Analytics & Big Data
Dec 12, 2016

Speech Analytics – Key design principles for a successful solution

In today's digital world, contact centers are not only receiving more attention than before, but also being transformed from cost-centers to value-centers. Daily customer interaction, which generates massive data, is "the new gold" behind the push to position contact centers as proactive listening hubs. This shift in perspective has led to quite a few solution accelerators in the market.

One such solution accelerator—the idea of leveraging customer conversation (speech analytics) in contact centers to identify areas of improvements proactively—generated a lot of interest as an innovative solution. It promised to solve myriad problems faced by enterprises today. Potential solution areas included multiple business dimension, such as market share, revenue, retention, compliance, productivity, and cost of operations.

This was almost half a decade ago. Most of the solution-offering players in mining the customer voice were technology-centric companies who had workforce management (WFM), workforce optimization (WFO), and contact center technology systems as their core offerings.

Now those promised potential business impact solution failed to deliver on ROI and in many speech analytics deployment it became a sunk IT expenses. This is primarily due to a lack of continuous actionable insights to enable business impact.

Institutionalizing uninterrupted tangible insights requires an advanced analytics engine on top of the rich complex data (“the new gold").

Speech Analytics journey

In many case, clients were presented speech analytics solutions as a magic wand for contact center operations and business IT teams, capable of delivering results immediately upon deployment. The solution is positioned by the technology product companies who are interested in selling licenses by presenting the huge business improvement possibilities of leveraging the off-the-shelf features of the product to drive improvements.

Absence of intelligent advanced analytics application complementing existing standard out-of-the-box reports, graphical user interface (GUI) to build queries and categorizations on the harnessed speech data is a primary driver for poor ROI of the solution. It is the most overlooked area (intelligent analytics engine and its application), when building a speech analytics road-map.

Over the last two years, there has been a rapid adoption of speech analytics solution. Various market reports show an increase in speech analytics market, and forecast its potential growth in coming years. With lot of learnings from past implementations, and not to be blind-sided by technology capabilities and features, many clients are asking: How do I ensure that speech analytics ROI can be accelerated, or in a worst-case scenario, investment is at least breakeven? Speech analytics, which started as an attractive solution offering, turned out to be a liability, and in several implementations have not added value beyond the traditional call quality and compliance scope of work.

Therefore, enterprises are being forced to look for alternate solutions. They are seeking expert guidance on how to pivot their approach to speech analytics, shifting from an IT-centric view to a value-centric approach that enables the continuous generation of insight to improve ROI.

A value-centric approach leverages Data-to-Insight-to-Action to run Intelligent OperationsSM.

The following section lays out key design principles that need to be considered when evaluating and implementing a speech analytics solution.

  1. Pricing model: Traditional pricing model is based on number of contact center agent , a fixed price model and does not add value to enterprise in terms of business outcome and will be upfront capex investment. Changing from fixed price agent based license model to contact volume-based subscription-driven transaction model will de-risk the entire project from being a high capex item. Volume based on-demand model will enable the business objective with an ability to scale up or down basis the business result
  2. Openness of the solution: Most of the solutions available in the market do not support raw data export from platforms for further analytics purpose. (And, even if they do, they come with high price-of-professional-services cost.) It is important to own the data, and to be at liberty to apply the harnessed data for advanced analytics model building and solution hypothesis validations
  3. Accuracy and ability to tune up to speech engine to increase the accuracy rate: Audio files and streams attributes such as compression rate, stereo channel recording and file format are important to improve the accuracy. Asides these key attributes, it is important that the speech engine should have the ability to be tweaked and optimized further to include business-specific keywords and utterance. Tuning the speech engine to accommodate the geography based voice accent and periodically amending the engine to include business rules and exceptions will further improve the performance.

    Advanced level of tuning includes self-learning speech engine that can be taught to correct and learn from the training data sets and eventually leverage that intelligence to self-learn and improve the overall accuracy of the analytics engine
  4. Scalability: While contact center is going digital, still voice is the most preferred channels by consumers for complex transaction. This is followed by other digital channels such as Web, Chat, apps and emails. These digital channels including voice channel generates large interaction data that needs to be harnessed at the industrial scale. Therefore, it is important that the proposed speech analytics solution have scale, without leaving a huge IT footprint, for optimal operational cost and maintenance
  5. ROI of the solution: Above all, the ability to link the deliverables of the advanced speech analytics solution to tangible business outcomes—whether through operating cost reduction, improved customer experience, or increased revenue and compliance—is critical to determining the success of the solution. Therefore, an ROI of at least 5X over a six-month term is required for the solution to be a viable business case and compelling solution for enterprises

Key factors to be considered while designing speech analytics solution

In summary, ensuring that speech analytics solutions meet the above-mentioned design factors will mitigate the potential risk of poor ROI. This will also help enterprises drive continuous improvement initiatives, remain competitive in the current landscape, and proactively meet their end-client needs to increase market share.

About the author

Raj Kumar Subramanian

Raj Kumar Subramanian

Assistant Vice President, Analytics & Research

Follow Raj Kumar Subramanian on LinkedIn