What to look for when buying – speech analytics software

Filed under - Technology,

This month, we’ve asked three key vendors – QPC, Verint Systems and Witness Systems – for their advice on speech analytics technology.

Speech analytics – what, how and why, why, why

By Martin Blacher, director of marketing at QPC

Contact centres and the enterprises that they serve can now gain valuable business intelligence from their voice recordings because of speech analytics. This technology allows businesses to search through recorded calls and find interactions that contain references to virtually any concept, topic or word. It’s like using Google but instead of searching the web, you search through all the calls you have recorded.

Interesting, but where is the value in that? Well, although speech analytics is relatively new, and the applications are still being developed, perhaps the list below might go some way to explaining why you might want to consider it. I’ve also highlighted some of the features you should look out for:

Market intelligence

An enterprise contact centre will take about a million calls a week. In those calls are valuable insights from your customers about what they want and why they want it. Think of it as the daddy of all focus groups. If the calls are recorded for quality or compliance purposes, what you need is a way of finding the calls that are relevant to what you want to find out.

With speech analytics you can look for the names of your competitors and find out how you compare. Search for ‘too costly’ or trend the number of results returned over time and see what your customers think about your price rise. Or look for one of your product names and see what buyers are saying about it. A point to note here is that speech analytics is valuable to all parts of your business; the contact centre is just a good place to collect information as, for many enterprises, it is the most important place where they meet their customers.

One of the things to consider with speech analytics is conceptual searching rather than just word searching. While a search engine like Google looks for a word or phrase that you have typed, conceptual search engines try to understand what it is you are looking for and bring back results accordingly. Taking the example above of looking for ‘too costly’ within your recorded calls, a word-spotting search will just bring back conversations that use those exact words. A conceptual search engine, on the other hand, will bring back conversations with words like ‘high price’ and ‘very expensive’ because it understands the concept that you are looking for. Conceptual search allows you to get more meaningful results more quickly because you don’t have to consider all the ways that someone might express something – for example, dissatisfaction with pricing.


Adherence to scripts is often deemed necessary for compliance to Financial Services Authority (FSA) regulated  transactions. Some organisations also prefer to use pre-defined language for certain parts of conversations as it helps them remove uncertainty that can lead to complaints and achieve compliance to standards like Energysure, for example. With quality monitoring, only a few calls per month are typically assessed by a team leader or quality department. Using speech analytics you can search for all the calls that don’t contain certain phrases so you can identify who is not using the correct terminology on a regular basis and address the issue accordingly.

Customer satisfaction and cost reduction

Within the contact centre a lot of cost reduction and customer service improvement emphasis is now being placed on removing the need for contact by identifying what is prompting customers to call and eliminating the problems that cause these. If there is a problem in any part of your business, you can bet that your customers tell your agents first. The difficulty is how do you find these conversations without listening to every call or hoping that agents report, or log  correctly, what may seem like a one-off incident?

Speech analytics makes locating causes of customer dissatisfaction much easier. You can simply look for the concepts that we always use in complaint calls, like ‘unhappy’ or ‘doesn’t work’. This gives quality and business process monitoring teams access to the content of interactions so they can find those ten calls about a model of mobile phone that doesn’t work properly before they become 1,000 or 10,000.

One of the interesting areas to look out for with speech analytics is its inter-relationship with enterprise search engines, which look for documents across the whole organisation. These search engines, like Google for instance, search for documents that could be a customer order history within a customer relationship management (CRM) system, product information in a PDF file at a branch, or frequently asked questions (FAQs) on an intranet.

With the search example above of ‘unhappy’, you might want to look at what was written in the CRM system as well as what was said during calls to locate those unhappy customers. Some speech analytics systems can integrate with corporate search systems to look through all data within an enterprise – not just calls, websites or files.

The applications for speech analytics in improving quality, compliance, business processes, customer satisfaction, sales and marketing are already compelling and are still being developed. They are moving the value of interaction recording far beyond that produced by quality monitoring and logging alone. If you have call recording, or are planning to implement it, it is well worth considering the additional uses that your call archives may serve for both your contact centre and other parts of your business to get additional return from your technology investment.

Moving from emotion detection to root cause and beyond
By Robert Wint, EMEA marketing director at Verint Systems (www.verintsystems.com)

It wasn’t too long ago that speech analytics was just a futuristic vision. The idea of getting technology to ‘listen’ to calls and analyse thousands of conversations to unearth valuable information and trends seemed far-fetched. But today it’s a tool that is being used by thousands of companies to get to the heart of better relationships with the customer.

The best starting point when buying speech analytics is to look at what this software actually is and what key areas it covers in order to help you decide which factors you need to address in your contact centre or business. Broadly speaking, speech analytics offers the ability to review spoken ‘unstructured’ data, such as the telephone conversation between an agent and a customer, and to deduce from it information that is of value to the business – or business intelligence as it’s often called.

However, as with all new technologies, there are varying degrees of sophistication available and each method produces very different outputs and associated business values.

There are four key areas that speech analytics covers:

Emotion detection

This will identify high levels of emotion or stress in vocal patterns and it was one of the first types of speech analytics to  evolve. As you’d expect, it focuses on variances in vocal pitch and volume. However, the technology can be very limiting as there are very few solutions available that will distinguish between, say, the agent’s and the caller’s voices. Also, it can be affected by variances in line quality, as well as the caller’s natural volume and background noise.

Word spotting

This will find calls with a particular word in them. At a basic level, word spotting is used to analyse each call to determine whether a particular word was spoken in that call – words such as ‘dissatisfied’, ‘unhappy’ and so on. These calls can then be flagged for review.

The limitations with this tool include the fact that there is no context applied to the spotted word. Thus a call in which the customer says they are unhappy with your service is treated in the same way as a call where they say they were unhappy with their previous supplier’s service. Furthermore, the technology often works in a single pass format. In other words, it will look for the word ‘unhappy’ but if you then want to find the name of your product, the analysis has to be repeated with the refined criteria.


Calls can be categorised based on what was said in a conversation. This approach builds on the word spotting technique but offers a major difference. Instead of searching for a single word, the technology uses a vast dictionary of known words and indexes each occurrence of all these words in every call. Each call can then be categorised and ranked based on known words or a collection of words that are likely to occur in a particular call type.

For example, complaint calls can be categorised as having two or more occurrences of the words ‘dissatisfied’, ‘unhappy’ or ‘disappointed’. Users are then able to drill down on each call category by looking for further words. The key advantage here is that the refined search is performed on the index and does not require a re-run of the speech analysis engine.

Root cause

It is often the case that it’s the things that you don’t know about that are the things you need to look out for. But how do you find out what these things are? This is where root cause analytics comes in to play. This technique builds on the call categorisation approach and utilises analytic techniques to spot trends in the use of words within a particular call categorisation. By comparing the number of occurrences of each word in the dictionary across all of the calls in the complaint category, this method can identify higher than normal occurrences of a particular word. For example, identifying a high usage of the words ‘Internet’ and ‘password’ in a growing number of service calls could highlight a root cause relating to log-in issues with an online service. This is an excellent tool for helping to mine issues that can’t be realised at a superficial level.

One step further

When you look to buy speech analytics you will probably be looking to improve the efficiency and effectiveness of your contact centre. In order to choose the type of software that suits you, you need to decide which issues you want to address through this technology, as per the key areas outlined above.

Taking it one step further, combining categorisation and root cause analysis will give you the ultimate deluxe set of tools to help you to establish what your customers are saying and why you are not dealing with the issues as effectively as possible.

It’s impossible for a contact centre manager or supervisor to have the time to listen to each and every interaction between each and every agent and customer. You might only listen to a handful of customer calls and, if this is the case, how can you possibly spot trends that are building within your customer base? All you can do is guess what the real problems might be. And if you’re guessing, any strategies you put in place to deal with issues will be guesswork too. So you won’t be in a position to monitor whether they are having any effect.

The power of this combination will not only present you with the words but also the context in which theyappear. This gives you an overview of all calls in a report that is meaningful in a business context. It could show that sales calls are not being handled appropriately, customers are not clear about instructions for using the website, and so on.

Whatever information you glean from speech analytics, the good news is that you will be in a position to rectify issues that might so far have evaded you, such as inadequate first call resolution rates.

Once again, the key to buying the best is to be clear at the start what issues you need to address and to buy software that is easy to use and that gets to the heart of the problem you are looking to change. Good luck in your search.

Fuelling the enterprise with customer intelligence By Richard Ray, head of business consulting – EMEA – at Witness Systems (www.witness.com)For most organisations, their most valuable business intelligence still comes from customer interactions – usually through the contact centre. When used correctly, this collective data should provide a critical
insight in to the effectiveness of their sales and service strategies, as well as their marketing campaigns. Yet all too often this essential business intelligence remains locked and hidden away in separate customer databases and business process systems such as operational or billing applications.But what can you do to make sure this doesn’t happen? The answer lies in successfully integrating the enterprise with the contact centre and using analytics to gather the essential customer intelligence that is needed to improve agent skills and service levels. Below are some of the key elements to consider when you’re buying analytics software, to ensure that happens…

Establish a core customer interaction recording foundation

1) Identify the reasons for capturing customer interactions.

Every technology implementation has to have clear, specific goals. Before you’re ready to roll out your customer interaction recording technology, decide what the focus of it needs to be to benefit your business. Today, focusing on quality of service alone often does not deliver the business value expected. So decide: are you capturing interactions for compliance purposes, process improvement, up-selling and cross-selling at the point of service, or for root cause analysis? Address these questions first and you’ll better define how your measurement system should operate, which will save time and money in the long run.

2) Ensure recording objectives map to contact centre/corporate business drivers.

Your project roll out team should include employees of different levels, functions and skills; this will help make the set-up and training more effective. A pre-installation workshop will help this team ensure that your corporate, contact centre and customer sales/service representative (CSR) objectives are all linked in to the project. It will also ensure that the intelligence you gather through agent evaluations and process analysis clearly supports your organisation’s overall business goals. Involve contact centre agents, supervisors, managers, trainers and head office personnel, as well as sales and marketing, to establish a comprehensive approach. This cross-functional team can help prioritise the competencies  hat are part of the CSR evaluation process, as well as identify key measurement criteria for process improvement and revenue generating initiatives.

3) Capture voice and data.

Many companies only get half of the overall service performance picture because they only record the voice transactions between agents and customers. By capturing both the voice and data – or screen sequences – taking place at the agent desktop, management can gain valuable insight in to how effectively agents manoeuvre through CRM systems, handle
screen navigation and respond to customer needs, which helps identify training needs. Forward-thinking organisations will also try to identify areas for process and technology improvements by analysing how agents are using contact centre technology and determining if the technology is performing as expected or designed.

4) Determine how customer intelligence will be used.

The recording of customer contacts in itself does not benefit your contact centre; rather, it’s what you do with the  information that counts. This is where you have to use analytics. What do you need the recordings for: to assess CSR skills and identify training needs, or to look at first-hand customer feedback? Capture the information you need to assess an agent’s effectiveness in your target areas of improvement, and then use it to improve their skills and your processes as needed.

Then go through the same exercise from the customers’ point of view. Analytics can provide a comprehensive view of the customer and allow agents to quickly understand the customer’s situation, anticipate the customer’s needs, and effectively resolve issues. This results in a more tailored customer service and faster issue resolution – delivering the level of customer service that inspires strong loyalty. Take the final crucial step of looking at the most common areas of service or process failure and then address at least the top five areas of deficiency.

Correlate direct revenue-supporting activities with organisational customer service and intelligence gathering functions

1) Establish ‘business rules’.

Identify the interactions that are most critical to your revenue, business and customer-focused goals, and set up user-defined rules to capture these contacts. This will provide you with all the information needed to make considered business decisions. Storing recorded contacts with advanced information about the transaction – like the reason for the call or the product(s) sold – will provide quicker access and deeper root cause analysis capabilities. This makes it easier for managers at all levels to get clearer visibility of contact centre performance.

2) Identify best practices.

It is key to establish criteria for good and bad interactions across all business areas. In most inbound contact centres, agents are not ‘natural’ salespeople; usually, only a small percentage have sales experience and are comfortable up-selling and cross-selling. By identifying, recording and analysing best practice sales processes, agents can learn, emulate and improve their skills. Take advantage of today’s technology that allows you to quickly deliver customised, best practices training to your agents.

For example: a major telecommunications company recently dramatically increased sales by capturing contacts and archiving those that demonstrated best practices for up-selling. Its ability to use customer interaction recordings in a CSR training capacity helped reinforce sales/revenue generation behaviours. Cross-sales per agent per month increased from 40 in the first month of training to 57 in the third month of training – a 43% rise. Why? Agents felt more comfortable selling through a best practice approach that was successful for their colleagues. Third party or off-the-shelf training content can work, but the relevance of best practice learning from your contact centre’s own recording interactions makes in-house development more effective.

3) Demonstrate impact on the business’s bottom line.

Most importantly, contact centres can contribute to the organisation’s bottom line. Historically, call centres have been  perceived as ‘cost’ centres, but using customer interaction recording technology enables them to directly impact revenue, as well as build brand loyalty through top-notch service. These solutions spotlight how contact centre personnel champion the customer and company cause, and why they are integral to the business as the strategic hub for customer intelligence.

4) Become a ‘change agent’ – apply what’s worked in the contact centre to other areas of the business.

The practices that have been successful in the contact centre can also be used in other customer-focused areas of the  business. By using recording technology, other departments that indirectly touch the customer can capture and review samples of business processes and perform root cause analysis to determine clear action steps for improving effectiveness. As in the contact centre, the rest of the enterprise can learn how their processes impact customers, as well as how they can help contribute to positive customer experiences.

Author: Jonty Pearce

Published On: 31st Jul 2006 - Last modified: 30th Jun 2017
Read more about - Technology,

Follow Us on LinkedIn

Recommended Articles

Graph in speech bubble - speech analytics concept
5 Benefits of Call Centre Speech Analytics
computer showing analytics
Speech Analytics - What to Look for When Buying a Solution
Wooden cubes with speech bubbles linked to each other with lines.
Speech Analytics 101: What Is Speech Analytics?
An Introduction to... Contact Centre Analytics