Regarding the transformations I mentioned, everything worked as expected. However, figuring out how it worked took a long time. The issue is that RudderStack's documentation does not explain the transformation flow well, making it hard to understand where the transformation was applied because of different sources. For example, Node.js had one format, and the frontend part with JavaScript had another format, and I did not know how to work with transformations between these nodes due to lacking documentation. I had to conduct many tests over months to gradually figure out how the transformation worked. I experienced another problem with name restrictions that were not documented. I remember using "name" in a transformation to transform from a generic name such as full_name to "name", but sending it to CleverTap did not work without providing any error. It merely did not arrive. It was not until I switched the variable name from "name" to "short_name" that it worked. It took two days to realize this issue was not in my transformation or the JSON I built but rather a name limitation not mentioned in the documentation. Essentially, the main problem is the documentation, which I find terrible. When it does exist, it is not very helpful. One problem I have seen in RudderStack's documentation is that because it covers many tools and different destinations, there is a lot that is poorly documented and does not explain how it works, which I believe is due to many different destinations and edge cases that are not mapped in the documentation. For example, in Meta Ads, there are usage permissions issues when you create a Meta app to use the API, leading to various permission conflicts regarding number verification that require an API call via Postman, among other potential problems. RudderStack can improve mainly in its documentation, which is lacking. It does not adequately explain how to use the platform, what edge cases exist, and issues such as forbidden names are not explained either. The documentation around Transformations needs to be better as well, to clarify RudderStack's data flow and how it treats incoming data from sources such as JavaScript, which would assist in understanding the expected JSON for each type of information. A potential problem beyond documentation relates to a limitation on API calls for external enrichment. Specifically, I have had a two-second lifetime limitation for the transformation to make an API call, wait for the response, and insert it into the flow. Two seconds can be a very short time.
