Artificial intelligence in the post-deep learning era
Chat and the Concurrency Myth
1. Article
Chat and the Concurrency Myth
The Impact of Simultaneous Chats on Customer
Satisfaction and First Contact Resolution
2. 2
The Concurrency Myth
Too Much of a Good Thing is Bad
You may have heard this argument time and
In designing [24]7 Assist, our intelligent chat
time again: chat is a very productive channel to
product, driving higher concurrency was an early
interact with customers because a chat agent
goal. One reason for this was the extremely
can talk to more than one customer at the same
time.
favorable math around raising concurrency.
However, it soon became clear from our
usability tests with actual agents that if
concurrency stayed at relatively high values of
2.5 or above for a sustained period of time, other
success metrics started to take a slide. There is
a direct impact on agents’ stress levels as well;
we have observed dips in ASAT (Agent
Satisfaction) scores with rising concurrency
which has led to personnel attrition. The simple
truth is that humans, even really well-trained
Unfortunately, only one part of that argument is
true. [24]7 research shows that while
ones, are not behaviorally suited to multitask
effectively.
concurrency - the average number of chats that
While your chat vendor might advise you to
an agent can handle simultaneously - is one of
target a concurrency of 4 or higher, the truth is
the factors leading to efficient interactions, there
that it is easy to achieve high concurrency
is a limit to how far you can stretch concurrency.
values and still be quite inefficient. For example,
More important, that limit is nowhere near as
a single “bad” chat with a high handle time can
high as many chat software vendors would like
easily push your concurrency up by a point, but
you to believe. Data shows that, as concurrency
will also cause your CSAT/ NPS and other
business metrics to nosedive.
increases beyond a point, other key success
metrics such as CSAT, First Contact Resolution
and Agent Satisfaction scores start getting
impacted.
Our best performing chat programs operate
within a concurrency of under 2, and this is
consistent with industry standards.
Customer satisfaction increases as concurrency decreases.
3. First contact resolution increases as concurrency decreases.
Indicative results from live chat operations over a 3 month period. Higher concurrency levels have an
impact on First Contact Resolution (FCR) and CSAT levels. This trend is consistent across programs
as well as over longer time durations. Source: [24]7 Customer Data.
The key to a high performing contact center is to
find a balance between efficiency and
effectiveness. Contact centers that operate at
high concurrency should track their CPH (chats
per hour) or cost-per-chat, as well as their
customer satisfaction scores, to determine the
right level of concurrency.
Designing for the Metrics that
Matter
We carried our learning back to the design table
with us. We chose to focus on the metrics that
drove real improvements in outcomes, such as
handle and response times, first contact
resolution (FCR) and incremental conversion lift.
There are a few use cases that can
Underlying all of this is one key strategic
possibly support higher concurrency
philosophy and it applies both to your chat
levels. Queues that are low in complexity,
centers and to your chat software: Most software
such as simple enquiries, may sometimes
is built around features. A chat solution should
support higher concurrency. On the other
be built around the outcomes from which you
end of the spectrum, in complex chat
want to benefit. In designing our [24[7 Assist
interactions such as a financial services
chat solution, we carefully handpicked the
queue involving the agent accessing
features that would drive better outcomes, and
multiple customer accounts
reimagined them with a laser focus on
simultaneously, higher concurrency can
result in agent error.
enhancing agent throughput and interaction
effectiveness. Here are just a few examples to
illustrate how we made our design principles
manifest in the [24]7 Assist agent workspace:
The Assist chat UI is built with a unique
tabbed view, called the carousel, which