How can Call Centre Data Analytics Improve Performance

How can Call Centre Data Analytics Improve Performance

Call Centres are a rich source of customer data and with the right approach, can lead to useful analytics insights. They amass a large volume of unstructured data which is generated from recorded voice calls and typically retained for audit purposes. A call centre is also often a critical touchpoint to resolve customer issues and identify cross-sell opportunities.

In many cases, these actions depend on having timely access to relevant customer data. Because of this, the ability to positively influence call outcomes is significantly increased when real time analytics are combined with sources of customer data brought together from across the organisation and presented in real-time. 

The increasing availability of cloud storage and compute for unstructured data is opening up opportunities for businesses to make better use of their call centre data for analytics and improve call centre performance. The ability to predict churn and make better cross-sell recommendations is increasingly straightforward to deploy in complex unstructured data environments.

How to Improve Call Centre Performance

Process Compliance

Some industries have strict rules about disclosures in a phone call. For example, according to the Australian Telemarketing and Research Calls Industry Standard 2017, a marketing caller must provide the following information as soon as the call starts: their name, their company name, the name of the person they are calling and the purpose of the call. Failure to adhere to regulation can lead to blacklisting or penalties which are very much in the interest of operators to avoid

There may be other script-based standards as well as general data collection needs about the call that need to be tracked. Additionally, there might be blacklisted phrases, jargon or slang which are discouraged from use. Keeping track of certain phrase usage can be a good way to gauge the uptake of a new script or set of call standards. 

Converting audio to text transcripts allows detection of key phrases which are useful in providing qualitative analysis of process compliance in call centres. These are linked to automated alerts and notifications which can be used to help understand compliance issues and help develop a strategy to address them. 

Quality of Service

Evaluating performance can often subjective and done on an ad-hoc basis involving a supervisor listening to randomly selected calls and judging them based on predefined evaluation forms. One review of relevant research highlights that evaluation is highly influenced by subjective factors such as perceived past performance and social ties. Additionally, poor sampling during this ad-hoc process can produce unrepresentative results.

Leveraging unstructured data analysis for audio can overcome the sample bias of qualitative supervisor reviews and automate the monitoring process, providing aggregated, quantitative assessments. 


Traditionally, this has been the best serviced area of call centre analytics, where data on queue times, call times, wrap times are required to manage service level targets down the operator level.

Unstructured data can help better differentiate the individual load on each operator each day by helping to understand the difficulty of each case and customer. This can be integrated into more queue management and smart customer routing in order to maximise service to callers flagged as priority.

Customer Satisfaction

Traditionally organisations rely on NPS surveys to assess the level of satisfaction that a customer has with a call centre interaction. This inherently suffers from sample bias and provides limited insight into the specifics of the customer scenario.

Leveraging emotion detection from voice recordings allows automated detection of sentiment in order to get a gauge on the customer’s interaction, potentially providing an additional vector through which to gauge satisfaction.

a desktop on a table telephone connected to a call centre analytics system

Key Techniques in Call Centre Analytics

Audio to Text Conversion

Creating text transcripts based on voice calls can now be created cheaply and at scale through a number of cloud-based services. These include Azure and Google Speech-to-Text, which can be deployed flexibly with a pay-by-the-hour of call time model. The text files which are created by this process can then be efficiently loaded into a Data Lake storage solution, complete with all necessary metadata, ready for further processing.

Intonation Analysis

There are a number of ways to perform sentiment analysis on voice call data, but the main two involve acoustic feature processing and semantic feature processing. Acoustic feature processing is language agnostic and does not rely on language semantics. It relies on tonal features of the voice, such as pitch and energy, analysing it against voice datasets that have been previously annotated to identify angry or otherwise emotionally volatile moments of conversation.  

Sentiment Detection from Text 

The second main method of performing sentiment analysis is semantic feature processing, which uses the output of the speech-to-text and processes it for relevant cues based on the type and frequency of words used. The effectiveness of sentiment analysis goes up when a hybrid acoustic/semantic model is used. We have previously written about the basics of NLP sentiment analysis and how it is deployed for opinion mining.


Storing and parsing large volumes of unstructured data produces a wealth of hitherto unavailable metrics and insights. However, it may be hard to know what to specifically do with it and identify how to use it in making productive yet unobtrusive changes. A system which automatically processes real time data feeds and creates automatic alerts based on pre-defined rules is a good way to take quick, data-driven action based on objective metrics without being beholden to performance dashboards and scorecards.

a cell centre worker sitting at a table in a call centre analytics environment

Call Centre Analytics for Predicting Customer Churn

In addition to compliance alerts, utilising call centre analytics to positively impact bigger picture customer relationship management is a key way to improve call centre performance. This can be particularly impactful for businesses who have time-based contracts with customers or who have a multi-tier or product offering.  

The ability to predict and prevent customer churn and identify cross-sell opportunities is based on timely data, and call centres often provide the missing link of when a customer may have entered a buying cycle. The ability to leverage a Customer Data Platform helps to more easily access data across a customer’s lifecycle with an organisation, to feed into a machine learning model.

Author of “Unstructured Data Analytics”, Jean Paul Isson, outlines four stages businesses can follow to successfully utilise this wealth of data. 

  1. Make sure all the data (structured and unstructured) is captured and stored.  It can then be cleansed, integrated and standardized in order for it to be valid and accessible in a centralised manner.
  2. Perform exploratory descriptive analytics of the structured data. This includes assessing distribution, multi-colinearity and correlation between variables. This correlation includes attempting to make links to churn status. Additionally, process and break down unstructured data with Text Analytics in order to make it usable for machine learning algorithms. 
  3. Implement machine learning algorithms and build predictive models in order to determine churn probability based on the variables and data derived from the previous steps.
  4. Take steps to optimise retention by setting up alerts and actions that should be taken based on the output of the predictive models. Establish KPIs based on customer lifetime value and implement a proactive customer outreach program. 

How to Implement Call Centre Analytics

Given the sheer volume of data which a call centre deals with daily, it is easy to feel like analysing it and implementing a data-driven strategy is a daunting prospect. Customer One View is a cutting edge solution which can help organisations of all types and sizes consolidate their customer data and use it to act on opportunities. 

  • Consolidate customer details from all of your different sources
  • Predict future demand
  • Know your customers better and address their specific needs
  • Identify up-sell and cross-sell opportunities as they arise