Using Speech Analytics to Measure Hybrid WFM Programs

First, you sent them home to work.

Now you call them back to the office. At least some of them.

Is there a difference in agent performance by venue?

During the early days of Covid, we provided in-the-moment coaching on coronavirus messaging.

As companies rushed to move agents offsite, we discussed how speech analytics could provide visibility into managing agents from home.

Now here we are, two years later. Many companies are wrestling with the agent location question. Should they permanently commit to locating agents at home? Or have all agents already returned to the call center? If not, should they maintain flexibility with a hybrid program? Many companies have maintained hybrid schedules post-COVID as a hiring and retention perk, so the question is more important than ever.

Speech analytics can provide the tools to inform this choice. Besides factoring in the agent preferences, here are three ways Zenylitics ‘Guided Insights as a Service’ analyzes the data to assist customers analyze performance between the two environments.

Enrich the Metadata

Tracking which calls are handled from home vs. the call center let’s you create cohorts for each locale that you can analyze. There are two ways to accomplish this. If agents work all of their time either home or ‘on prem’ you can assign them to teams, enabling the use of scorecards and dashboards to analyze them as a cohort.

If agents work some of their time at each venue, you can add metadata to identify where they reside during the call. Then you can analyze within CallMiner as you would any metadata field. If changing the metadata itself is not practical, you can export the data into Excel or your BI tool and analyze there. The key is having a way to identify which call takes place where.  For example, by tracking associating the call ANI …., you should be able to create look ups that make it easy to identify each ‘on prem’ vs. ‘at home’ cohort for analysis.

Compare Efficiency

One important dimension to review is agent efficiency. How does agent behavior differ when they work from home vs. in the call center? You analyze this by comparing ‘at home’ efficiency stats vs. ‘on prem.’ How does talk time per agent compare by locale? How do agent handle times differ? Especially telling can be scenarios in which agents work some of their time in both locales.

If handle times vary, look deeper at specific attributes of conversions. For example, within a given locale, is there more silence? If so, what is driving it? Is there more agent overtalk? If so, is the agent environment influencing the tension in the conversations?

What about background noise? One Zenylitics customer had a concern around background noise with their agents working from home. We did an analysis looking at auditory understandability issues as well as transcription confidence. We were amazed at the results. Auditory issues dropped almost 20% with agents working from home, and transcription confidence improved. Our customer was shocked. They began considering a hybrid program in their post-COVID plans, even before the pandemic had slowed.

One important caution about drawing conclusions: Be careful making inferences from small data sets. Make sure any differences actually are statistically significant.

Analyze Outcomes

Measure the ‘on prem’ vs. ‘at home’ cohorts by the desired call outcome. Are there differences in customers’ emotional states like satisfaction or dissatisfaction? How do call dispositions between location types differ? Is there a variance in success rates between ‘on prem’ and ‘at home?’

If you do find meaningful differences, you should further analyze what drives those differences. Is there greater evidence of rapport building and use of soft skills in one cohort over the other?  In a revenue-driving use case, are agents being more directive in one of setting over the other? Finding the root cause of varying performance can help identify what behaviors to coach and increase agent self-awareness about how their surroundings might influence their behavior.

One important analysis detail: Make sure the data you compare does not acoustically differ in ways that would skew the results. Does transcription confidence vary enough between the two environments that the frequency of category hits might materially differ. Identifying such differences upfront can avoid a lot of wasted effort. If environments are materially different acoustically, you may want to address THAT as an issue to ensure a better customer experience of interacting with your business.


The events of the last two years have changed this industry forever. If you aren’t considering WFH or hybrid work environments, your competitors are — and they’re recruiting your staff! Hiring is more difficult than ever. Make sure you leverage every insight available to you when making this impactful decision.



This website stores cookies on your computer. Cookie Policy