How customer analytics can measure CX cause and effect
All businesses want to understand causality – which levers to pull to have the greatest effect on financial results – and assumptions about causes and effects underlie all business thinking.
For example, a strategic objective of high customer intimacy implicitly assumes that greater understanding of customers will enable the delivery of an enhanced experience. Then that this enhanced experience will increase revenue from existing customers and generate more referrals.
But these assumptions are not always surfaced, meaning that correlations suggestive of causal flows remain unexamined, to the detriment of business performance. In such circumstances, dashboards are populated with metrics that may reveal what is happening, but not why.
This problem can be overcome by taking a three stage approach.
Surface assumptions and turn them into testable hypotheses
Firstly, be explicit about the assumptions being made and turn them into ‘if…then…’ hypotheses. Then lay them out as links in a chain. In itself this reveals potential risks in logic as the greater the number of assumption links, the greater the number of potential failure points and the greater the chance of the hypothesised causal flow not materialising. It is also necessary to try to step outside traditional mental models and identify alternative hypotheses as this will reduce the risk of confirming weaker hypotheses because stronger ones have not been tested.
So in the example outlined above, there are two links in the causal chain with three points of measurement – the intimacy of customer-facing processes (internal perspective), how customers experience that intimacy through the service provided (external perspective), and how that experience then impacts purchasing behaviour (outcome perspective).
If we look at those three perspectives in terms of the metrics to be tracked:
- Internal perspective: This incorporates the operational metrics that cover the delivery of products, services and support to customers across the research, purchase and usage cycle. These metrics will span a number of functions – marketing, sales, delivery, customer service and finance. Across these functions what customer intimacy means to the business must be turned into metrics. For example the frequency of sales visits made, the completeness of data held in the CRM system, the number of customized propositions submitted, the number of potential issues that were proactively resolved, the number of no-fault emergency deliveries completed, etc.
- External perspective: The external perspective should obviously include measures for customer satisfaction or Net Promoter Score. But other measures of customer reaction – such as the number of complaints or issues raised, service requests, help page visits, even how the product is being used if sensor data is available – also indicate how customers are experiencing the products and services being provided. Text analytics can also be used to track the frequency of key words used in qualitative feedback provided.
- Outcome perspective: This layer includes measures for rates of customer acquisition, retention, growth, and profitability – plus any other metrics that the business believes is critical to financial success (e.g. measures of recency and frequency of purchase).
Design the dashboard and underlying data model
Once the metrics have been identified, the dashboard can be constructed to reflect the three perspectives and surface the genuine linkages that exist. At an aggregate level, this makes it possible to see how changes over time in a customer-facing operational metric flow through to how customers notice those changes and respond, (in some cases it will only be when a metric rises above or falls below a certain threshold that customers perceive the difference). As a result both the base assumptions and alternatives can be tested.
The data warehouse that sits beneath the dashboard needs to be designed to contain all the underlying data at a customer level. This provides the drill down capability for identifying and analysing pockets of difference.
Explore via segmentation
This drill down capability is critical for the third phase. In addition to providing visualisation of results, the dashboard solution needs sufficient analytical capabilities to enable segmentation of results, according to their dispersion. Results can be segmented according to the internal, external or outcome perspective.
So for example, how have customers who suffered a delayed delivery reacted in feedback given and subsequent purchase behaviour? What have customers who rated the service as 4 (out of 10) or below experienced in common and how have they reacted? And for customers who have shown significant increase in value or frequency of purchasing, what might have influenced that?
Such an approach ensures that the insights gained from natural experiments – unintended variations in the service provided – can be gleaned. Only correlation is being measured in these instances and this does not prove causation – only that the relationship is worth exploring. But causation can be proved through structured experimentation – focusing on a process metric to see how an improvement impacts experience and outcome measures. And ensuring that what is measured reflects assumptions about causation enables hypotheses to be examined, enabling businesses to better understand what drives superior financial performance and therefore deliver it.
Jack Springman is head of customer analytics at Sopra Group in London.