Best Practices for Realizing a Rapid ROI from Speech Analytics



A large number of organizations continue to believe the vendor hype that speech analytics is simple, despite the many articles and analyst reports to the contrary. It's not easy if you want useful and actionable results. It may not be overly taxing to use a speech analytics solution to find key words and key phrases, but they are of limited use unless you understand the context of these words and are able to identify correlated terms and concepts and the interpret them correctly. While it might be interesting to see the number of times a competitor's name is mentioned by customers in a defined time period, for example, it is much more useful to understand why that rival is suddenly becoming a topic of conversation.

Speech analytics is and should be treated like any other business intelligence (BI) application. It should be staffed by analysts, quantitatively oriented employees who are comfortable analyzing numbers and finding business trends. It should not be staffed by contact center quality assurance experts or other people who are not accustomed to working with numbers or using automation to conduct sophisticated Boolean-based searches. Analysts are trained to employ broad-based thinking to correlate statistical results with what is happening anywhere in the organization and throughout the business world at large.

Another major misconception about speech analytics is that it can be managed on a part-time basis. This is not a good idea, as a part-time manager will not have enough opportunity to properly set up searches or perform the time-consuming iterations required to optimize findings and avoid false positives. (Some calls are mistakenly believed to be relevant, but upon further review--when people listen to actual call snippets--prove not to be.) In fact, an essential operational best practice is to conduct a search, sample it to identify the percentage of false positives, and if it's too high, modify and rerun it. It can require as many as 20 to 50 searches to compile a good, clean data set for each business issue. This will take days, particularly while learning to use the application. This methodology is very common for BI solutions, but is not typical for contact center applications, which are expected to deliver usable findings immediately.

Companies that follow these best practices have a much better chance of realizing a strong and rapid return on investment from their speech analytics implementation:

1. Create a separate speech analytics team with a minimum of two resources; the number will vary based on the needs of the organization and the volume of data and searches.

2. Staff the team with quantitatively oriented business analysts who are highly experienced in analyzing large volumes of data and are comfortable using an analytics application to conduct searches.

3. Thoroughly train the speech analytics analysts to use the application and to understand your business. A good speech analytics analyst should be able to identify issues and provide a subtle and nuanced interpretation of the underlying business drivers.

4. Require all speech analytics analysts to spend at least one week listening to calls prior to formal training, so that they have context when trying to interpret what is happening in phone conversations.

5. Ask analysts to vet their findings with relevant department manager(s) prior to sharing the information with the rest of the company. This will improve the overall validity of the findings.

6. Start small; do not try to do too much all at once with speech analytics. Allow the analysts to concentrate on two or three issues at a time.

Besides selecting the right application, companies that are succeeding with speech analytics are investing in highly trained staff resources and refining the analysis process on an ongoing basis. Speech analytics is a sophisticated enterprise application that can provide valuable and actionable findings and insights to enterprises, but it takes a great deal of effort and the right people to make it work. For help in selecting the right speech analytics solution for your company and to see more best practices to help you succeed with your initiative, see DMG's "2012- 2013 Speech Analytics Product and Market Report".


Donna Fluss, president of DMG Consulting LLC, is an expert on contact center analytics and the back office. She has 30 years of experience helping organizations build contact centers and back office operating environments, and assisting vendors to deliver competitive solutions. She can be reached at Donna.Fluss@dmgconsult.com.