Best practices for Amazon Connect outbound campaigns
The topics in this section explain best practices for outbound campaigns. These practices can increase agent productivity, help you comply with regulations, and help protect the integrity of your phone numbers.
Contents
Choose the right mode for voice communication
Amazon Connect outbound campaign provides several types of voice communication. The following sections describe each type so that you can implement the campaign that best meets your needs.
Contents
Predictive (Agent assisted voice)
When agent productivity, cost per calls, or contact center efficiency are critical metrics, use predictive. Predictive mode anticipate that many calls won't be answered. They counterbalance that by dialing as many phone numbers in a list as possible during an agent's shift by making predictions about agent availability.
The predictive algorithm calls ahead based on certain performance metrics. This means that calls can be connected before an agent becomes available, and a customer is connected to the next available agent. The predictive algorithm constantly analyzes, evaluates, and makes agent availability predictions in real-time so that agent productivity and efficiency can improve.
Progressive (Agent assisted voice)
When you need to reduce answer speeds, use progressive. The progressive mode dials the next phone number in a list after an agent completes the previous call. It only makes as many outbound calls as there are agents available.
You can use integrated answering machine detection to help identify a live customer pickup or a voicemail, and customize your contact strategy accordingly. For example, if a person answers a call, you can present options for them to select. If a call goes to voicemail, you can leave a message.
You can also manage pacing by specifying capacity for each campaign. For example, you can send more voice notifications faster by setting a higher capacity for a given agentless campaign compared to other dialer campaigns.
Agentless (Automated voice)
You use agentless mode to send high-volume personalized voice notifications, appointment reminders, or to enable self-service using the Interactive Voice Response (IVR) with no agents needed.
Agent staffing best practices for voice communication
When call recipients answer a call and hear silence in return, they often hang up. For predictive mode, use the following best practices to help reduce that silence:
-
Ensure that you have enough agents logged in to your call queue. For more information about staffing, see Forecasting, capacity planning, and scheduling in Amazon Connect.
-
Consider using Amazon Connect's machine learning services.
-
Forecasting. Analyze and predict contact volume based on historical data. What will future demand—the contact volume and handle time—look like? Amazon Connect forecasting provides accurate and auto-generated forecasts that are automatically updated daily.
-
Capacity planning. Predict how many agents your contact center will require. Optimize plans by scenarios, service level goals, and metrics, such as shrinkage.
-
Scheduling. Generate agent schedules for day-to-day workloads that are flexible, and meet business and compliance requirements. Offer agents flexible schedules and work-life balance. How many agents are needed in each shift? Which agent works in which slot?
Schedule adherence. Enable contact center supervisors to monitor schedule adherence and improve agent productivity. Schedule adherence metrics are available after the agent schedules are published.
-
Connection latency best practices
Successful outbound calling campaigns avoid silent calls, the period of silence after a person answers a call and before an agent comes on the line. Legal requirements to limit the number of silent or abandoned calls and keep the called party informed may also apply. You can configure Amazon Connect in different ways to reduce call connection delays.
Contents
Outbound agent-staffed calling
When using the Check call progress flow block:
-
Call Answered branch - Remove all flow blocks between the Check call progress and Transfer to queue blocks. This minimizes delay between the dialed party saying hello, and the agent answer time.
-
Not Detected branch - This branch should be treated in the same manner as a Call Answered with routing to a Transfer to queue block. This branch is used when the ML-model was not able to classify the answer type. Since this could be a voicemail or a live person, you can play a message before to the Transfer to queue block in the event that there is a voicemail answering a message can be left.
For example, "This is Example Corp. calling to confirm your appointment. We couldn't tell if you or your voicemail answered this call. Please stay on the line while we connect you with an agent."
Outbound agentless calling
Outbound campaigns often use custom greetings and self service functions. Do not use Lambda functions to get contact attributes. Instead, provide customer data (attributes) via the campaign segment. Use these attributes from the campaign segment to play custom greetings.
-
Example - Call Answered or Not Detected: "Hello,
$.Attributes.FirstName
. This is$.Attributes.CallerIdentity
calling to confirm your upcoming appointment on$.Attributes.AppointmentDate
at$.Attributes.AppointmentTime
. If this is still a good time and date for you, just say, "Confirm". If you would like to use our self service system to modify your appointment, just say, "self service" or stay on the line and we will connect you with the next available agent." -
Example - Voicemail with or without beep: "Hello,
$.Attributes.FirstName
. This is$.Attributes.CallerIdentity
calling to confirm your upcoming appointment on$.Attributes.AppointmentDate
at$.Attributes.AppointmentTime
. If this is still a good time and date for you, we will see you then. If you would like to modify your appointment, please call us back at$.SystemEndpoint.Address
to reschedule your appointment" -
Error branch - Occasionally there could be an issue that causes a call to follow the Error branch. As a best practice, use a Play prompt block with a message that applies to the contact that was dialed, with instruction to "Please call us at
$.SystemEndpoint.Address
to confirm or reschedule your appointment." Do this before the Disconnect / hang up block in case the call recipient answered, but an error occurred in the processing.

Whisper and queue flow best practices
-
Remove the Loop prompts from the Default customer queue flow and replace them with End flow / Resume.
-
If agents don't answer within 2 seconds of calls going to queue, you can minimize silent calls by using Loop prompts and play a message for the customer. The following image shows a typical flow block with a Loop prompt.
-
Use the Disable agent whisper and Disable customer whisper options on the Set whisper flow block. This is so customers perceive less connection latency as part of an outbound campaign. The following image shows the location of the Disable agent whisper setting on the block's properties page.
User administration best practices
We recommend setting the following options for your users to reduce connection times. To access these settings, in the Amazon Connect admin website navigate to Users, User management, Edit.
These options apply to soft phones only.
-
Enable the Auto-accept calls. It reduces the potential for call connection latency/delay after a dialed party answers.
-
Set After Contact Work (ACW) timeout to 30. Minimizing the ACW time will optimize the dialing algorithm when using Predictive dialing campaigns.
-
Enable persistent connection. This maintains agent connection after a call ends. It enables subsequent calls to connect faster.
The following image shows the Settings section of the Edit user page.

Workstation and network best practices
The following best practices can help optimize agent efficiency by ensuring adequate hardware and network resources.
-
Ensure that agent workstations meet the minimum requirements. For more information, see Agent headset and workstation requirements for using the Contact Control Panel (CCP).
-
Ensure that the agent has the CCP or agent workspace open and present on their desktop. This reduces the time spent bringing the screen to the front before greeting the caller.
-
On the local network, ensure that the agents are connected to a LAN. This mitigates potential wireless network latency
-
If possible, minimize the geographic distance between the AWS Region that hosts your Amazon Connect instance and the agents that interact with the outbound campaigns. The greater the geographic distance between your agents and the hosting Region, the higher the possible latency.
Note
Outbound campaigns have limitations on the numbers that agents can dial, depending on the
origin of the Amazon Connect instance. For more information, see the Amazon Connect
Telecoms Country Coverage Guide
Testing best practices
As a best practice, run tests at scale. To achieve the lowest call connection latency, use outbound campaigns to make hundreds of thousands of continuous calls to mimic your production environment. Call connection latency can be relatively high when making a handful of campaign calls.
Best practices for answering machine detection
To use Answering Machine Detection (AMD) in a campaign, use the Check call progress flow block. It provides call progress analysis. This is an ML-model that detects an answered call condition so that you can provide different experiences for calls answered by people and calls answered by machine, with or without a beep. The flow block also provides a branch for routing calls when the ML-model can't distinguish between people and voicemail, or when errors occur in call processing.
AMD uses the following criteria to detect live calls:
-
Background noise associated with a pre-recorded message.
-
Long strings of words such as "Hello, I am sorry I missed your call. Please leave a message at…"
-
A live caller saying something similar to "Hello, hello?" followed by a post-greeting silence.
Forty to 60-percent of calls to consumers go to voicemail. AMD helps eliminate the number of voicemail calls over live calls. However, the detection accuracy has limitations.
-
If the voicemail greeting is a short "Hello" or includes a pause, AMD detects it as a live customer (a false negative).
-
Sometimes a long greeting by a live customer is incorrectly detected as a voicemail (a false positive).
-
There is a small delay while the system connects the call to an agent, which could result in the customer possibly hanging up.
-
PBX (private branch exchange) numbers with multiple levels of voicemail prompts are not supported.
The pros, cons, and best uses of Answering Machine Detection
The use of Answering Machine Detection (AMD) may not comply with telemarketing laws. You are responsible for implementing AMD in a manner that is compliant with applicable laws, and you should always consult your legal advisor regarding your specific use case.
Use case 1: AMD is on and leaving automatic voicemails
-
Pros – Agents primarily interact with live calls 95-percent of the time, maximizing talk time. AMD can leave automatic voicemails if a voicemail is detected.
-
Cons – The technology leaves a voicemail 50-percent to 60-percent of the time due to false positives due to the large variety of answering machine types. Also, AMD can irritate customers because it adds a short delay to live calls.
-
Best uses – Calling consumers during the day when you may get a large quantity of answering machines and it’s not urgent to ensure every call receives a voicemail.
Use case 2: AMD is on but not leaving automatic voicemails
-
Pros – Agents primarily interact with live calls 95-percent of the time, maximizing talk time.
-
Cons – Cannot leave any voicemails. Adds a delay to live calls which can annoy customers.
-
Best uses – Calling consumers during the day when you may get a large quantity of voicemails and you don’t want to leave any voicemails.
Use case 3: AMD is off and agents can leave manual voicemails
-
Pros – Voicemails can be left 100-percent of the time.
-
Cons – Agents must determine whether they are receiving a live call or voicemail. Must manually leave a voicemail. Most time consuming and can lower the number of calls your agents make in a day.
-
Best uses – Calling consumers or businesses and leaving customized voicemails.
Use case 4: AMD is off and agents can leave a prerecorded voicemail
-
Pros – Agents can leave a personalized, pre-recorded voicemail 100% of the time saving significant time by avoiding repeating the same message over and over with 'Voicemail Drop'.
-
Cons – Agents must determine whether they are receiving a live call or voicemail. More time consuming than AMD but quicker than manually leaving a voicemail.
-
Best uses – Calling consumers or businesses and leaving generic voicemails.