Guy Kawasaki, Apple’s former chief evangelist, once said , “Ideas are easy. Implementation is hard.” His boss at Apple, Steve Jobs – who I had mentioned in my earlier blog – would’ve agreed. It’s great that airlines are investing in speech analytics solutions for their contact centers. The potential benefits are just too irresistible. However, are they drawing out forward-looking plans for implementing these solutions? The benefits will follow only if pragmatic implementation plans are in place. However, this is easier said than done.

First, let’s debunk a couple of myths around speech analytics.

  • It is plug and play technology
    • Not quite. Speech analytics solutions need to customized. Moreover, they need to be continually fine-tuned to meet evolving business needs.
  • It can work wonders on its own
    • Well, not really. Airlines should proactively invest both time and resources to administer, analyze and regularly ‘fine-tune’ the system.

Now, let’s take a look at the implementation of speech analytics in airline contact centers. There are four main stages in the implementation – Call Injection, Fine-tuning, Discovery and Reporting.

Speech Analytics

Call Injection involves feeding conversations from source systems such as call recorders into the speech engine. The associated metadata — agent’s name, customer’s basic information, day/timestamp — is also fed into the engine. The speech engine converts audio (including acoustic signals such as agitation or noise) into the text using speech recognition.

Fine-Tuning is an ongoing process that involves building specific call categories in the call library by identifying relevant keywords or phrases. Once the call categories are created, fine tuning is conducted to ensure that the tool correctly segregates calls based on defined categories. Sample calls are analyzed manually and validated against the analytical tool’s findings.

Discovery involves the system automatically analyzing customer interactions and assigning them to pre-defined categories. Key metrics are used to measure various performance indicators such as agent quality, customer satisfaction, emotion and first contact resolution. The insights derived from this stage are used to create actionable reports for relevant teams.

Best Practices for Implementation

It’s important to deploy a team of experts equipped with sound business knowledge to manage the speech analytics process. Equally important is the attention to details during implementation.

The following stages should be carefully planned and executed.

Speech Analytics

Building a Speech Library: A content audit to build a comprehensive speech library is the first and most critical step in implementation. This requires listening to hundreds of calls to gain insights into the real ‘voice of the customer.’ The audit helps capture the actual phrases used during customer interactions. These inputs can be used to create relevant categories in the call library.

Setting Confidence Level: The confidence level for a keyword or phrase reflects the degree of certainty required of the speech engine to recognize them. Confidence levels can be set for each word or phrase in the speech library.

Auditing and Fine-Tuning: Ongoing auditing involves listening only to that part of a call containing the identified phrase. This exercise is carried out to compare actual words spoken during the conversation with words identified by the speech engine. In case of an erroneous identification, the system is fine-tuned to avoid such errors in the future.

If airlines plan meticulously and execute the implementation skillfully, speech analytics will give them the required return on investment within a short period of time. A third-party vendor with the right capability and domain expertise (not just technology skills) can help airlines in this process.

Learn more about WNS’ travel industry analytics solutions, delivered using a deep domain expertise to deliver a profound bottom-line impact.

For additional information on this topic, please read our previous blog post about speech analytics.

Join the conversation