Customer contact analytics as-a-service

Most contact centres manage a range of customer contact types, such as calls, chat, email and social media. These customer conversations contain a wealth of analysable data that can be used to drive out costs, increase revenue, manage compliance and improve customer experience. It is now possible to buy “analytics as a service”, combining analytics software with human smarts.

Contact centres record and listen to calls for a variety of reasons – to spot-check for compliance with regulations, processes and standards; to assess agent performance and identify customer service training gaps; and to tease out insights into customer behaviour and preferences.

Because it unlocks this valuable operational and customer insight, analysing calls is one of the most profitable activities a contact centre can do. Unfortunately, limited resources means that QA and supervisory staff are only able to listen to 1% to 2% of calls, so that insight remains forever hidden in the remaining 98-99% of calls. Current regulatory pressures are mandating firms to monitor and review a greater percentage of customer contact. AI enables automation of a number of these processes without increasing labour costs.

Now, advances in technologies such as speech-to-text, NLP (natural language processing), and machine learning enable contact centres to analyse 100% of customer contacts.

Intelligent QA – humans and AI working together

Humans are, of course, much better than machines at picking up on the context-sensitive linguistic, social, and commercial cues in conversations between your agents and customers. The human brain is trained to understand that context and tease out the causal factors which explain why people do the things they do, and what informs their decisions.

No machine can truly understand everything, however machines can be trained to seek and identify patterns. By deploying analytical QA software, the QA staff score and analyse a statistically relevant sample of contacts that result in a rich, substantial data set covering agent and business intelligence.

During the ‘human’ review, calls are tagged for specific use cases like pressure selling, mis-selling, a good / bad customer experience, a first time resolution – and so on. After modelling this data, you will have a matrix of indicators specific to your own operation that can be used to automatically generate scores and flag up issues for every call.

What you can do with speech analytics

Analyse a huge volume of calls, in near real-time, to provide immediate feedback to operational teams and flag up compliance or QA issues as they occur.
Identify the financial and reputational risk of bad practices like non-compliance and mis-selling.
Identify high and low performing agents, and understand what separates one from the other.
Lift ‘average’ performance by identifying skills, knowledge and CX gaps, then patch with training and process changes as well as monitor the improvement.
Improve speed to competency for new recruits and reduce attrition rate.
Improve customer satisfaction and other key metrics, reduce customer churn, predict future behaviour and the impact of potential changes.

Insights as-a-service

Until recently computationally intense AI processes that require lots of data storage were only available to larger organisations with significant budgets. The rise of cloud computing has changed all that, and customer contact analytics can be provided on an ‘as-a-service’ basis.

The best providers don’t just bring software and machine learning algorithms to the table, they have data specialists and experienced contact centre analysts that you can call on to help build your training data set and the analytical tools you will use to mine call recording transcripts.

Do be aware that some managed service vendors are not based in the countries they sell to, and offshoring data is something that can affect compliance with local privacy legislation and PCI-DSS.

Four levels of QA

Manual QA	QA staff and supervisors listen to and score a small sample of call recordings.	Pick up a small % of compliance, QA, and mis-selling issues, plus ad-hoc customer feedback.
Analytical QA	A combination of human and machine analyse a sample of calls.	Score calls and dive deep to identify business intelligence and root causes behind the metrics.
Automated QA	AI trained to listen to high volume of calls, can even be in near real-time.	Automatically score all calls in near real-time, flag issues as they happen, gain deeper insights into customer and agent behaviour.
Predictive Analytics	Take human and machine post-call analytics, combine with customer data, to predict future customer behaviour.	Predict the next contact, NPS score, customer sentiment and demands, and how likely the customer is to churn.