The third consideration for Flex text-based communication channels from my introductory post is “What type of interaction is it?”. This topic can end up being a bit more philosophical. That being said, it does drive real-world decisions in how you architect your overall Flex approach. But it will likely take a combination of historic understanding of your customer interactions and some trial and error to get right.
In this post, we’ll look at the third focus area…
What type of interaction is it? Near real-time (web chat as an example) or more asynchronous (such as SMS).
The terms synchronous and asynchronous have a lot of different connotations to developers, but may not immediately make sense in this context. It’s probably easiest to think about the oldest type of customer interaction in the contact center: voice. With voice communication, each agent can really only be engaged with a single customer at at time. You can put the customer on hold or consult with your supervisor or another agent, but generally you are focused on a single conversation at a time. From a Flex standpoint, you can only handle a single Task when the channel type is voice. Voice conversations typically require your full attention. So this is the far end of the synchronous or “real-time” spectrum. In Flex terms, the voice channel should only ever have a capacity of one.
Omni-channel changes that whole paradigm, particularly with text-based channels. If you think about one of the earliest contact center evolutions after voice, it’s probably some variation of chat. Web chat or “instant messenger” types of applications have been around for quite awhile now. These interactions could often be much like voice in that responses from both (or the group) can fly back and forth quickly. Both sides might be typing at the same time. But then sometimes one party would step away or get interrupted. You might send several responses and eventually ask “are you still there?”. Sometimes you might stop chatting at the end of a day and pick it up the next day. This is what I mean by the other end of the spectrum or asynchronous.
So already we can start to see how web chat or similar channels (now Slack, Teams, Facebook Messenger come to mind) can be very synchronous or very asynchronous. Or the communication can go back and forth over the spectrum multiple times over the course of the conversation. The same interaction can move back and forth along that line, even with the same agent involved.
How Async Changes Everything
When communication with a customer starts to move closer to async, things get interesting. A lot of the standard rules you may be used to in voice interactions (or even web chat) go out the window. It’s not so much that is changes the mechanics of how the channel works. But it really changes the number of use cases you need to think about and support.
You will really start to see this with some of the newer native channels in Flex. SMS. WhatsApp. Facebook Messenger. Line.
It starts to become challenging to determine customer expectations. If the customer sends you an SMS, do they expect a response in 10 minutes? An hour? A day? Might depend on the customer. Might depend on what they need help with or their intent. How do you determine an appropriate SLA for your SMS channel? And how do you make sure your agents have the right amount of work at any given time? Or to step up one more level, how do you staff effectively?
Over time with voice, customer expectations have generally become much more concrete. Don’t make me go through 20 layers of IVR. I don’t want to sit in a queue for more than X minutes. I don’t want to finally get an agent and then get transferred to a different queue to wait some more time because you put me in the wrong queue to start. Don’t put me on hold for too long. Give me some additional options during long waits, like callbacks.
These types of questions are still being worked out for text-based channels. They will vary based on your industry, your customers and the types of interactions you have with customers. Metrics become very important to help you hone in on exactly what your customer expects and needs. One nice thing about text-based channels is you already have the transcripts. You can think about deep machine learning or AI to use customer intent, sentiment and other indicators. This can drive your SLA determinations down to the queue, customer or even Flex Task level if necessary.
Is The Conversation Over?
A big question: when is the interaction “done” or “over”? Not as clear as before. The answer usually ends up being something you need to answer based on your own business rules and customer interactions. This question will really challenge the conventional approach to standard contact center KPIs. It also makes WFM more challenging since customer communications may often span shifts or agent availability.
With voice, you have a clear determination when the conversation is over; the call ends. It’s not always that simple with text-based channels. Sometimes there is a natural end to the conversation and sometimes there is not. You may need to rely more heavily on your agents to figure this out, based on real-time feedback and comments from the customer. You will often see this with web chat, where the agent asks the customer if there is anything else the customer needs help with. This can work, particularly when web chat is towards the synchronous end of the spectrum.
With phone-based channels like SMS, it may not be so straight-forward. Sometime the customer just stops responding. They get distracted or mostly got the answer they needed. Or maybe they lost service or left their phone at home or can’t use their phone during a work shift. You might not know for sure. Ultimately you will have to use a combination of customer feedback, agent intuition, historical data and possibly time-based rules to decide when a conversation is over. And you have to be prepared if the conversation starts up again. If the original agent isn’t available, accurate context is important. Customers don’t like starting over with the same questions.
You will also need to think about how these channels impact agent capacity. While voice interactions are usually effectively limited to “one at a time”, many agents can quite successfully manage multiple text-based interactions simultaneously. The pace of communication (how quickly a customer is responding) can also vary a lot between channels and between customers. Metrics become important in helping you to best determine how to approach WFM (Workforce Management) to support an optimal capacity for your various agent teams. Cross-training will give you more flexibility. Flex allows you to treat text-based channels in a similar fashion in how the agent interacts, which makes this easier from a training perspective.
While some of the standard KPIs for contact center still make sense for text-based channels, others don’t provide a lot of help anymore. Concepts like how long the customer waits in queue or to get the initial response still matter. You probably still want to know how many active conversations each agent or team is managing at a given time.
But concepts like handle time don’t make as much sense when text-based channels are mostly going in an asynchronous direction. You still need some concept of handle time, but it’s really more important to think of focus. Which conversation is the agent spending time on right now? What activities are they doing? Are they typing responses or looking at customer history or finding a product? Are they consulting with a supervisor and how does that work?
If it takes twice as much time on your support team to handle a question over web chat vs. SMS, that matters for staffing and agent capacity. Do some customers speak with agents more frequently or require more time generally before they are satisfied with the interaction? As concepts like “conversational commerce” and establishing and maintaining relationships with customers become more of the norm, having the right metrics to track and manage agent staffing and load levels becomes even more crucial.
We’re just looking at the tip of the iceberg here. We could probably take 20 or 30 of the most common KPIs and look at how those metrics might vary across channels. Not only voice vs. text-based, but also within individual types of text-based channels or channels that tend to be more synchronous or more asynchronous. Hopefully this gets you thinking about all the new opportunities omni-channel first contact center platforms like Flex can open up.
We still need to talk about the various capabilities unlocked by text-based channels in Flex. As I’ve already mentioned, Flex collapses a lot of the base functionality onto Twilio’s Programmable Chat platform. So there is a lot of overlap in the way text-based channels work. Some key differences as well. And you can develop your own text-based channels if one of the native ones doesn’t completely meet your needs. We’ll talk about this in the next post.