Speech Analytics is on the verge of becoming a billion-dollar industry within the next few years and the number of applications it has in the modern contact center have grown wildly. But how did we get here? Where exactly are we now? And, where are we headed next?
Speech Recognition – a Brief Timeline
The road to speech analytics started with the development of speech recognition, a process known in today’s industry as Automated Speech Recognition (ASR) or Speech to Text (STT).
The Twentieth Century
The 1950s – Bell Laboratories develops “Audrey”, a system that is able to recognize numbers spoken by a single human voice.
The 1960s – IBM introduces its “Shoebox” technology, capable of recognizing 16 words and the numbers 0-9. By the end of the decade, speech recognition is able to identify words with up to four vowel sounds and nine consonants.
The 1970s – The US Department of Defense develops the “Harpy Speech System” which can recognize over 1,000 words.
The 1980s – Probability models are introduced that help to predict the chance of particular sounds being particular words.The number of words that speech recognition technology can identify rises to several thousand.
The 2000s – Google introduces voice searches and is able to collect voice data from billions of its users. Accuracy of prediction and transcription rises dramatically.
The Dawn of Speech Analytics in the Early 21st Century
Speech recognition is one thing, but it wasn’t until The U.S. government got heavily involved that the days of speech analytics began. The catalyst for this was 9/11, after which intelligence agencies began using speech recognition and indexed audio to spot keywords and phrases that might indicate security risks!
Around this time two new developments also took place – namely the indexing of audio based on phonemes rather than whole words (which helped to increase accuracy) and Large Vocabulary Continuous Speech Recognition which used dictionaries to allow for a larger recognized vocabulary and even better transcription accuracy.
Now, the technology was really getting somewhere.
As we look at how speech analytics is utilized in a contemporary contact center environment we can see that these foundation technologies, lile Phonetics, LVCSR, ASR, and STT, are all still in play.
The list of applications that contact center managers have found for the technology is growing all the time, currently including:
- Quality Management – using 100% call auditing and automated scorecards to exponentially scale up the ability to monitor call quality and agent performance.
- Risk Mitigation – screening every call for risk, including the use of mandatory statements, proper greetings, escalation language, and more
- Agent Coaching – using aggregated insight from 100% of an agent’s interactions to provide objective, data-driven coaching.
- Data and Insight Generation – using conversations analytics to understand customer intent and experience on a scale that wasn’t before possible. Using this insight to inform business decisions across departments – not just in the contact center.
On top of these established use cases, more functions and uses are being layered on to help give a more and more detailed and holistic picture of what is happening in the contact center. We’re talking about things like emotion detection, sentiment analysis, talk-over analysis, and real-time analysis.
It’s real-time analysis and alerting that’s been most in demand lately, allowing for calls to be analyzed while they’re in progress. As with post-call analysis, this functionality is great with scripting adherence, risk mitigation, and regulatory compliance.
Artificial Intelligence is another new development, the most common application being “Natural Language Processing” (NLP). NLP takes the unstructured text and transforms it into normalized structured dialogue, which allows for cleaner, richer analysis along with the ability to drive machine learning. Think of products such as Siri, Alexa, and Cortana.
Some of today’s contact centers are using AI for predictive analysis. This is helping them to predict the outcome of calls, predict agent or customer behavior, and better understand the customer experience.
Additionally, more speech analytic vendors are branching out into text analytics. In the age of digital content and social media, we are seeing a paradigm shift with more chats, emails, text, and social media interactions being brought into analytic solutions, so that they too can be categorized, scored, and coached on. This is beginning to give organizations a comprehensive, omnichannel view of what’s going on in their contact center.
What Will Speech Analytics Look Like in Contact Centers in 2025? Our Predictions:
- Analytics becomes a still-growing billion-dollar industry.
- It’s progressed from a “good-to-have” solution to a “must-have” for all contact centers who want to compete in an economy where customer experience is the key competitive differentiator and strict consumer protection regulations are many.
- Analytic vendors have added all the latest and greatest innovations allowing their clients to collect the richest data with which to drive organizational decisions.
- The use of prosodic data is now the norm with most vendors. This analyzes the acoustic attributes of voice recordings, such as tone, pitch, pace, emphasis, etc. With this, clients can get far greater insight into call sentiment and can easily spot and address bad customer experiences or agent misconduct, or, conversely, agent best practice and delighted customers.
- Companies are using AI-driven behavioral prediction to identify any gaps or problems in their products, marketing, back office, and customer experience (CX).
- Vendors have begun adding survey ingestion into their platform allowing their clients to automatically incorporate the results of voice-based post-call customer satisfaction surveys into their performance analytics platform.
- Machine Learning (ML) will continue to advance contextual accuracy allowing callers to have a loyalty score based on their interactions with the organization. This will be an instrumental metric for many organizations looking to improve their customer experience.
- Time to ROI has decreased dramatically with many vendors providing pre-configured categories, scorecards, etc. out of the box.
- ROI has increased as organizations are cutting costs on analytical teams and resources typically needed to spin up analytic platforms. This is due to the AI capabilities along with Machine Learning. Data is readily available following implementation.
The Future of Speech Analytics Looks Bright
So, of course, these are just our predictions, but based on the level of investment and competition that is taking place in the speech analytics space right now the future of the industry looks incredibly bright.
As we continue to be more technically advanced, we will continue to see innovations in speech analytics that will give us better, more accurate data faster than ever before.