Speech analytics? Its audio transcription with text searching, right ?

For the third time in as many weeks we had a conversation with a prospect that went along the lines of:

#1 “We looked at speech analytics a while ago, and we get it, “check 100% of calls, automate agent scoring, compliance checking, coaching effectiveness, discover new customer insights, improve the customer journey & the customer experience”.

#2 “Loved the idea in principle, but it turned out to be pricey and complicated”

#3 “Anyhow we think we can do this in-house, quicker and cheaper.  We can we push the audio through Google and run a search on the transcript”.

Transcription vs. speech analytics

Transcription services have improved hugely in accuracy and speed, and costs have come down dramatically over the last few years, making on demand transcription very accessible, such that generating an accurate transcript is straightforward & inexpensive.

So far so good, and useful if you prefer to read text rather than listen to a call, with the benefit that you can often zip through a transcript and find relevant points far quicker than wading through a whole recording.

To create meaningful analytics to achieve the use case benefits quoted above, could just involve a word  search of the text. However, and this is the crunch point, whilst searching for words and phrases does of course offer some value, speech analytics is far more than “transcription + word searching”.


To draw on an example, checking for script compliance, particularly in regulated environments such as financial services, can be a key requirement for every call.

Whilst finding an important word or phrase is perfectly possible, firstly, how laborious would this be as a manual exercise, over multiple call and agents, and secondly, even if the search process is automated, how much meaningful insight would you derive from word searching taken in isolation?

Speech analytics, at speed, can assess virtually any number of calls, in near real time and deliver the results users are looking for, against their pre-determined criteria. 

Data is good – more data is better still !

This is where speech analytics really comes in to play. Not only key words and phrases but other useful indicators can be identified to deliver deeper insights. For example:

  • The positioning of words or phrases at a specific point in a call
  • Gaps in the call, silences, one party talking over the other and so on.
  • Sentiment measures such as anger, disappointment, stress.

In addition, meta data such as customer and campaign details, agent information or call disposition codes, can be added to the transcript to provide deeper insight for each call, by agent, across a team or a company as a whole.

And whilst we are on the theme of “data is good and more data better still !”  drawing in other data sources, such as from a CRM or invoicing system, can deliver a more detailed picture,  for example when undertaking discovery or problem analysis.

The upshot

So what does this mean in practise?

Pulling together the multiple data “pools” mentioned above and running pre-defined algorithms means that complex searches to identify and classify call outcomes become possible. There are many and varied uses including:

  • Call scoring – how well did the agent “hit” all the key factors we want to measure?
  • Immediate alerts for significant breaches or problems.
  • Pick up important calls that might need attention, for example hot leads, cancelling customers or high value customers.
  • Target coaching to an agents specific needs.
  • Provide regular feedback so agents can “self-police”.
  • Link agent feedback to engagement and incentive programmes to drive best behaviour and long-term change.


if speech analytics is on your “to do list” we would be happy to discuss how our services could support your business. Whether its lifting sales, improving agent performance and retention, understanding how customers interact with your brand or a raft of other uses, we are always happy to chat!

Andy Andreou

Sales Director