Congratulations! You Just Bought Speech Analytics! Now What?

“Purchasing a speech tool is like hiring someone to remodel your house only to discover that he can also repair your dented fender, prepare your taxes and whip up gourmet dinners for your family.”

David Patchen, Partner, Maintrax
David Patchen, Partner, Maintrax

If your city’s police chief recruited officers and trained them effectively, but failed to deploy them appropriately or manage them in a disciplined, strategic way, the public’s confidence would undoubtedly go down while the crime rate went up.

So it is with speech analytics. You may have done your homework in selecting just the right vendor and in following best practices for implementing your new speech tool. But if you haven’t given proper consideration to how to optimize deployment of your new technology, management will quickly lose confidence in it—and it would be a crime to waste all that potential.

The following seven post-deployment tips will help you to police your speech initiative and allow you to protect (your organization’s investment) and serve (your key stakeholders).

1. Develop a speech analytics subject-matter expert

Many organizations now understand that speech analytics isn’t a plug-and-play solution like other contact center technologies. Still, many businesses fail to designate the right individual to manage the technology once the deal is inked.

Contact centers often assign this task to someone in operations, typically on the telephony side, but in our opinion, the person given this responsibility doesn’t need to be tech savvy. More important is that they should be solutions-driven and business- curious. Why? He or she is likely to answer to a wide range of internal customers.

Dan Wilson of the Skipton Building Society had the foresight to recognize this distinction from the onset. Skipton, a Yorkshire-based organization, is the U.K.’s fourth largest building society, with 220 customer-facing agents who handle about 96,000 inbound/outbound calls a month. Wilson serves as Skipton’s business improvement leader, overseeing an internal team of consultants comprised of speech analytics, telephony, communications and business process specialists.

To make the most of Skipton’s investment in speech analytics, Wilson brought on Nicholas Boon as a full-time resource tasked with “owning” the tool. Boon’s first few months were spent becoming intimately familiar with the technology and experimenting with it to learn its strengths and flaws.

He then focused on expanding the list of phrases in the speech library, testing them against calls to determine the hit accuracy, making tuning adjustments to improve detection, summarizing the resulting business intelligence, and building reports and visual representations of the data for dissemination throughout the organization.

Skipton gained some quick wins along the way, which was important for achieving enterprisewide buy-in. Even so, Wilson stresses the importance of taking a long-term view. In his estimation, it took nine months before the tool was operating as it should. “You need to put in the time and effort to investigate and tune,” Wilson says.

Absent that investment of time, energy and focus, Wilson believes that Skipton may have found itself in the same position as another U.K. call center that purchased and deployed the identical tool but didn’t derive much value from it because they hadn’t dedicated resources to optimizing it.

2. Create a “closed-loop” feedback system

It’s imperative that your anointed subject-matter expert doesn’t exist in a vacuum. Optimize the impact and ROI of your speech tool by disseminating the actionable data it produces with key stakeholders throughout your organization.

One of Wilson’s original intentions was to use Skipton’s new speech tool to identify calls from vulnerable customers—those who required additional support due to disabilities, stress, relationship breakdowns or financial difficulties. Identifying and reviewing those calls “really changed the way we look at vulnerable customers and how best to serve them,” Wilson acknowledges.

However, it didn’t take long before other departments within Skipton got wind of the amazing new technology that was revolutionizing call center operations. Once the word got out, says Wilson, “it spread like wildfire.”

Before long, Wilson and his team found themselves answering to an ever-widening number of internal customers across the organization. To ensure that each of those 10 to 15 stakeholders’ needs are being met, Boon consults with them on the front end to completely understand what they want to achieve, and meets regularly with them on the back end to make sure what’s being delivered is providing the necessary intelligence. That feedback loop allows him to make adjustments that increase efficiency and provide even greater value.

Post-Implementation Pitfalls


  • Failing to engage key stakeholders throughout the organization who will directly benefit from the business intelligence produced by speech analytics.
  • Not appointing project champions with the authority to carry out cross-departmental change or the skills to raise awareness and foster enthusiasm across the enterprise.
  • Failing to accurately define the scope of the project, or to benchmark relevant changes before and after implementation.
  • Racing out of the gate. New speech users should start slowly and take the time to get more comfortable with and knowledgeable about the speech tool.
  • Juggling multiple long-term objectives from day one. Look for quick wins so you can point to a track record of success when engaging in internal debates.
  • Choosing success criteria which are not measurable. Prior to implementation, it’s crucial to factor in your business rules and objectives, define the endgame and determine how to evaluate results.
  • Choosing projects that do not promise to deliver significant business change.
  • Treating the speech initiative purely as an IT project instead of assigning business analysts to define objectives, follow through on project delivery and effect organizational change.
  • Failing to put change management processes in place that are appropriate in size and scope to the insights emerging from the analytics solution.

3. Employ “change agents” who are receptive to new ideas

Any organization considering a speech initiative needs to ensure that it is open to change and that change agents—people with the responsibility and enthusiasm to do what needs to be done, regardless of where it takes them—are strategically in place.

That’s why the choice of whether to implement customer contact analytics should not be left to the IT department, or even to the business users except perhaps in clearly defined cases of QA automation or compliance monitoring. Without executive-level engagement, your speech initiative will be unduly vulnerable.

Fortunately, solution providers are beginning to offer executive-level consulting programs that help companies prepare for change by restructuring their processes and strategies to take advantage of potential insights and outcomes produced by analytical findings.

4. Be ready to go when it’s go time

Data, no matter how rich and revealing it is, is worthless unless acted upon. If a business isn’t willing or able to act upon the insights produced by speech analytics, there’s no point in even taking the first step.

We worked with a prominent health care insurance provider to identify key segments of its customer base that were most likely to churn. However, after investing in a sophisticated speech tool and three months’ worth of professional services, the insurer’s managers allowed themselves to be distracted by other priorities and brushed aside data that could have saved them countless customers and many thousands of dollars.

Watching valuable business insights get swept away by the winds of indifference can be frustrating. Wider business insights, which are likely to cross over several departments and “fiefdoms,” are often at even greater risk, especially if the organization is plagued by political infighting and competing priorities.

When actionable business intelligence actually gets acted upon, the impact can be profound. A client in the cable TV industry purchased a speech tool to identify at-risk customers who had given subtle indications they were likely to defect. Given the potential ROI, the company assembled and tasked a team of loyalty experts to place outbound calls to high-risk customers within 48 hours of the initial customer call.

The initiative was extremely successful. Of the at-risk customers reached live, 20% were considered by agents to be irrefutably “safe” with the revenue stream protected. Of these, 60% were retained by extending a special offer, while 40% were retained based on the resolution of a technical issue.

Fully three-quarters of the rest of the at-risk customers who were reached were deemed by agents to be safe by virtue of the courtesy call itself. Many customers who appeared to be on the verge of disconnecting expressed gratitude for the company’s attentiveness and sensitivity to their situation.

Thanks to executives who made this speech initiative a priority and had the foresight and commitment to see it through, a potentially dire situation was transformed into a vibrant success story.

5. Don’t clutter your speech library with phrases that don’t provide actionable business value
The best advice I can offer new speech users can be distilled down to one word: discernment.

Before investing time, money and other precious resources in deploying your new speech tool, weigh the potential value of each bit of knowledge you hope to discover.

Analyzing every call to ensure that agents are using proper language sounds good in theory. Too often, however, speech users expend their energies on analyzing common elements of the discussion that won’t offer commensurate value. I’ve seen organizations make a concerted effort to analyze their agents’ exact greetings only to conclude, “OK, 94% are using the proper greeting. But so what?”

Setting precise objectives that promise to yield high-value business intelligence is like looking for $10 bills instead of pennies. You can certainly use your speech tool to find an unending supply of pennies, but it’s not a productive use of your time or the tool’s capabilities.

Besides, generating a large amount of data will only be useful if you have the extensive resources to review, interpret and analyze it all. The more you can narrow your focus, the more valuable the results.

On the other end of the spectrum, I’ve seen organizations waste countless hours preparing for the occurrence of highvalue phrases that are so rare that the value they provide doesn’t justify the expense to find them.

For instance, a health care company that we worked with was determined to categorize each call by medical condition, and “doula” was included among the phrases that would identify pregnancy-related calls. The effort was eventually abandoned after the organization came to the conclusion that:

  • Phonetically speaking, “doula” is a difficult word to accurately detect;
  • “Doula” is a term that’s rarely used by callers; and
  • There was a stronger likelihood that the caller would use a more common term that would reveal the nature of the call, such as “pregnancy,” “maternity” or “due date.”

6. Listen to calls to better understand the authentic voice of the customer

If you forge ahead on a speech analytics project without listening to random recordings from the appropriate business queue, you significantly lower your odds of producing useful business intelligence.

Guessing about “the voice of the customer” is not a suitable substitute for actually hearing it. Listening to hundreds of actual calls from start to finish is a tedious process, but it’s the best way to identify the meaningful keywords and phrases that determine customer intent and predict outcomes.

Take the marketing director of a credit union that had just deployed speech analytics. He was disappointed to learn that very few members were bringing up the loyalty program he had recently launched. After doing some investigating, we found that customers were indeed mentioning the program with great frequency. It’s just that customers were largely ignoring the official package names that were incorporated within his speech library and instead were asking about “perk points.” It’s a good example of the gap that can exist between expected and actual customer language, which can best be revealed when operations and marketing departments cooperate.

The quality of your speech library will make or break your speech analytics initiative. It’s vital that you build a customized speech library and then apply sophisticated methodologies to continually fine-tune search results until you reach the sweet spot of accuracy that’s thoroughly aligned with your business objectives.

A collections agency was about to declare their new speech tool useless because the tool had identified only 10% of the recordings their agents had manually flagged at the time as “Do Not Call.” It turns out that the agency was relying solely on the canned, out-of-the-box “Do Not Call” phrases that came with the software. By taking the time to understand the language their customers were actually using, they greatly expanded the number of relevant phrases in their system and significantly increased the number of DNC captured.

Nicholas Boon from Skipton can attest to the power of truly understanding the voice of the customer. He began to reverse- engineer calls that were known to contain certain topics to better understand the phrases that customers were actually using. “It’s improved our results tenfold,” he says.

Developing and Deploying a Speech Guru


  • Executives who wish to use the tool to keep their finger on the pulse of the business.
  • Quality management supervisors who want to monitor their teams with greater efficiency.
  • Sales VPs who wish to employ the best practices used by the most successful agents.
  • Marketing science administrators who want to better understand the customer experience.


  • Performing a “content audit” of calls to benchmark how often particular phrases actually occur.
  • Auditing data on a semi-regular basis to ensure that results are delivering actionable business intelligence.
  • Adjusting settings to reach the optimum level of accuracy.
  • Building algorithms and rules to improve the precision of the intelligence being produced.
  • Adding and testing new phrases relevant to the issues at hand and removing phrases that don’t provide meaningful insights.
  • Conducting “live control tests” to assess how phrases are likely to perform before applying them to actual recordings.
  • Creating reports and developing a system to disseminate them to appropriate stakeholders within the organization.

The responsibilities of a speech guru may not require a full-time position; at the same time, we’ve worked with companies that have hired two or more people to manage the initiative. Given the importance of the function, it’s a smart play to establish redundancy in the event someone leaves.

Ultimately, when it comes to speech analytics, the better the effort, the better the results.


7. Keep an open mind

Purchasing a speech tool is like hiring someone to remodel your house only to discover that he can also repair your dented fender, prepare your taxes and whip up gourmet dinners for your family. Once you start using your speech tool, you’ll be amazed at all the things it can do.

Dan Wilson of Skipton notes that speech analytics “opened up new avenues of investigation” his company hadn’t considered, such as compliance, financial crime and fraud protection. It’s imperative, he says, “to keep an open mind” about all the management information that can be gleaned from a speech tool.

Although it wasn’t on Wilson’s radar when Skipton invested in speech analytics, he and Boon have since used it to better understand the root cause of customers’ calls. The resulting insights not only provided agents with hints and tips on how to better communicate with customers, they’re beginning to drive strategy related to Skipton’s transition to nontraditional communications, such as website interactions, text messaging and video conferencing.

The value that speech analytics has added to Skipton comes at a critical time. After the Brexit vote, the financial market in the U.K. has become even more competitive due to lower interest rates generating even lower margins. “The one area where we can really set ourselves apart is customer service,” Wilson explains.

Closing Thoughts

Savvy speech users discover that, post-deployment, the learning never ends and the benefits keep accruing. The more familiar you become with the features and functions, the better you can leverage both the art and the science of speech analytics to deliver actionable—and ongoing—business intelligence.

David Patchen, partner of MainTrax, is a highly respected independent voice in the speech analytics industry.

– Reprinted with permission from Contact Center Pipeline,

Related Articles

Back to top button