How Accurate is Conversation Analytics?
Part of our job on Blog.LogMyCalls.com is educating the marketing and sales universe about Conversation Analytics. Last week we wrote a blog post called Does Conversation Analytics Actually Work?
In this blog post we'll answer a similar question: how accurate is Conversation Analytics? My guess, when someone asks this question, is that they want to know what percentage of words our speech recognition technology gets right.
The short answer to that question is about 80% to 85%. But the long answer is that it doesn't really matter. Here are a few reasons why.
When you or I get on the phone and talk to someone we generally hear between 80% of and 85% of the words spoken on the call. And immediately after a relatively short conversation with my wife (say 45 seconds or less), I could probably only transcribe 30% of the words on the call. I simply won't remember everything that was said.
However, I still almost always get the gist of the call.
I know how excited my wife was to talk to me, what she wanted, action items from the call, and what I'm supposed to grab at the store and bring home for dinner.
But...wait. How is this possible when I only remembered 30% of the call immediately afterwards?
It's possible because the words only matter to the degree that they're in context.
Conversation Analytics analyzes calls and extracts dozens of already-baked indicator metrics. These indicators are everything from lead score and sales readiness of a caller, to conversion data and prospect follow-up action. Conversation Analytics analyzes literally thousands of pieces of data for each indicator. The algorithms are extremely complex and powerful.
Thus, it doesn't really matter if words are missed.
Aoccdrnig to a rseearch taem at Cmabrigde Uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are, the olny iprmoatnt tihng is taht the frist and lsat ltteer be in the rghit pclae.
Even though the letters are in the wrong place, you still know what the paragraph says. Your brain fills in the blanks.
That's the same way Conversation Analytics works. It looks at the whole of the conversation to derive a conclusion. So, if it misses a few words here and there (which it will), it really won't make a difference.
Even though the algorithms and the phrase combinations Conversation Analytics is looking for are complex, they're also quite specific.
For example, Conversation Analytics could even miss a majority of words on a call, but still hear the word 'Visa' followed by several numbers and then the phrase 'expiration date' followed by a few numbers and easily derive that a transaction took place. It would know the caller purchases even if those were the only words on the call that Conversation Analytics got correct.
by McKay Allen, Inbound Marketing Manager