I'm not sure why Zendesk deleted this tip I wrote previously, but I keep backups of everything, and figured I'd put it back up, since this is still something people are asking about chat, and seems helpful. 

When working with chat data, it became clear that when you have automatic chat assignment on, there is a need to better understand why a chat was marked as Missed. I did some brainstorming with team mates, and we broke down missed chats into a few buckets:

1. Nobody was available - Someone chatted in, waited a reasonable amount of time, but no agents were assigned to the chat as they were busy, and then the chatter left, without being served.
2. The chatter left quickly - Someone chatted in, then left very quickly, before an agent could join. 

3. Tried Agents - Someone chatted in, waited, agents were assigned to the chat, but none of the agents picked up in the 30 second window.

As you read these, you might start to see why these buckets are needed. If you had a high volume of missed chats for scenario 1, you need more agents on chat. If you had a high volume of missed chats for scenario 2, adding more agents would not help, but a trigger message about waiting for help might. For scenario 3, instead of adding more agents, you might want to investigate why those available chat agents aren't picking up chats on shift. 

Ok, so we know that we want to break down the Missed Chat total and label them, how can we do it? Through the power of custom formulas!  And good news - it’s a single formula. 

Before we get to the formula, I want to address why there’s added complexity to the formula that seems unnecessary. The built-in attribute “Chat Wait Time” is calculated incorrectly, so we cannot use it, and must build our own from scratch. You could make this formula cleaner by creating a custom metric for “Corrected Chat Wait Time” and inserting it in place of the string "((VALUE(Chat no reply time (sec))+VALUE(Chat duration (sec)))/60" but for simplicity’s sake, I’ve just included that calculation directly in my formula.

In my formula, I’m saying a wait time equal to, or greater than 2 minutes, and trying at least 1 agent, should be categorized as “Tried Agents”. In this scenario, agents have 30 seconds to pick up a chat, so this felt reasonable. You may want to adjust the duration and number of agents, if you give your agents more (or less) time to pick up a chat.

IF [Chat completion]="Missed"

AND ((VALUE(Chat no reply time (sec))+VALUE(Chat duration (sec)))/60)>=2

AND (VALUE(Engagements))>1

THEN

"Tried Agents" 

Next, is the section for “Left Fast”. Again, you may want to adjust the formula. This is set to if the chat wait time was less than 2 minutes, marked them as “Left Fast”:

ELIF

[Chat completion]="Missed"

AND ((VALUE(Chat no reply time (sec))+VALUE(Chat duration (sec)))/60)<=2

THEN

"Left Fast"

Finally, the situation where someone waits more than 2 minutes, but nobody was assigned to the chat (IE all agents were busy on other chats), is simply the remainder.

ELSE

"Nobody Available"

ENDIF

Altogether:

IF [Chat completion]="Missed"

AND ((VALUE(Chat no reply time (sec))+VALUE(Chat duration (sec)))/60)>=2

AND (VALUE(Engagements))>=1

THEN

"Tried Agents" 

ELIF

[Chat completion]="Missed"

AND ((VALUE(Chat no reply time (sec))+VALUE(Chat duration (sec)))/60)<2

THEN

"Left Fast"

ELSE

"Nobody Available"

ENDIF

It’s important to remember that if you adjust the chat wait times, that you do not leave any gaps between these “brackets”. If you set “Tried Agents” to 5 minutes or greater , you need to set “Left Fast” to 5 minutes or less, or else people who waited 4 minutes, will get counted as “Nobody Available”, which is not what you want. 

Here’s the end result, the top Metric is Missed Inbound Chats, and I've filtered to a specific time range.