Talkdesk® Live™ Dashboards and Explore™ are now using a New Data Model which increases your visibility into call metrics. Accompanying this increased visibility are updates in how Talkdesk records certain call data.
The purpose of this document is to provide an overview on how the methods used in Live Dashboards and Contacts Historical Reporting in Explore differ from the old data model and how you can best leverage it for your organization.
Note: This is a high level overview of the new methods and metrics used in Live Dashboards and Explore. For in-depth examples on these metrics and how they might differ from the old reporting model, see the article New Data Model Metrics Examples.
The New Data Model records the different stages of an interaction with more precision. These divisions are defined below.
A segment is a small component of an overall interaction between a caller and an agent. Examples include Queue, Ring, Talk, Hold, After Call Work (ACW), among others.
Segments can overlap for one interaction. For example, two segments coincide after the interaction is transferred: one for when the first agent is in ACW, and another for when the customer is placed into the new, post-transfer queue.
A contact is a collection of segments that generally starts with a wait segment and ends with another wait segment, an ACW segment, or the termination of the contact. Contacts are created when the interaction reaches a queue.
Contacts are the primary object from which contact center metrics are calculated. Each segment of a contact, the duration of those segments, and the events within the contact are used to report the dimensions and metrics for contact center KPIs.
Interactions are a collection of contacts. They include all events from the time a call is connected to a flow to the end of the call, including ACW. Contacts are only created when an interaction reaches a queue, therefore interactions that don’t reach the queue don’t have contacts.
Interactions can also exist without contacts. This occurs when an interaction reaches your contact center but it does not reach a queue. This is known as pre-call data.
Interaction (old model) vs. Contact (new model)
Previously, Talkdesk’s reporting model was limited to recording data at the interaction level. This is acceptable in certain instances but becomes problematic in others (especially for transferred calls).
To address these limitations, the methods used to populate the New Contacts Data Model data operate differently in some notable ways. As Talkdesk transitions to New Contacts Data Model, your organization will have access to both interaction-level reporting through the old model and contact-level reporting through Live Dashboards, and Contact Historical Reporting in Explore.
Please consider these guidelines when deciding to view data through the old and new models.
You should be using interaction-level data if you are interested in your customers’ experience in your call center. For example:
You should be using contact-level data if you are analyzing agent-facing metrics. For example:
This new resource for recording call data introduces updated methods for reflecting it, as described below.
Whereas the old data model used the ending time of interaction events to determine which day calls would be counted towards, the New Contacts Data Model uses the beginning time of contacts to record activity. The advantages of this change include insight into the path that calls take, as well as visibility into when contacts arrive, rather than exit.
Queues and Ring Groups
The old data model had limitations reflecting queue usage over the entirety of an interaction. Generally, it would simply apply the latest set of ring groups to a call if it underwent multiple routing events.
However, this falls short of building a comprehensive picture of your organization’s routing activity. To build a more actionable narrative of how calls are travelling through your organization, the new model now records queue-specific data at the contact level. This ensures that activity such as transfers don’t skew data by omitting routing behavior prior to the call’s final queue.
Reporting will use the ring group the call arrives in when it initially encounters an A&D component. Subsequent A&D components encountered through the branch of the flow will be treated as attributes, just as Studio treats them.
When an agent transfers a call to another ring group (queue), reporting will recognize that as an intentional change of queue rather than the flow’s attempt to widen the eligible agent pool to answer calls that are waiting.
This treatment of contacts will provide a clearer picture of the incoming volume, as well as the total amount of time a caller waited before his call was answered.
With the new reporting component, there's an increased flexibility on building the queues. By using this component in your flow you’ll be able to clearly identify when you want your queue to change (please note that this component is in Early Access; more details here).
New and Updated Metrics
This section introduces and defines key new metrics used in Live Dashboards. While it’s not a comprehensive list, these definitions represent the major changes implemented in the new reporting model.
Inbound contacts include all contacts that reach a queue, regardless of the contact’s final status (abandoned, short abandoned, answered, etc.). Many of the contact types below can be considered subsets of this definition.
Abandoned contacts are inbound contacts where the caller disconnected after entering a queue but prior to getting connected to an agent.
Calls which disconnect while ringing an agent also fall under this definition in the new model (the old model considered them missed calls). The rationale for this change is there’s no experiential difference between waiting in the queue and ringing from the customer’s perspective. In other words, it’s less valuable to treat these outcomes differently since the customer is ignorant whether an agent is being rung.
Pre-queue abandons are contacts where the person who initiated the inbound contact disconnects prior to entering a queue. Calls that end during an IVR prompt count as pre-queue abandons.
Note: For now, these are not included in Live Dashboards. They’ll be accessible through the old reporting model during the transition.
Service level percent is the number of contacts answered within your organization's waiting time threshold. This value is configurable through Admin > Preferences as well as through numbers’ custom settings.
Live Dashboards will calculate service level as follows (see New Data Model Metric Examples for more information).
[All contacts answered within threshold / (All inbound contacts - Short abandoned contacts)] x 100
Additionally, Live Dashboards will now show a contact-level wait time for calls, meaning queue-specific metrics for how long customers wait before an agent answers will be available.
Note: You should never attempt to measure agent-level dimensions using service level percentage as it would have no statistical significance.
Average Handle Time (AHT)
Average handle time (AHT) is a new metric which measures how long an agent is occupied while handling a contact for a given queue. Individual values for handle time are a sum of a contact’s talk, hold, and ACW time. Expected use case for this metric is for staffing and forecasting.
Note: “Hold time” above refers to the time where the customer is explicitly put on hold by an agent. It does not count the time spent in a waiting queue before an agent answers.
- Talkdesk’s new data model used in Live Dashboards divides calls into segments, contacts, and interactions. The new model relies heavily on contact-level metrics.
- Date filtering now keys-off beginning time, not the end time.
- Contact-level reporting enables better visibility into wait times and queues.
- Talkdesk’s new reporting metrics include inbound contacts, abandoned contacts, and AHT.
- Service level is calculated using a new formula.
To see specific examples of these metrics and how they differ from the old reporting model, please see the article New Data Model Metric Examples.