How to handle unexpected contact spikes with Amazon Connect

Historically, and especially when there are events that are not anticipated, contact center managers have struggled to handle an unexpected increase in call volumes. If not handled properly, this can cause poor customer experiences, such as long wait times or callers unable to reach the contact center, resulting in loss of revenue. Traditional contact centers are especially strapped when faced with a surge of unexpected calls because of limited ability to scale quickly. Alternatively, they have to plan and pay for peak capacity at all times. Inflexible seat-based licensing and contact volume-based architectures are difficult and expensive to scale on short notice. In addition, it is challenging to quickly change the customer experience in reaction to long call queues. Recording time sensitive prompts, updating them in the system, and changing call routing requires professional services and takes several days to weeks.

Amazon Connect was built from the ground up to be 100% cloud based with easy to use UIs to quickly make changes to contact flows, automatic scaling, and consumption based, pay as you go pricing. Amazon Connect enables businesses to react quickly to customer demand by on-boarding remote agents in minutes, editing contact flows easily, and quickly creating automatic caller experiences that are dynamic and personalized.

This blog post serves as a guide to provide actions and measures that you can take with Amazon Connect for handling high contact volumes gracefully, in turn providing your customers the best experience possible. A few topics we will cover are:

• Add agents quickly – remote or in office
• Automate frequently asked questions (FAQs)
• Provide estimated wait times
• Offer callbacks
• Manage queue depth
• Manage Service Quotas
• Monitor call volumes
• Use best practices for API calls

Finally, we will provide guidance on how to engage with AWS Support as well.

Add agents quickly – remote or in office
In traditional seat-based licensing models for contact centers, there is always a limit on how many users are configured for surge capacity. However, with Amazon Connect you can quickly add new users to meet increased demand of call or chat volumes and scale down whenever needed, only paying for what you use. Amazon Connect uses secure webRTC for agent softphones; allowing them to work from anywhere – all you need is a laptop with an internet browser and a headset.

You can also reach more customers with web and mobile chat with Amazon Connect and expand the capacity of contact center agents to serve more customers. A contact center agent will be able to handle one voice call or they can handle up to 5 chats simultaneously. Customers benefit from the ability to engage with the contact center in a medium that is comfortable for them. Customers may not always be in a place where they can talk, so having the ability to chat with an agent provides customers with additional contact options as well as allows agents to handle more contacts at one time. Chat transcripts can be helpful for quality monitoring as well as chat bot training.

Automate Frequently Asked Questions (FAQs)
An effective way of diverting contacts from your agents is by automating frequently asked questions. For example, in a voice call you can advertise your website’s FAQ link on the IVR (Interactive voice Response) or during queue wait times using Amazon Connect’ s loop prompts. These messages could redirect customers to the FAQ links addressing a chunk of calls. These messages can also be leveraged during both open and closed hours to address customer queries at all times.

Additionally, you can use Amazon Lex bots to facilitate self-service in a conversational style, which gives a better customer experience. You can build Amazon Lex conversational bots and deploy it on your contact center IVR. Self-service options designed to address commonly asked queries are especially useful as they allow agents to spend time on meaningful and complex conversations. You can build one Amazon Lex bot and deploy it both on your voice or chat channel. You could read the Amazon Connect with Amazon Lex Press or Say Input blog post or the Create a questionnaire bot with Amazon Lex and Amazon Alexa blog post for some ideas.

Provide estimated wait times
Unexpected call spikes can result in longer than usual wait times. One can improve customer satisfaction by providing them a proactive notification of this information. Letting the callers know of long wait times is respectful of their time as well as sets appropriate customer expectations. With Amazon Connect, you can view the real-time metrics for your contact center queues. You can add prompts to advise customers of higher-than-normal call volumes and update the contact flows based on the queue wait time or queue size. For example, when wait time is > 2 minutes -play a high call volume message. When wait time > 5 minutes- play a high call volume message and offer a callback. Under extremely high call volume and limited agent staffing, you can check queue status and play a prompt advising customers to call back at a less busy time, and disconnect the call.

The prompt block in Amazon Connect enables you to play an audio prompt. Alternatively, you can choose the language and the voice and play a text to speech prompt. You can use SSML tags for a more personalized experience for your customers.

Offer callbacks
A queued callback as shown in the sample queued callback can be offered when a customer has been waiting in queue to be connected to an agent. A queued callback means that a customer can provide a phone number that they would like to be called on when an agent becomes available instead of waiting in queue.

Additionally, callers will continue to hold their position in queue when they opt for a callback. This leads to a better customer experience and saves money because you only pay for the customer-connected time with Amazon Connect.

Manage queue depth
Setting queue limits helps contact center managers define the queue depth and automate call routing actions when the queue capacity is reached. Having manageable queue depths can build agent morale and also be a better customer experience by customized routing when the queue is at capacity.

You can set queue limits for heavy hitter queues such as collection queues around payment due dates, travel cancellation queues during inclement weather or other disaster situations. This can be done by setting the maximum contacts in queue value and then accordingly define queue treatment for those queues. Contact center managers can then set up overflow routing using the transfer to queue block.

Manage Service Quotas
Service Quotas are the maximum number of resources that can be created in an AWS Service. Unless otherwise noted, each quota is Region and instance specific. You can request increases for some quotas, while other quotas cannot be increased. These are the Amazon Connect Service Quotas. When contact center managers monitor their Amazon CloudWatch alarms configured for Amazon Connect and determine that a value needs to be increased, they can file a request using Service Quotas. To obtain the current quota value for any specific Service Quota, you can raise a ticket through the AWS Support console. Customers who have AWS Enterprise Support have the additional option of engaging their AWS Technical Account Manager (TAM) to help assist with this.

Additional metrics for monitoring and alerting and a sample setup is shown in the Monitoring Amazon Connect with CloudWatch Dashboards and Alarms blog. It shows how contact center managers can use SNS to fan out notifications such as sending an email and trigger a Lambda function. The Lambda function could be set to initiate an outbound call to notify administrators. It can also enable Amazon an Connect contact flow to play a high-volume message for future calls and request customers to call back at a later time.

Monitoring call volumes
You can check concurrent call volumes by monitoring the ConcurrentCallsPercentage quota for Amazon Connect. You can set up a CloudWatch alarm that is triggered on the ConcurrentCallsPercentage metric for Amazon Connect. When the ConcurrentCallsPercentage metric reaches 80–85% of the limit for an Amazon Connect instance, the contact center managers can initiate business processes to reduce contact handle times. At 90-95% of the limit, automation can be enhanced within Amazon Connect contact flows to address increased volumes. Automated actions such as offering callback until the ConcurrentCallsPercentage metric falls below 90%, or playing a message that informs callers of high call volumes and requests them to call again later. If the ConcurrentCallsPercentage metric is expected to remain high, you can open a support case for a service quota increase with business justification using the AWS Support Center.

Use best practices for API calls
Contact center teams should ensure that custom applications leverage best practice to call Amazon Connect APIs less frequently, and write the result set to a caching solution. Applications can then look up the information from the cache, instead of calling the API each time. Architecting the applications in this way ensures that they will perform well under high contact volumes. Data populated by the API does not change drastically within a short time frame, so using slightly older data for the most part will not negatively affect decision making based on this data. This approach will need some fine-tuning to make a judgement call on how frequently the APIs need to be called to write to the cache without compromising the freshness of the data. For example, for the GetCurrentMetricData API gets the current metrics related to the agent and contacts. Refreshing the result set every 90 seconds may suffice for lower call volumes but may need to be refreshed more often if the volumes are higher. So, you could start with calling the API every 90 seconds, and if the data in cache is very different from the API call 90 seconds later, then you can change to 60 seconds or lower. For the ListQueues API lists a summary of the queues. The queue configuration is fairly static, so refreshing the data every hour or every four hours may suffice. Amazon Connect customers have successfully implemented cache solutions using Amazon ElastiCache. Other options would be to use API Gateway caching, or storing the data in DynamoDB with the TTL set to a desired value. Additionally, applications should be resilient to failures and use retries and backoff. If an API call fails, the applications should wait for some time (backoff) before retrying the API call.

Getting Help
Should you have any questions or need guidance on any of the topics covered here, we are here to help. You can reach us via the AWS Support Center. For AWS customers that have Enterprise Support, make sure to engage your TAM to help with any support-related items and to help escalate any urgent issues.

Conclusion
Using these suggestions and recommendations, contact center managers can gracefully handle unexpected call volumes resulting in a positive customer experience and effective handling of call volumes. As explained in the sections above, this can be done with ease and with little to no coding in Amazon Connect. Having queue limits, queued callbacks, Amazon Lex bots for self-service, combined with the ability for agents to work from anywhere gives you the flexibility to customize the experience to meet your business needs. Amazon Connect enables you to scale up to handle the unexpected high call volumes and scale back down once the demand normalizes, while only paying for customer value minutes.

Anjali Khambete, SA Amazon Connect, AWS and Terry Chapman, TAM, AWS