Real-time metrics definitions - Amazon Connect

Real-time metrics definitions

The following metrics are available to include in real-time metrics reports in Amazon Connect. The metrics available to include in a report depend on the report type.

Tip

Developers can use the GetCurrentMetricData API to get a subset of the following real-time metrics from the specified Amazon Connect instance.

Abandoned

The count of contacts disconnected by the customer while in the queue during the specified time range. Contacts queued for callback are not counted as abandoned. When you create a customized real-time metrics report, to include this metric, choose a Queues report for the type. On the Filters tab, choose Queues, then on the Metrics tab you'll have the option to include Abandoned.

Abandonment rate

The percentage of contacts that were disconnected by the customer while in the queue. Contacts queued for callback are not counted as abandoned.

Active

The count of active slots. This number is incremented for each contact where the contact state is either Connected, On Hold, After contact work, Paused, or Outbound ring.

In the GetCurrentMetricData API, this metric can be retrieved by using SLOTS_ACTIVE.

ACW

The count of contacts that are in an AfterContactWork state. (After contact work is also known as After call work.) After a conversation between an agent and customer ends, the contact is moved into the ACW state.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_AFTER_CONTACT_WORK. The name of this metric is confusing because in the Amazon Connect console, ACW counts the number of contacts that are in an ACW state, not the number of agents.

To learn more about agent status and contact states, see About agent status and About contact states.

Agent Activity

If an agent is handling a single contact, this metric may have the following values: Available, Incoming, On contact, Rejected, Missed, Error, After contact work, or a custom status.

If an agent is handling concurrent contacts, Amazon Connect uses the following logic to determine the state:

  • If at least one contact is in Error, Agent Activity = Error.

  • Else if at least one contact is Missed contact, Agent Activity = Missed.

  • Else if at least one contact is Rejected contact, Agent Activity = Rejected.

  • Else if at least one contact is Connected, On Hold, Paused, or Outbound contact/Outbound callback, Agent Activity = On contact.

  • Else if at least one contact is After contact work, Agent Activity = After Contact Work.

  • Else if at least one contact is Incoming/Inbound Callback, Agent Activity = Incoming.

  • Else if agent status is a custom status, Agent Activity is the custom status.

  • Else if agent status is Available, Agent Activity = Available.

  • Else if agent status is Offline, Agent Activity = Offline. (After an agent moves to Offline, they disappear from the real-time metrics page in approximately 5 minutes.)

If a manager is using the Manager Monitor feature to monitor a particular agent as they interact with a customer, then the manager's Agent Activity will display as Monitoring. The Agent Activity of the agent who is being monitored is still On Contact.

Agent First Name

The first name of the agent, as entered in their Amazon Connect user account.

Agent Hierarchy

The hierarchy the agent is assigned to, if any.

Agent hung up

The count of contacts disconnected where the agent disconnected before the customer.

Agent Last Name

The last name of the agent, as entered in their Amazon Connect user account.

Agent Name

The name of the agent, displayed as follows: Agent Last Name, Agent First Name.

Agent non-response

The count of contacts routed to an agent but not answered by that agent, including contacts abandoned by the customer.

If a contact is not answered by a given agent, we attempt to route it to another agent to handle; the contact is not dropped. Because a single contact can be missed multiple times (including by the same agent), it can be counted multiple times: once for each time it is routed to an agent but not answered.

This metric was previously named Missed.

Agent non-response without customer abandons

If a contact is not answered by a given agent, Amazon Connect attempts to route it to another agent to handle; the contact is not dropped. Because a single contact can be missed multiple times (including by the same agent), it can be counted multiple times: once for each time it is routed to an agent but not answered.

AHT (Average Handled Time)

The average time, from start to finish, that a contact was connected with an agent (average handled time). It includes talk time, hold time, After Contact Work (ACW) time, and agent pause duration (which applies only to tasks).

AHT is calculated by averaging the amount of time between the contact being answered by an agent and the completion of work on that contact by an agent.

API contacts handled

The count of contacts that were initiated by an API operation, such as StartOutboundVoiceContact, and handled by an agent.

Availability

For each agent, the number of available slots they have that can be routed contacts.

The number of available slots for an agent are based on their routing profile. For example, let's say an agent's routing profile specifies they can handle either one voice contact or up to three chat contacts simultaneously. If they are currently handling one chat, they have two available slots left, not three.

What causes this number to go down? A slot is considered unavailable when:

  • A contact in the slot is: connected to the agent, in After Contact Work, inbound ringing, outbound ringing, missed, or in an error state.

  • A contact in the slot is connected to the agent and on hold.

Amazon Connect doesn't count an agent's slots when:

  • The agent has set their status in the CCP to a custom status, such as Break or Training. Amazon Connect doesn't count these slots because agents can't take inbound contacts when they've set their status to a custom status.

  • The agent can't take contacts from that channel per their routing profile.

In the GetCurrentMetricData API, this metric can be retrieved by using SLOTS_AVAILABLE.

Available

The number of agents who can take an inbound contact. An agent can only take inbound contacts when they manually set their status to Available in the CCP (or in some cases when their manager changes it).

This is different from how many more inbound contacts an agent could take. If you want to know how many more contacts an agent can have routed to them, look at the Availability metric. It indicates how many slots the agent has free.

What causes this number to go down? An agent is considered unavailable when:

  • The agent has set their status in the CCP to a custom status, such as Break or Training. Amazon Connect doesn't count these slots because agents can't take inbound contacts when they've set their status to a custom status.

  • The agent has at least one contact ongoing.

  • The agent has a contact in a missed or error state, which prevents the agent from taking any more contacts until they are flipped back to routable.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_AVAILABLE.

Average Active Time

The average time, from start to finish, that an agent was working on a contact (average active time). It includes talk time, hold time, and After Contact Work (ACW) time. It applies to both inbound and outbound calls.

Average Active Time does not include any time spent in a custom status; Contact handle time (CHT) does include the time spent in custom status.

Average Agent Pause Time

Average time that the contact was paused after being connected to agent during an inbound or outbound contact.

SUM(agent_pause_time)/Number of Contacts that were paused

Average API Connecting Time

The average time between when a contact is initiated using an Amazon Connect API, and the agent is connected.

Avg abandon time

Average time, in seconds, that abandoned contacts were in the queue before being abandoned.

Avg ACW

Average time, in seconds, that contacts spent in the After contact work state, during the specified time range.

This is not the average amount of time agents spent on contacts.

To learn more about agent status and contact states, see About agent status and About contact states.

Avg callback connecting time

The average time between when callback contacts are initiated by Amazon Connect reserving the agent for the contact, and the agent is connected.

No equivalent to this metric is available in the GetCurrentMetricData API.

The following image shows the five parts that go into calculating Avg callback connecting time: Amazon Connect assign work item to agent, agent accepts work item, connection build time, network connection time, rings for customer. It also shows what is in the agent event stream: Connecting, Connected or no answer.


                        The five parts used to calculate average callback connecting
                            time.

Average customer hold time

Average time that customers spent on hold while connected to an agent. This is calculated by averaging CustomerHoldDuration (from the contact record).

Average greeting time agent

This metric represents the average first response time of agents on chat, indicating how quickly they engage with customers after joining the chat. It is calculated by dividing the total time it takes for an agent to initiate their first response by the number of chat contacts.

Note

This metric is available only for contacts analyzed by Contact Lens conversational analytics.

Avg hold time

Average time, in seconds, that a contact in the queue was on hold.

This metric doesn't apply to tasks so you'll notice a value of 0 on the report for them.

Avg incoming connecting time

The average time between when contacts are initiated Amazon Connect reserving the agent for the contact, and the agent is connected.

In the agent event stream, this time is calculated by averaging the duration between the contact state of STATE_CHANGE event changes from CONNECTING to CONNECTED/MISSED/ERROR.

No equivalent to this metric is available in the GetCurrentMetricData API.

The following image shows the three parts that go into calculating Avg incoming connecting time: connection build time, network connection time, and requesting agent to accept. It also shows what is in the agent event stream: Connecting, Connected, Missed, or Rejected.


                        The three parts used to calculate average incoming connecting
                            time.

Avg interaction time

Average time, in seconds, that contacts were connected to and interacting with agents. This does not include hold time, agent pause duration (which applies only to tasks), or time spent waiting in the queue.

Avg interaction and hold time

Average time, in seconds, that contacts in the queue spent interacting with agents and on hold. This is calculated as follows:

Avg hold time + Avg interaction time

Average interaction time

Total time that agents spent interacting with customers on inbound and outbound contacts. This does not include Customer hold time or After contact work time.

Average interruptions agent

This metric quantifies the average frequency of agent interruptions during customer interactions by dividing the total number of agent interruptions by the total number of contacts.

Note

This metric is available only for contacts analyzed by Contact Lens conversational analytics.

Average non-talk time

Average of total non-talk time in a voice conversation. Non-talk time refers to the combined duration of hold time and periods of silence exceeding three seconds, during which neither the agent nor the customer is engaged in conversation. To calculate non-talk time, we add up all the intervals in which both participants remained silent and then divide this total by the number of contacts.

Note

This metric is available only for contacts analyzed by Contact Lens conversational analytics.

Average talk time

Average time that was spent talking during a voice contact across either the customer or the agent. This is calculated by summing up all the intervals in which either an agent, a customer, or both were engaged in conversation, and then dividing it by the total number of contacts.

Note

This metric is available only for contacts analyzed by Contact Lens conversational analytics.

Avg queue answer time

Average time, in seconds, that a contact was in the queue before being answered by an agent. This is calculated using the amount of time that the contact was in the queue, not any time that the contact spent in prior steps of the flow, such as listening or responding to prompts.

Average resolution time

The average time, beginning from the time a contact was initiated to the time it resolved. The resolution time for a contact is defined as: beginning from InitiationTimestamp, and ending at AfterContactWorkEndTimestamp or DisconnectTimestamp, whichever one is later.

Avg outbound connecting time

The average time between when outbound contacts are initiated by Amazon Connect reserving the agent for the contact, and the agent is connected.

No equivalent to this metric is available in the GetCurrentMetricData API.

The following image shows the four parts that go into calculating Avg outbound connecting time: agent call customer, connection build time, network connection time, rings for customer. It also shows what is in the agent event stream: Connecting, Connected or No answer.


                        The four parts used to calculate average outbound connecting
                            time.

Callback attempts

The number of contacts where a callback was attempted, but the customer did not pick up.

Callback contacts handled

The count of contacts handled by an agent that were queued callbacks.

Capacity

Displays the maximum capacity that's set in the routing profile currently assigned to the agent. This column can be filtered by channel.

If an agent's routing profile is configured to handle either one voice or up to three chats, then their maximum capacity equals three, when not filtered by channel.

Consult

Deprecated May 2019. When used in a report, it returns a dash (-).

The count of contacts in the queue that were handled by an agent, and the agent consulted with another agent or a call center manager during the contact.

Contact State

The state of the contacts the agent is currently handling. The state can be: Connected, On Hold, After contact work, Paused, Incoming, Calling, or Missed contact.

For queued callbacks, the contact state can also Callback incoming or Callback dialing.

If a manager is using the Manager Monitor feature to monitor a particular agent as they interact with a customer, the manager's contact state is Monitoring; the agent's contact state is Connected.

Contacts Abandoned in X

The number of contacts abandoned between 0 and X, for example, from 0 - 60 seconds.

You can create custom thresholds for contacts abandoned. For more information, Custom service levels.

Custom contacts abandoned is localized to the report where the metric is applied. For example, you create a report that has a Contacts abandoned in 75 seconds metric. You leave the page and then create another report. The Contacts abandoned in 75 seconds metric won't exist in the second report. You'll need to create it again.

Contacts Answered in X

The number of contacts answered between 0 and X, for example, from 0 - 60 seconds.

You can create custom thresholds for contacts answered. For more information, Custom service levels.

Custom contacts answered is localized to the report where the metric is applied. For example, you create a report that has a Contacts answered in 75 seconds metric. You leave the page and then create another report. The Contacts answered in 75 seconds metric won't exist in the second report. You'll need to create it again.

Contacts transferred out

The count of contacts transferred out from queue to queue, and transferred out by an agent using the CCP.

Duration

Amount of time that the agent has been in the current Agent Activity State.

Error

A count of agents in Error state. An agent is included in this metric if they miss a call or reject a chat/task (most common). They could also be counted if there is a connection failure.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_ERROR.

Handled

The count of contacts in the queue that were answered by an agent.

Handled in

The count of incoming contacts handled by an agent during the specified time range that were initiated using one of the following methods: inbound call, transfer to agent, transfer to queue, or queue-to-queue transfer.

Handled out

The count of contacts handled by an agent during the specified time range that were initiated by an agent placing an outbound call using the CCP.

Hold abandons

The count of contacts that disconnected while the customer was on hold. A disconnect could be because the customer hung up while on hold, or that there was a technical issue with the contact while on hold.

In queue

The count of contacts currently in the queue. The queue is updated when the contact is routed to the agent, before the agent accepts the contact.

To learn how this is different from Scheduled contacts in a callback scenario, see How Initial delay affects Scheduled and In queue metrics.

In the GetCurrentMetricData API, this metric can be retrieved by using CONTACTS_IN_QUEUE.

Max queued

The longest time that a contact spent waiting in the queue. This includes all contacts added to the queue, even if they were not connected with an agent, such as abandoned contacts.

NPT (Non-Productive Time)

The count of agents who have set their status in the CCP to a custom status. That is, their CCP status is other than Available or Offline.

Tip

Although agents aren't routed any new inbound contacts while their CCP status is set to a custom status, it's possible for them to change their CCP status to a custom status while still handling a contact. For example, let's say an agent is being routed contacts very quickly. To go on break, they set their status to Break proactively, while still finishing up the last contact. This allows them to go on break and avoid accidentally missing a contact that's routed to them in the sliver of time between the last contact ending and setting their status to Break.

Because agents can be On call or doing ACW, for example, while their CCP is set to a custom status, this means it's possible for agents to be counted as On call and NPT at the same time.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_NON_PRODUCTIVE.

Non-talk time percent

The non-talk time in a voice conversation as a percent of the total conversation duration. To calculate non-talk time percent, Amazon Connect totals all the intervals in which participants remained silent (non-talk time), and then divides this total by the total conversation duration.

Note

This metric is available only for contacts analyzed by Contact Lens conversational analytics.

Occupancy

Percentage of time that an agent was active on contacts. This percentage is calculated as follows:

(Agent on contact (wall clock time) / (Agent on contact (wall clock time) + Agent idle time))

Where:

  • (Agent on contact + Agent idle time) = total amount of agent time

  • So (Agent on contact)/(total amount of agent time) = percentage of time agents were active on contacts.

Important

Occupancy doesn't account for concurrency. That is, an agent is considered 100% occupied for a given interval if they are handling at least one contact for that entire duration.

Oldest

Length of time in the queue for the contact that has been in the queue the longest.

In the GetCurrentMetricData API, this metric can be retrieved by using OLDEST_CONTACT_AGE.

On contact

The count of agents currently on a contact. An agent is "on a contact" when they are handling at least one contact that is either connected, on hold, in After contact work, paused, or outbound ring.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_ON_CONTACT. This metric used to be named On call. You can still use AGENTS_ON_CALL to retrieve data for this metric.

Online

The count of agents who have set their status in the CCP to something other than Offline. For example, they may have set their status to Available, or to a custom value such as Break or Training.

The Online metric doesn't tell you how many agents can be routed contacts. For that metric, see Available.

This metric can be confusing so let's look at an example. Say you see this in a Queues report:

  • Online = 30

  • On Call = 1

  • NPT = 30

  • ACW = 0

  • Error = 0

  • Available = 0

This means 30 agents have set their status in the CCP to a custom status. 1 of those 30 agents is currently on a contact.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_ONLINE.

Queue

The name of the queue associated with the contact the agent is currently handling.

Queued

The count of contacts added to the queue during the specified time range.

Routing Profile

The routing profile for the agent.

Scheduled

The count of customers in the queue for which there is a callback scheduled.

To learn how this is different from In queue contacts in a callback scenario, see How Initial delay affects Scheduled and In queue metrics.

In the GetCurrentMetricData API, this metric can be retrieved by using CONTACTS_SCHEDULED.

SL X

Percentage of contacts removed from the queue between 0 and X after being added to it (Service Level). A contact is removed from the queue when one of the following occurs: an agent answers the call, the customer abandons the call, or the customer requests a call back.

For X, you can choose from pre-set times in seconds: 15, 20, 25, 30, 45, 60, 90, 120, 180, 240, 300, and 600.

Custom service levels

You can also create custom service level metrics. You can also choose from additional durations, such as minutes, hours, or days.

You can add up to 10 custom service levels per report.

The maximum duration for a custom service level is 7 days. That's because in Amazon Connect you can't have a contact that goes longer than 7 days.

Staffed

The count of agents who are online in the CCP, and not in NPT (a custom status).

Another way of thinking about this is, there are two scenarios in which Staffed is not incremented:

  • The agent's status in the CCP is set to Offline.

  • The agent's status in the CCP is set to a custom status.

For example, let's say an agent sets their status in the CCP to a custom status such as Break and they make an outbound call. Now the agent is On call, but Staffed is 0.

If the agent sets their status in the CCP to Available and makes an outbound call, the agent is On call and Staffed is 1.

This metric is available on the Queues report.

In the GetCurrentMetricData API, this metric can be retrieved by using AGENTS_STAFFED.

Transferred in

The count of contacts transferred into the queue during the specified time range.

Transferred in by agent

The count of contacts transferred in by an agent using the CCP.

Transferred in from queue

The count of contacts transferred into the queue from another queue during a Customer queue flow.

Transferred out

The count of contacts transferred out of the queue during the specified time range.

Transferred out by agent

The count of contacts transferred out by an agent using the CCP.

Transferred out from queue

The count of contacts transferred out of the queue to another queue during a Customer queue flow.