In our org we want to measure the "handle" and "answer" metric values of message interactions for all agents through APIs, but I can't seem to find those metrics in the endpoints I've searched.
I've checked in these endpoints which seem like the logical options:
api/v2/analytics/conversations/details/query /api/v2/analytics/conversations/aggregates/query
And in the available list of metrics I can't seem to find those 2.
Is this by design or I'm searching in the wrong place?
Those aren't metric keys. You're probably looking for tHandle and tAnswered. They are conversation aggregate metrics (per https://developer.genesys.cloud/analyticsdatamanagement/analytics/metrics#conversation-aggregate-metrics), so they can be requested using the conversation aggregate query. Keep in mind that if there is no data for a given metric, it won't be present in the response; only metrics for which data exists are returned.
n: counts — A count, such as the number of interactions over SLA (nOverSLA).
You'll want to look at metrics starting with n for count-based metrics. You might consider nConnected. The context of this statistic is how you've filtered your query. So if you run the same query as a detail query (instead of aggregate), you'll see the list of conversations that are being counted so you can validate that your filter is working as you expect.
Can you clarify why getting a count of connected conversations for a query filtered to your liking doesn't give you a count of how many conversations were handled/answered?
Because our client wants to differentiate between all the interactions that have been "answered" and "handled" by all agents.
They calculate two different internal metrics thanks to those values with the times the agents spent interacting in the platform and the number of interactions they had in a given period
Sorry, I'm failing to grasp how that's different from a count of connected conversations. "interacting in the platform" and "interactions" aren't Genesys Cloud terminology. It might be helpful for my understanding if you can use Genesys Cloud-specific terms to describe the specific actions you're attempting to measure.
Based in what you said "connected" can work instead "answered" correct?
But would it work too for "handled"?
Because if I'm correct, "handle" measures all the interactions that were managed by the agent from start to finish. While "answer" only measures if they interact with it initally.
This is because our client wants to know which interactions were closed by the agents vs just simply replied. Because in our client's business there are many transfers.
Like I said, they use the "handle" and "answer" to calculate other metrics based in the differences I mentioned to you
It sounds like maybe you're looking for wrap up codes? This is where using Genesys Cloud terminology is very important. In Genesys Cloud, "connected", "handled", and "answered" would both mean that a participant had a segment in a connected state, and nothing more. This concept of what type of task the agent performed while interacting with the conversation is a business process of the customer, not an inherent trait of Genesys Cloud. Wrap up codes are a way for an agent to express how their interaction with the conversation concluded within the context of the customer's business processes.
If the agent assigns a wrap up code, you can use that in your filter. So if you have a wrap up code named "handled", use a filter for that. The context of nConnected in the response is the count of conversations that have that wrap up code on any participant. Repeat the query but for the "answered". If you want that count per user, use group by user id.
If you're looking for more than a total count of conversations that had that wrap up code at any point, such as "conversations that were 'handled' but never 'answered'" (using your terminology as wrap up codes), you'll need to perform a detail query and then post-process the results using your own custom logic. Analytics filters only perform positive matching, so you can't do a "this but not that" type of filter.
edit: Also consider if you can do a little math with the stats from multiple queries. Maybe subtracting the "answered" conversations from the "handled" conversations would give you a count of "conversations that were handled and not answered".