In April Nexidia announced Workbench 6.0, which contained some key improvements to their core software development environment. The 6.0 release included a new scanning engine which can simultaneously monitor and search up to 1000 audio streams in real-time on a single server, allowing large contact centers to do away with legacy call logging equipment as they can capture and analyze audio directly from the switching environment. If tied in with the contact center’s customer interaction databases, this can potentially provide agents with guidance during a call.

The release also included automatic audio classification, which identifies specific characteristics of audio, such as the language that is being spoken, or non-speech audio activity that is present in the audio stream. The later includes things like music, background noise, or silence, for example. The importance of this is that these activities can be analyzed to help determine characteristics of different calls, such as why some calls are excessively long. For example, long periods of silence could mean that an agent is sitting there trying to figure something out, leaving the caller tapping their heels, and perhaps indicating a need for agent training. Audio classification can also segment recordings based on other audio factors such as gender of the speaker.

In this morning’s release, Nexidia added even more value for the contact center. Enterprise Speech Intelligence (like that name) 8.0 combines speech analytics with contact center data to improve first call resolution. It does this by identifying call patterns by type, and performs speech analytics on all calls by type, or all calls by customer depending upon what the contact center manager decides to drill down into for information. It creates a topic report, identifying the most common repeat call drivers. It also identifies calls with inordinate amounts of occurrences of non-talk time or high-talk time.

What do they mean by call patterns? Call patterns are repeat callers, who they are and why are they calling. For example, are there subsets of callers who habitually call back or is there some issue that makes callers call back, and if so, why and when. Probably the most important aspect of this new release is identifying the relationship between the initial call and repeat calls. Putting these two together, a supervisor can look at calls by type, pull out, for example, callers that callback after thirty days, and drill down to figure out why. An example that Nexidia gave of this was of one wireless company that was offering a one time credit to customers, however the credit didn’t show up on the first page of the bill, but was buried many pages down. This caused a huge spike in call volume when customers called back after they got their next bill. With the new analytics this was uncovered quickly and resolved by putting the credit on the first page.

Also included in this release is the ability to analyze calls by individual speaker (agent or customer). The ESI Forensic Search tools allow contact center managers to query, report on and visualize individual speaker data including agent cross-talk with the customer (which might indicate a need for agent training), or periods of non-talk, etc.

Finally, something that Nexidia prides itself on is the speed with which they can do all this. In Workbench 6.0, for example, Nexidia more than doubled the performance of its audio indexing function, indexing recorded content up to 207 times faster than real-time per CPU core, breaking their previous company benchmark indexing speed of 83 times faster than real-time. At those speeds, customers can fully index over 20K hours of content per day on a single server. With this morning’s product release, customers can now integrate multiple servers together to double or triple this capacity. This sounds impressive, but it is even more so if you see Nexidia’s comparison slide on their solution versus the rest of the industry. I didn’t do this analysis, but their slide shows that a company would need 26 servers to do 20K hours per day, or 65 servers to do 50K hours a day, compared to 1 and 2 servers with Nexidia. Your mileage might vary, I’m just reporting here.

In all, I think that Nexidia is providing some pretty solid enhancements with both of these releases.