Customer satisfaction 2.0: 3 steps to achieving customer delight with automated predictive analytics

Dan Somers, CEO Warwick Analytics

“Customer experience is the new marketing,” Steve Cannon, president of Mercedes- Benz USA famously said. And many would not disagree. Whether you are using net promoter score (NPS) or customer effort score (CES), you have customer satisfaction key performance indicators (KPIs) in place. But how do you know that these measures are useful predictors of behaviour?

Most marketers and customer services managers accept that customer satisfaction KPIs focus on historic customer feedback with fixed measures and outcomes, and are actively provided by the customer, which may be self-serving and unreliable. Yet what else are managers supposed to do? It is seemingly not obvious how else to measure the success of customer services, particularly where there is a departmental separation between customer services, marketing and sales.

Welcome to customer satisfaction 2.0

By introducing predictive analytics, you can turn this concept on its head and ask, “what are the outcomes you are actually looking for?” Maybe you want to know about a very specific target group or how a specific issue might directly affect sales, now or in the future, or to come up with a dynamically-generated, holistic KPI which best predicts organic sales and loyalty.

Predictive analytics helps you answer these questions and more. It’s ultimately about optimising and prioritising the actions to do with customer satisfaction so that they directly influence organic sales, ie, recency, frequency, monetary value (RFV) and/or word of mouth (WOM).

So what are the steps involved? There are three steps which can be undertaken by data scientists or, even better, with automated predictive analytics:

Step 1: Simplify

The first step is to cluster your data, such as the voice of customer (VoC) data, so that semantically-similar reviews and comments are aggregated. This data can be social media, reviews, web logs (if e-commerce), loyalty data, customer relationship management (CRM), point of sale (POS), surveys, etc. 

Similarly, customers can be clustered by attributes and behaviour. This provides value in itself in terms of insight, as well as providing the foundations for further analyses. Note that there are nuances in terms of complaints and reviews which are from positively or negatively-motivated customers. In other words, they may not necessarily be representative. 

Segment normalisation analysis is required to give a more accurate picture. Further, if a business is multichannel, then the reviews might only come from e-commerce sales and need to be normalised, too. There are different techniques for clustering depending on the data and the data scientists’ preferences. 

Automated predictive analytics is a new breed which does not require pre-defining terms or feature extraction. It can be of great assistance in this step, as well as automating the subsequent steps to get to a usable answer quickly, in a matter of minutes rather than weeks (data scientists spend circa 80% of their time manually transforming, cleansing and preparing data). Dictionaries are an output to be validated, rather than an a priori input.

Step 2: Predict

Once the data is simplified and aggregated, there are a variety of statistical and machine learning techniques which can take the results from step 1 and fast forward them to see how a specific issue or sales pattern will look at any fixed point of time in the future. Which issues are the most significant now or growing? How will they affect organic sales? Is a specific problem affecting a specific customer segment?

Again, automated predictive analytics can greatly help here - all of the predictive model building and validation occurs in an automated workflow which learns and improves as new data arrives. A reliable forecast of issues and their effect on organic sales can be generated.

Step 3: Recommend

Many commentators refer to prescriptive analytics as insight which is actionable. This can take various forms, such as a traffic light indicator, eg, of a customer likely to churn, or a recommendation to do a specific action, such as a pre-emptive outbound customer care call. These in turn might be sets of business rules or root cause, eg, if a customer has complained, then they are 50% more likely to leave if there’s a further poor experience in the future.

So many times businesses try to solve known issues in the wrong area or starting point. Often, the root cause is quite removed from the symptomatic pain points and only with complex automated predictive analytics can this be clearly identified and resolved.

By pulling together these three steps and finishing with a set of prescriptive analytical results, it is easy to see how CSat 2.0 can enhance customer services. Now there is potentially a prescriptive action with a quantified percentage effect on organic sales for every corrective and preventative action that customer services can take either towards individual customers or customer segments. 

Actions can be prioritised on predictive behaviour and the closed feedback loop between sales and customer services can be achieved. While this can be achieved stepwise with a team of data scientists and state-of-the-art tools, it can be shown that automated predictive analytics can quickly arrive at step 3 and dynamically update as new data arrives.

Conclusion

So, we can see that by first aggregating the data in a normalised way, then predicting outcomes and finally making clear recommendations for improvement, new KPIs are generated. These can still be used to remunerate and monitor managers, albeit aligning the outcomes of the company more without re-engineering the business. Also, by further understanding the link between customer satisfaction and organic sales (or RFV), then the costs of the customer services department can be optimised and not overspent on non-priority actions.

Automating predictive analytics can enable staff to work smarter and faster, not requiring the analytics team to spend 80% of its time preparing the data. And, of course, it’s not only about optimising scores, it’s about making sure every single customer has the best chance of a positive journey with your company. 

By using predictive analytics, if broken down into these three simple steps, businesses can catapult customer satisfaction from a simple series of KPIs to a far more information-driven, accurate and proven journey - customer satisfaction 2.0.

Please note that blogs are the sole view of the author and that they are not neccesarily the view of IQ ddg Ltd and should not be interpreted as advice. Please read our full disclaimer

Chief executive officer, Warwick Analytics

You have....



to be GDPR compliant.

Register with us for all the news

Sign-up to hear about the latest DataIQ news, content and events.