What is the difference between dropped chats and missed chats?

Return to top

13 Comments

  • Igor Czajkowski

    There is something still unclear in dropped chats metric. At the mentioned "Metrics..." article is said that dropped chats are "accepted by or assigned to the agent where the visitor responds to an agent message, then leaves the chat". Does it mean if at the end of the chat the customer writes "thank you, goodbye" and closes the chat or the page with the chat - it will be counted as dropped chat? If yes, are there any good practices to influence statistics of dropped chat by such cases?

    2
  • Erik Lynch
    Zendesk Customer Care

    Hi Igor,

    You are correct. If the chat ends with the visitor entering a reply, then quickly dropping off before the agent can respond again, that is being counted as a dropped chat. This is a tricky thing to filter out. The data stored from the chat doesn't have a smooth way of differentiating between a chat that had gone on for a while and ended with "goodbye" as compared to a chat that had gone on for a while and suddenly ended without reaching a conclusion. 

    While there isn't a perfect solution to this, you could try out a few different attributes based on what you find causes dropped chats the most. For example, you could include attributes for Chat Agent Messages Brackets and Chat Visitor Messages Brackets. With those, you could see how many dropped chats had very few messages exchanged, or perhaps had an uneven number of messages sent between the agent and visitor. Those could be useful pieces of information in assessing whether chats were truly dropped mid-chat or if they may have concluded naturally with a final comment from the visitor.

    I'll relay this feedback to our team as well, but I hope this at least helps clarify things!

    0
  • Kelsa Henry

    Hi Igor, 

    I fervently support your point. The lack of proper "drop chat metric" in Zendesk has forced us to go the extensive route of data mining in order to understand the nature of this issue. 

    from our research, Zendesk can identify % of chats which unexpectedly ended on the agents end.

    Can we get a metric which identifies:
    1) "Chat which unexpectedly ended" from agents side
    2) "Chat which unexpecetedly ended" on the end user side. 
    3) "Total chats which unexpectedly ended.

    regards 

    2
  • Qin Brian

    Hi ZD: Is there possible to use function in ZD to separate the chats which are really dropped? Can we find type is chat.memberleave in ZD query? Then can we separate the member is agent instead of user?

    0
  • Marco
    Zendesk Customer Care
    Hi Qin,
     
    Thanks for your query. Dropped chats will only really have one factor to take into consideration, which is that the chat was ended by the agent with an unanswered message from a visitor. Not sure what you mean by finding the type that is chat.memberleave in ZD query, as well as separating the member, but dropped chats will only be counted if the agent leaves with an unanswered message, hence if it is the end user that leaves, then it is not counted as a dropped chat. 
     
    Hope this helps clarify it for you! Cheers! 
    0
  • Qin Brian

    You can see the details I got from API of audit.

    It's too complex to use API for analysis.

    IS there convenient way in ZD's dashboard to find the leave member is agent or customer?

     

    0
  • Marco
    Zendesk Customer Care
    Thanks for getting back to us, Qin. Unfortunately there would be no way of doing this on the dashboard level. You could look into your metrics which are missed chats, dropped chats, and completed chats which would give you an idea. More information here: https://support.zendesk.com/hc/en-us/articles/4409149177242
     
    As mentioned before, dropped chats will only count chats where the agent left the chat with an unanswered message by the customer, but if it is the agent that left the chat without an unanswered message then it would be counted as a completed chat. 
     
    Hope I was able to clarify it. Thanks! 
    0
  • Zachary Hanes

    Above, Marco says this:

    dropped chats will only count chats where the agent left the chat with an unanswered message by the customer

    Mark's original answer says this:

    Dropped chats are the number of chats where the conversation with an agent ended with an unanswered message from a visitor.

    In the linked documentation, it says this:

    The number of chats accepted by or assigned to the agent where the visitor responds to an agent message, then leaves the chat.

    Mark's answer seems to suggest that it counts as a dropped chat even if it was ended by the end user. Marco is saying it only counts if it was ended by the agent. 

    Can someone from Zendesk tell us which is correct? Do the metrics distinguish between a chat dropped by an agent and one "self-dropped" by the end user?

    1
  • Dane
    Zendesk Engineering

    Zachary Hanes,

    A chat will be considered as a Dropped Chat when a conversation ended, either closed by an agent or end-user, and there is a message that was not answered by an agent. I understand that there are also some similar requests to modify how Dropped Chats are calculated. As it turns out, it's not yet on our roadmap. You can also try to use Messaging, which does have different classifications of conversation types.

    0
  • Bobby Koch

    Linked out page is dead

    0
  • Brett Bowser
    Zendesk Community Manager
    Thanks for the heads up Bobby! I updated the article with the correct link. Cheers!
    0
  • Brian

    Hello ZD. If we were to change the origin of a missed/drop chat from chat to email, would this throw off our D_COUNT for missed chats in Explore? Or are D_COUNT missed chats tracked separately from the ZD fields?

    0
  • Dane
    Zendesk Engineering
    Hi Brian,
     
    If you are pertaining to just simply changing the channel of your responses, it will not have any effect on it as they are from two different datasets.
     
    Hope this helps!
    0

Please sign in to leave a comment.

Powered by Zendesk