Using predictive analytics to create a reliable crystal ball
Global accounting and reporting senior director at Huawei, David Wray, explains how predictive analytics can enhance customer service
Add bookmark
Following his session at CXN Live: Predictive CX, global accounting and reporting senior director at Huawei, David Wray discusses how companies can eliminate data silos and use predictive analytics to gain intelligent customer insights.
What has your CX career journey looked like so far?
David Wray: For me, it is a journey that started by accident. I have been in finance pretty much my whole career and one aspect that became interesting to me was looking at data in the context of, for example, generating additional revenue.
So, I started looking at data. More specifically, how I could use a variety of sources I already had access to, like customer buying patterns, and considering how I could use it to influence [the customer] experience.
That has really been my career journey - looking at data and seeing what [new] insights I can create. My journey has not necessarily been typical, but it is one that has developed quite naturally.
Read: The Big Book of Customer Insight and Analytics 2021
What are predictive analytics and how can these systems enhance customer service?
DW: Predictive analytics look at data and extrapolate out to predict [an outcome], whether it is a purchasing pattern, product success or a [new] market. [These systems examine] an element and [forecast] where it is going to go, somewhat like a crystal ball. However, as a company, we want something more reliable than a crystal ball. So, predictive analytics uses data, statistics, and some basic assumptions to map out patterns, connections, and algorithms. It essentially leans on math to create causal relationships, but also foresees where [events may head], based on what happened in the past. Past experience is one of the basic assumptions we need to continuously monitor, otherwise the predictive nature of the analysis could easily go astray leading to poor decision making.
If we observe a product going through design, manufacture, sales and post-sales support, we could run [predictive analytics on] the [experience and operational] data from each stage. If we start to see an issue or opportunity in one part of this process chain, we can get an idea of where this is likely to [end up]. For example, if we start to develop a product and notice some customers are talking about a [needed] feature, we can see what would happen in terms of sales if we built this feature into the tool.
Get the latest CX resources straight to your inbox
Sign-up to the CX Network newsletter today to stay informed about new CX resources and industry updates.
How can companies combat data silos so they can deliver predictive customer experiences?
DW: Firstly, it is important that companies have a good sense of what data they actually have. Get an inventory of the data you have across the organization. Seriously consider about putting it into [a repository] whether it is a data lake or some other cloud environment which breaks the data silos. This will allow you to start to extract value from the data, whether its structured or unstructured, [in the form of] intelligent analytics insights.
A second thing that companies need to make a conscientious choice about is master data management. Give common naming conventions to data points, and keep the same terms for that data point throughout the company. It is incredible the number of companies that do not have this consistency. This set-up leads to an extremely inefficient environment, and a significant loss of potential value, because the data has to be cleaned to a point where it can be mapped and connected to other related and relevant pieces of information.
Read: Customer Experience Predictions Report: 2022
What is your golden rule for using CX metrics to optimize experiences?
DW: One of my golden rules is: never use CX metrics (or any other metric for that matter) without first having a hypothesis around what the data ought to be telling you. I always want to know that what comes out of the proverbial black box is sensible. I personally never use data that emerges from any system without asking myself whether its insights are reasonable.
A second golden rule is validating the data and insights with other people. Ask other departments in the company whether the [results are what] they would expect. I don't know of anyone in a reasonably sized company that has a complete view [of a business] with the degree of depth and breadth of understanding necessary to make decisions in a vacuum. So, it is really important people validate what they are seeing with other teams - it is a form of sanity check. It is about using insights on an informed basis, and if we are going to make growth, transformation or investment decisions, then we should only make them with data we know is reliable.
