We are working with the Platform API to retrieve conversation details using the endpoint /api/v2/analytics/conversations/details/jobs. Our goal is to determine whether a conversation (chat, email, voice, or callback) has been abandoned, offered, or answered based on the API results.
We run the query with only a time filter for the previous day, and the results are stored in a SQL database. We aim to add three new columns in the table to represent these conversation statuses as booleans (1 or 0).
However, we have not been able to find consistent variables or patterns across all conversation types that help us clearly distinguish between these statuses. We've tried multiple approaches but haven't found a reliable method. Has anyone else been able to find a reliable way to do this?
There is no consistency because those events mean different things in different media types, and some of them are mutually exclusive and some aren't.
A hypothetical conversation could contain two voice legs (inbound and outbound) and a callback, all with different metrics there depending on how you're setup and what happened in the life of the call. If it was for example answered on the inbound voice leg, needed follow up, had a callback scheduled, and then was contacted again later on the outbound leg there are two different Answer records for that conversation, and two different offers. Those metrics also don't occur on the same participants. A customer can abandon, an agent has to Answer and an ACD gets the offer in most cases. A call can get offerred to an ACD, Answered by an agent, transferred to another ACD where the customer gets fedup and abandons or gets answered again. Either way that one call had two offers and one or more answers plus may have also abandoned. And that's all just on voice. Messaging is asynchronous so likely even more confusing if it even has those concepts at all, and chat and email are probably on par with voice in complexity except of those three only chats would have any sort of true abandon concept.