The Insights Dashboard is a powerful aggregation tool that allows users to filter on specific dimensions like call type, timeframe, number prefix, country code, and automatically updates the data visualizations to take into account the filter criteria. Key sections of the Dashboard include the total number of calls placed on Twilio, the average length of those calls, who hung up, call completion rate (ASR), calls with high PDD, and calls whose quality performance may have been impacted by network conditions. Call data can be visualized in a number of ways for each metric, and calls can be exported to .csv.
The user experience of voice communication is measured by two key components - call connectivity and audio quality. Connectivity issues can originate from the user’s handset, the last mile link between their phones and the cell tower, their carrier network, or from within Twilio’s network. Congestion, spotty coverage, handoff failures and numerous other issues could prevent you from reaching your users, or cause connected calls to drop. Voice Insights Connectivity dashboards enable you to rapidly identify connectivity problems by giving you real-time visibility into four key metrics:
- Connection Rate: The percent of total calls that reached the destination. Canceled and failed calls are excluded from this calculation.
- High Post Dial Delay (PDD): Calls where the number of seconds elapsed between dialing the last digit of the phone number and the start of ringing exceeds local thresholds. High PDD can sometimes be reported as "dead air", "one way audio", or even "dropped calls".
- Network Affected: Percentage of calls adversely affected by poor network quality. Issues that indicate poor network quality include: packet loss, jitter, and latency.
- Who Hung Up: Breakdown of calls in which the SIP BYE message was received from the caller or callee. The BYE message signals the termination of an answered call. Calls that have not been answered will not have a BYE, so they will be excluded from the results.
To view connectivity analytics for your account enable Voice Insights for Carrier calls from Insights settings.
Start by selecting the filters as shown below.
1. Call Type Selection: Choose from the various call types that can occur on the Twilio platform.
2. Date Range Picker: Required. Select any date range between 1 hour and 30 days. Metrics on the entire dashboard are calculated for calls made from the current account in the selected date range.
3. Analytics filters: You can narrow down the metrics you are viewing by choosing to view the data for a specific phone number, number prefix, country, device type, and users' carrier network. The applicable values are listed below.
|Phone Number||The phone numbers to which calls were made or were received from.|
|Carrier||The carrier names of the user's phone number.|
|Country Code||ISO Alpha-2 country codes|
Landline, Mobile, or VOIP.
|Number Prefix||The area code of US based numbers. The available area codes are based on the prefixes of phone numbers using which the calls were made.|
|SIP Response||The last SIP response code received in the dialog.|
4. Value Selector: Will pre-populate with most frequent options. You can select multiple values from the options available.
5. Direction Filter: Helps you narrow down the data for the "To" phone numbers (destination), or for "From" phone numbers (source). By default, data is aggregated for both directions.
Our trends graphs show comparisons of call volume and average length of call (ALOC) for the selected time frame. Call volume and ALOC are classic telecom metrics and changes in these should be investigated closely; for example, a very common pattern when a quality problem is introduced is the number of calls to increase, while the ALOC goes down. This is due to people being unsatisified with the quality of their experience, ending the call early, and placing another call (or calls) in an attempt to get an acceptable experience. You can also select-to-zoom on these graphs for greater detail.
|Total Calls||Total number of Twilio calls made from this account over the selected date range.|
|Average Length of Call||Sum of duration of all calls divided by the total number of calls made from the account.|
Connection Rate allows you to filter by call state.
The Voice Insights account-level dashboards provide trends on the following metrics:
|Who Hung Up||
Who disconnected the call, based on the direction of the SIP BYE event. Percentage is calculated based on the total number of errors with respect to the total calls.
|Connection Rate||Total calls completed with respect to the total number of calls. Documentation for the applicable call status values is here.|
Post-dial-delay (PDD) is the number of seconds elapsed between dialing the last digit of the phone number and the start of ringing.
The tab represents percentage of total calls with high Post-Dial-Delay (PDD), compared against 99th percentile of all Twilio calls to the destination country.
This tab shows percentage of calls adversely affected by poor network quality. Issues that indicate poor network quality include packets arriving out of order (jitter), never arriving (packet loss), and taking a long time to arrive (latency). For Carrier and SIP/Trunking call types, high packet loss is defined as 5% cumulative packet loss, high jitter is defined as average jitter of more than 5 milliseconds, and high latency is defined as Twilio-internal RTP traversal time of greater than 150 milliseconds. For SDK calls high jitter is defined as 30 milliseconds or higher for 3 out 5 consecutive samples, and high packet loss at 1% or higher for 3 out of 5 consecutive samples.
|Who Hung Up Values||Description|
|Caller||The BYE message was received from the "From" number|
|Callee||The BYE message was received from the "To" number|
|Error||A SIP error code was received on the call. The count is based on the total number of server or global errors seen.|
The call logs view helps you drill down into a subset of calls with specific issues. To get to the call logs, click on the phone icon towards the right of the screen as shown below. These results can also be exported to a comma-separated value file.