Containment rate is a well-established performance metric in the customer service world, but not as well-known as NPS or hold time. I would say it’s on the “wonkish” side as far as metrics go.
So, you can imagine my delight to see “containment rate” mentioned in a New York Times Op-Ed last week! Not surprisingly, it was in the context of a “robots-will-take-our-jobs” thesis, but it at least was a more nuanced approach than what we usually see in the mainstream press.
Why Containment Rate Matters
In general, we say that a transaction was “contained” if it was completed entirely through self-service and didn’t require an agent. It’s pretty straightforward to talk about the containment rate of a particular channel for a particular transaction. For example, you could look at people that started the “make a reservation” process on the website and see how many completed it.
We can also talk about the aggregate containment rate across all channels and all transactions. This indirectly answers the question, “How much are human agents needed?” Since companies are eager to reduce labor costs, raising containment rate is a logical goal.
But here we can run into trouble, because there’s a lot of detail that you can skip over when you move from the narrow containment rate (one specific transaction on one channel) to the aggregate containment.
Containment vs. Coverage
One common mistake is to ignore the effects of coverage. Coverage refers to which transactions are available as self-serve. Coverage is never 100%. There are always transactions that are too rare or too tricky to implement in self-serve.
One of the big fallacies you hear when people discuss AI and customer service is confusing containment with coverage. For more on this, see “How to Think about Chatbots in a Big Picture Kinda Way.”
To summarize the issue: People think AI is going to expand coverage – that transactions which currently need agents will become self-servable. Yet, what’s happening today, and the only thing plausibly happening in the near-term, is an increase in containment: Self-serve systems getting marginally better at the transactions they already do.
Handpicked related content for you:
Read the Contact Center Trends 2020 report here.
Friedman’s Stream of Consciousness
Now back to the NY Times article by Thomas Friedman. Here’s the relevant paragraph:
“Today ‘virtual agents’ — using conversational interfaces powered by artificial intelligence — can increasingly understand your intent … [so] machines can answer many more questions than non-machines, also known as “humans.” The percentage of calls a chatbot, or virtual agent, is able to handle without turning the caller over to a person is called its ‘containment rate,’ and these rates are steadily soaring.”
Did you catch it?
He starts by saying, “machine can answer many more questions …” i.e. an increase in coverage. But then he switches to, “… percentage of calls a chatbot can handle …”, which he correctly calls containment rate. (So, partial credit, I guess.)
Unfortunately, Mr. Friedman doesn’t cite any data or sources, so I’m not sure what his inspiration was behind this paragraph. “Soaring” is quite a strong claim! If anyone knows, please drop me note!
Later in the article, the CEO of a large BPO firm is quoted as saying that people are looking to reduce their agent headcount to 1% of current levels. Yet, data continues to show increases in the number of agents, rather than a decrease. For more on that topic, see “AI is Not Reducing Call Center Agent Employment.”
Several trends are poised to make a significant impact on contact centers in 2020. This white paper explores key areas that are sure to shake up the industry.v
- Preparing for the second text-revolution
- Making your contact center omnichannel ready
- Moving your call center to the cloud
- AI trends for the coming year
- Plus so much more