The seven evolutionary stages of analytics in the contact centre: How evolved are you?
What typifies the evolution of analytical capabilities in the contact centre - and where are you?
In my last article, I made reference to using data and insights to understand where you are today with Experience Management (XM), where you want to be tomorrow (and why), and what it's going to take to get there.
In this article, I will share with you my framework covering the evolution of analytical capabilities, especially in a contact centre environment (but it equally holds true elsewhere), what typifies each level and examples of their application. This is predicated on the principle that the better your data & analytics, the better your decisions, and that better decisions are more valuable to the business.
This does not imply that moving up a level removes the need for the previous one; all the levels are necessary and useful, even when you've reached the peak of the pyramid. I also accept that many organisations will have invested in analytical point solutions at a level higher than their average capability (e.g. RPA for inbound call triage, RDA for agent assistance) and whilst they may be getting a good return from those investments, that does not mean there isn't even more value to be had from a strategic, integrated approach.
Customer data platform
Without data, you can't have analytics, and a customer data platform (CDP) is just a place where we can bring together all the customer data that is useful. The term ‘Single Customer View’ is hardly new, but many organisations have yet to integrate their data sources to produce a truly comprehensive one. Without it, it is impossible to generate a complete picture of each customer, their behaviour and what’s driving it (what is important enough to them that they are prepared to act). If you don’t know that, how can you be more relevant and serve them better?
I've seen countless examples of agents 'cutting-and-pasting' between multiple operational systems - their applications are not operationally integrated, and often neither is the underlying data. In these examples, data performs a transactional, or record-keeping function, but is not really suited to formal analysis.
Many data management technologies offer the ability to build a CDP without basing it on traditional data warehouse architecture and can also leverage cloud technologies to offer low-cost, scalability and agility. However - do get some expert advice to be sure that what you gather and store (and how you store it) is appropriate and fit for purpose (e.g. useful). Get the CDP design wrong and you are storing up problems for later.
Operational analytics does what it says on the tin - it collects and presents data about the health of the operational processes. Every contact centre I have ever visited has invested in some form of operational analytics - it's often what powers their big-screen dashboards, the weekly business reviews, their agent performance reviews and so on.
Whilst operational analytics is absolutely necessary for running an efficient business, most of the measures used are not valuable to either customers or other business stakeholders: Customers only care about their call, not your average handle time for other callers; the Marketing Director isn't interested in your agent utilisation rates, only in customer outcomes, and so on. It is operational analytics that are often at the heart of SLAs for contact centres or BPOs (Business Process Outsourcers) and they tend to be quantitative, retrospective and introspective. That said, they are very good at identifying costs of operation.
By the way, I occasionally put measures of customer experience (CX) in this layer - for example, using surveys to generate a score for current customer satisfaction, without the objective analysis of why the score is what it is. Not every organisation makes a determined effort to understand the drivers, instead focusing on arbitrary absolute measures (for example 'CSat must remain above X', 'our rNPS should be Y% higher than last year') - the challenges of measuring and interpreting customer experience / satisfaction will be the subject of a future article.
Diagnostic (explanatory / descriptive) analytics
As the name would suggest, this is often the first time that 'analytics' are used to find deeper meaning in the data - the cause(s) behind an effect; For example, 'Why did the average call waiting time go up last Thursday?' At this level, it is more often than not a human evaluation using simple reporting tools or spreadsheets - manually correlating several data points to surface an insight ('Last Thursday we had fewer agents available to handle calls because of staff sickness.').
However, humans are not good at seeing the insights buried in large or complex data sets, and if you are not sure what you are looking for, there's a good chance you won't spot the cause (or even the effect). Experience will take you so far, but to get deeper into the meaning will take something more.
One example of Planning Analytics in the contact centre is workforce management (WFM) which is a special case of demand forecasting. This can range in complexity from fitting a trend line to a collection of data points (with some manual adjustment to take into account other environmental factors), to sophisticated tools that can accommodate influences like seasonality, shift patterns, agent competence, team composition, etc.
Planning analytics typically extrapolates a past trend into the future, based on both data and some assumptions. What it typically does not do is fully account for the influence of multiple variables on a forecast - that's the role of...
Predictive analytics evaluates the influence of multiple factors to explain a past behaviour and predict a future one (to 'model' the causes behind an effect). It uses statistical techniques to identify previously hidden patterns, trends and groupings. For example, a customer's decision to buy can be based on a combination of factors; need, price, delivery schedule, etc. For different kinds of customers, different criteria apply.
Analytical techniques like regression analysis identify what variables contributed to past purchasing decisions of a group of customers and then finds other potential customers who are (statistically) similar and may purchase in the future. Predictive analytics allow us to begin to anticipate the future and to serve individual customers better - each according to their individual need.
Most contact centres are already collecting large quantities of customer data which, if analysed correctly and targeted appropriately, can turn a costly inbound service call into a potential sales opportunity or driver of loyalty.
Predictive analytics, particularly when combining data from a variety of sources, can provide insights into past and future customer behaviour that is valuable to the wider business. Most contact centres are already collecting large quantities of customer data which, if analysed correctly and targeted appropriately, can turn a costly inbound service call into a potential sales opportunity or driver of loyalty.
Optimisation analytics allows an organisation to go beyond answering 'What is likely to happen?' to 'What's the best possible outcome I can achieve, based on a number of assumptions and restrictions?'. For this reason, it is sometimes called 'goal / constraint analysis'.
For example, I have 2 potential customers I would like to win, and a £10 incentive budget. Customer-1 loves my product and will purchase without an incentive, Customer-2 is less well disposed towards my brand and will need at least a £7 incentive to buy. I could give both customers £5 each and get a sale to Customer-1 but not Customer-2. However, by using optimisation analytics, I could decide to give £0 (nothing) to Customer-1, £7 to Customer-2, sell to both, and save £3 on my incentive budget as well. (Not a perfect analogy, but hopefully you get the picture).
The 'in-silico' simulation techniques of optimisation analytics allows a business to evaluate a range of scenarios and pick the one that best meets its needs (goals). Its constraints can be absolute (e.g. a budget), tactical (e.g. a need to promote product-X) or strategic (e.g. 'no customer should receive more than 3 unsolicited emails in a year' - in which case, you better pick the right 3!).
Prescriptive analytics is where we embed and automate some of our analytical decision-making within our operational systems. It can include Artificial Intelligence (AI), the Machine Learning (ML) upon which it depends, and broader automation - for example Robotic Process Automation (RPA) - which is often customer-facing, and Robotic Desktop Automation (RDA) - which is usually agent-facing.
Prescriptive analytics has tremendous potential for allowing organisations to automate operational delivery; lowering costs, improving consistency, and freeing up resources for more complex, human-centric tasks. However, to be really effective, you do need experience in the other forms of analytics - what ML learns is dependent upon the data it is presented with and how the results are interpreted - get either wrong and your AI can suddenly start performing erratically, even detrimentally to your business. A good understanding of data and analytics, within an operational context can help prevent such occurrences.
Data and analytics are not new - the tools of data science have been around for decades (even machine learning) and yet many contact centres are still operating at the lower-half of the pyramid - focussed predominantly on its cost of operation and performance SLAs. Is it any surprise then that they are often accused of being a 'cost centre'?
However, that doesn't need to be the case - a contact centre is a potentially rich source of data into customers, the market and even the company's overall standing. If that data could be merged with other organisational data and analysed, the insights derived could be used to drive up revenues, reduce costs, improve loyalty and even enable business transformation.
Many analysts have identified that contact centres need to become more relavant to both customers and to the wider business. As their role expands to include new channels and methods, more 'operational' decisions will have to be made on the basis of reliable data and enhanced analytics and to be more agile as well.
I expect to see the emergence of more analytical Centres of Excellence that bridge the gap between channels / contact centres and the rest of the business - small teams of analytically-savvy experts that develop the insights and models that enable operational transformation. I would even go so far to say that if you have not considered analytics as part of your transformation strategy, you are not only missing an opportunity but increasing risk.
If you would like to talk more about how to move your organisation up the analytics value pyramid or create an Analytics Centre of Excellence, please feel to contact me for an initial discussion and please leave comments.
You might also be interested in
Peter is an expert in using a combination of data and behavioural sciences to lead transformation in the field of Experience Management (XM); encompassing Customer Experience (CX), Employee Experience EX) and Partner Experience (PX) within an omnichannel environment (and...