credits).
Let us try to break down this problem.
You mentioned that you have tested out the flow on web, and you see the tracing working fine. And that might not help much here.
This is what I would do before getting into the technical audit of my implementation and ecosystem.
Try to zero in on the Transactions that are not recorded by AA.
I am sure you must have other associated information for the orders missing in AA. See if you have anything that links this missing order to the recorded visit in AA. For instance, any order placed must be through a registered and existing user - you might be tracking encrypted user ID if the same user is visiting web. Try to see if that user had any activity on web when the order was placed in this external data source.
Plus check if there are filter applied (IP Exlcusion, URLs, etc.) at a report suite level or if you are using VRS.
I forgot to mention one thing. We migrated our site to new platform and the page load time improved from 8 sec to 3 sec. Since then we started seeing the gap in orders. Do we need to make some changes in adobe launch to load quickly and capture values?
You would need to perform tests on purchases to ensure you are not dropping any orders.
Did you do a QA of changes and actually test/verify purchases and all related metrics?