FEATURE1 May 2019

Listen and learn

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

B2B Features Impact Innovations Technology Trends

When businesses carry out ‘voice of the customer’ programmes, the questions they ask must be carefully considered. It’s not about getting the answers a business wants, it’s about getting the answers it needs. By Tim Phillips.


When Rosemary Hadden, head of research and insight, joined npower six years ago, the company propped up the Which? annual survey of customer satisfaction for utility companies. “We were bottom of the pack – and by quite some way,” she recalls. 

A programme to improve the customer experience was in place, but its results were hard to understand because – as Hadden describes it – “the view was nearly everything was broken”.

Listening to the voice of the customer (VoC) for a company in npower’s situation is messy and challenging. “In the energy sector, everything you measure is interrelated,” Hadden says. “All of the touchpoints influence one another, so it’s important – from a voice of the customer perspective – to look at the way you analyse and interpret the data, rather than do it in silos.” For example, unhappiness with price affected the rating for customer experience. When bills went up across the industry, negative sentiment on social media would spike – and, for some customers, this was about their bills. In the 2018 Which? survey, 58% of former customers said price was among their top three reasons for switching – while, for some, it was media criticism of npower as one of the most recognisable ‘big six’ suppliers.

Indicators in different parts of the business also sometimes pointed in different directions. So while post-contact surveys sent through email and text message were positive about the performance of npower’s contact centres and webchat service, its relationship surveys showed customer service was consistently rated as poor.

The last time Impact reported on the voice of the customer in 2017, we focused on what we called ‘the multichannel, always-on, technology-driven world of much customer experience research’ that technology platforms could provide. While the potential to analyse and act on large volumes of customer data has increased dramatically, the challenge for many users is to create VoC programmes that take us beyond the dashboard – that give insight into why customers respond the way they do, and what can be done about it. Productised, automated tech-driven VoC platforms may help many organisations, but companies such as npower – with complex, multilayered challenges – risk neither diagnosing the problem nor solving it adequately.


“I would describe a lot of VoC platforms as a kind of customer-centricity theatre,” says Felix Koch, regional CEO, EMEA and APAC, of C Space. “It allows heads of insight or CMOs to say ‘of course we are listening to the customer all the time. We have 24/7, always-on dashboards’.”

Susanne O’Gorman, global head of customer experience at Kantar Insights, adds: “Our clients struggle with three aspects of monitoring VoC. The first is how to make sense of all this information and data they are gathering. Very often, they have KPI systems that are not aligned with the data. Second, how do they turn this into something actionable and, third, how can they understand what the return is on this investment? Each is a huge challenge.”

So how do organisations with complex needs and a desire to improve, such as npower, get beyond the dashboard? Mark Barrett – a managing director of Watermelon Research, which created npower’s VoC programme – says: “You can collect the number and the feedback, but it’s what you do with it – how you choose to fix it, what you expect to happen, and how you offset the costs – that matters. Some companies are measuring a number rather than starting with the issue.”

O’Gorman agrees that this is an essential first step, and says meeting those challenges is primarily about careful design – and using the appropriate methods to collect, analyse and act on data – not the data itself. “First you need buy-in from the organisation,” she adds. “What are the needs of the stakeholders? What are their information needs? How are they going to act on it?”


The way Hadden’s team resolved its contradictory indicators shows why a business governance plan is essential. “We did a deep dive with individual customers and started to realise they’d had to contact us multiple times because the problem hadn’t been sorted,” she explains. “So each transaction in the call centre tended to leave the customer happy at the point the post-contact survey was conducted, only for that emotion to evaporate when the operational side of the business did not fulfil the promises made by customer service. The (non-trivial) solution to the problem lay in a completely different part of the business.”

Demonstrating how accountability spans the organisation, Watermelon and npower have found that automated sentiment analysis can be more complex than it appears. “Social media data is used to generate complaints, or public review sites tend to be people looking for a way to complain. This has a value, but it’s perhaps not the best way to drive strategy,” Barrett explains.

His colleague, Sian Kerr, also a managing director of Watermelon Research, explains that one of the ways the company has been able to help is by creating econometric modelling of different VoC input data in real, or near real, time. So while there are multiple drivers of a metric such as customer satisfaction, what matters for response is their relative influence.

Factor analysis shows that the strong negative sentiment towards npower when prices change is a combination of a transient media-driven effect and an underlying price effect. “We instantly see spikes in social media with negative reactions,” Kerr says. “When the Customer Satisfaction Score (CSAT) drops, we can understand how much is driven by their reaction to negative press, compared with the actual customer experience.”
This means Hadden has more clarity in her reports to npower managers. “Whenever I go to the leadership team, I make sure I know what caused the scores to go in a particular way. I’m not just going in there to say the scores have gone up or down, with no idea what might have influenced it,” she says.

In 2018, for the first time since Which? performed its annual survey, npower was not at the bottom of the table. While it has a long way to go to match the ratings of the smaller companies at the top, it no longer lags behind other large suppliers. Which? was grudgingly supportive: “The foundation for improvement has been there for a while. Our evaluation of npower’s procedures and practices, including on product management and customer service, has found it to be well above average for three years running.”


Collecting data

Watermelon’s approach to VoC stresses the heterogeneity of customers, as well as of challenges and clients. Importantly, this means its VoC platforms may make use of old-school forms of collecting data, such as CATI or face-to-face, because it has found it is important that the way data is collected matches customer experience. “Do what is appropriate and relevant to reach the customer, not just what you can do on the platform,” Barrett recommends. “Mirror the channel through which they deal with you.”

Real-time feedback is exciting, but not every piece of analysis on the dashboard needs this attribute. And while many VoC platforms stress automation, Watermelon has invested in maintaining its capability in analytics and techniques such as CATI, which it uses to generate insight for clients and other research agencies that are trying to deepen their VoC ability.

One of the related pitfalls of data-driven VoC programmes is that they may tempt businesses to obsess on local, transactional peaks and troughs at the cost of insight into a bigger picture. Charlotte May, head of group customer research at Legal & General, has been running its VoC research for four years, during which time she has concentrated on building a longitudinal model of how customers view the company. “There was a real need for it to become much more of a barometer of how our customers were thinking and feeling, so we could use that to help shape and inform far more constructive actions,” she says. “The analogy is, you could use a camera and take some really beautiful pictures of the scenery, but if you have a film camera, you can get a view of the nuances in the weather. You get much richer detail when you look longitudinally.”


Balancing reactions

Looking only at the snapshot view can lead to over-reaction to events, May says, but also under-reaction to important trends. It’s hard not to react to a seasonal spike in the data, or two inflammatory comments in a post-contact survey, but that might not be the best course of action – taking a longer view might show up more important insights. “Potentially, those might be the only two comments you’ll ever get in that one area, in two years. Whereas something that’s not as provocative could have a much more positive or negative impact on your overall customer performance,” May explains.

An example is the analysis of Net Promoter Score (NPS) data. While low scores are potentially a cue to act, those in the neutral zone can also create insight into how to improve the business – although, in many NPS surveys, customers giving a mid-range rating are not asked why. “We flip the question round and get them to tell us what would have made the experience a nine or a 10,” says May. “By understanding what would have taken it from good to great, that’s really helpful, because you’ll often get a very different type of insight into your customers.”

Another reason to take a cautious view of what the dashboard tells us is that valuable insight is scarce for many companies – though that’s sometimes not obvious watching metrics bouncing up and down. This is even the case with transaction data, argues Clive Humby, formerly of dunnhumby, where he was a pioneer of informing VoC programmes through data analytics using Tesco Clubcard. Now he is the chief data scientist at Starcount. He points out that most consumers think about brands rarely, and without a strong opinion either way. So, as with Hadden’s call centre ratings, we should be sceptical about attributing meaning to the data we collect during the transaction or at infrequent touchpoints.

“We were really lucky when we worked in grocery, because you buy so much in a grocery store that your transactional data is very valuable,” says Humby. “It tells us all the basics about your ability to cook, whether you’ve got a family, whether you’re economical or treat yourself. Virtually every other business in the world isn’t like that; you’ll be lucky if you get 12 transactions a year. So what does that say about customers? Not a lot.”

The challenge is to hear a customer’s voice during the huge part of their lives when they are not transacting with you, or even thinking about your brand. This contextual information is often ignored in VoC platforms, but is a factor in many metrics. This type of outside-the-customer-experience listening is, by definition, not explicitly captured in much VoC data, but does drive the numbers.

For example, an energy supplier suffering from strong negative sentiment after a price rise may not find that business suffers, because every other supplier feels the same effect.

“I would argue that the most important aspect of a customer that anyone needs to understand is, are we part of their consideration set – and, if we are, how big is that consideration set?” Humby says. “It’s almost a share of wallet measure. There are early indications we might get something in that space from people such as Mastercard and Visa; they won’t give you personal data, but they will give you patterns.”

Social data – the focus of much of Starcount’s activity – is one way to deepen that non-transactional insight. Social listening is a valuable way to hear about an emerging customer problem or event to track individual complaints, but it has a negative bias (you don’t tend to care much about a brand until it annoys you) and a tendency towards groupthink. Starcount, instead, has identified a way to hear the customer’s voice by looking at the things that groups of customers like or follow precisely when they are not interacting with the brand. “On its own, social is not the answer, but aligned to the transactions you have – perhaps to market research, perhaps to other third-party data, such as traditional geo-demographics – it helps you understand where you’re winning and where you’re losing,” Humby explains.


Business value

Aligning VoC platforms to business needs and creating effective plans to act on what they tell us are both complex – but evaluating whether that has created value for the business can be even harder. Hadden’s insight that “everything you measure is interrelated” is common to many types of research. At Watermelon, Kerr argues that regression analysis in real time can reveal the largest influences on customer experience. The same analysis can also uncover the likely effect of improving individual metrics. It is, she says, an essential starting point for the root-cause analysis that decision-makers need when presented with customer problems or opportunities, suggesting both a course of action and the likely return.

At least at the beginning, when there are many improvements that can be made to customer experience, it may be more practical to target project-based discrete “wins”, says Kantar’s O’Gorman. That means designing a business goal that can be captured in data and explained to people in the business who will make it happen. “The ability to resolve a customer service issue with the first call is one of the most effective drivers of customer satisfaction and loyalty, for example – and you can say ‘these are the actions using the programme that we know will help us to improve first-call resolution’. Then you can estimate what this means in terms of putting dollars against the investment.”

Creating discrete campaign-driven improvements to the customer’s experience is also the thinking behind the VoC programmes that Feefo, a public review site, is beginning to run with its clients (see ‘Feefo Feedback', below). This can also be tackled, says Kerr, by making sure that VoC platforms deliver only relevant information to each part of the business. While call centres can focus on driving metrics known to influence business performance, decision-makers in other parts of the business may get more long-term, top-level information. On this scale, Watermelon clients can interactively and collaboratively build action plans on the platform.

First-hand knowledge

On 3 January 2017, Mark Zuckerberg posted a blog on Facebook saying: “My personal challenge for 2017 is to have visited and met people in every state in the US by the end of the year.” The road trip predictably earned a lot of media coverage for Zuckerberg’s “deep reflection” about the role of Facebook in ordinary people’s lives, and there always seemed to be a photographer to record it. He got to do many things that Silicon Valley billionaires usually don’t do, such as drive a tractor and eat jello. The tacit admission is that Facebook’s CEO, who was a billionaire in his early 20s, had oodles of real-time and longitudinal data about the voices of ordinary people – more data than any CEO in history – but he had almost no first-hand experience of the lives they led.

Some researchers claim that experience of these first-hand customer voices is essential for making VoC relevant to the business. At C Space, Koch says organisations should always supplement their data with personal interaction – and that doesn’t mean observing customers from behind a two-way mirror. The danger of data-driven platforms, he warns, is the implication that face-to-face interaction is now optional. “Managers can say, ‘we don’t have to meet customers, we don’t have to do the messy thing of speaking to them, because we know what they feel and we know what they like’.”

C Space listens to the voice of the customers by immersion programmes, taking executives to eat in their own restaurants and browse in their own shops, often places they rarely visit. “It’s very human to want to retreat to an area where you are not confronted with difficult news and difficult conversations, because this is about being faced with something that’s broken about your company,” Koch explains. C Space also uses a ‘buddy system’, an app that creates a one-to-one relationship with a customer to message throughout the day – in Koch’s words, “injecting some humanity into a process that is devoid of emotion”.

Injecting humanity can also inform managers about customer priorities and irritations that they never considered as relevant. “We’re seeing a lack of empathy,” says Craig Scott, founder and CEO of Mingle. Inspired by the example of legendary P&G CEO AG Lafley, who insisted on a customer visit in every location, Mingle will work with a small group of managers to meet users one to one – in their homes, their workplaces, or even, in one case, at a beer festival. The brief is to find a real-life response to the important questions that they need to investigate. “A spreadsheet doesn’t tell you that. Big data doesn’t tell you that. Epos doesn’t tell you that. You have to see it with your own eyes,” Scott says.

He explains that data creates what psychologists call ‘effective empathy’ – a way of evaluating the voice of the customer while still keeping them at arm’s length. “But if you get out of the office and see for yourself, you learn in a way that’s called ‘cognitive empathy’, which is far more long lasting.”

Their experiences are fed back in a workshop, and this becomes the spark for new learning that Scott calls the “Mingle tingle”. This tingle could be to use a new insight into ‘why’ (a kitchen appliance client was surprised to find that its customers were shoving their appliances into cupboards, because kitchen gadgets are universally regarded as ugly), or simply to understand context to help them decipher the data.

The idea is not to replace quant-based VoC data, Scott says, but to inspire and supplement it. In an era when many executives are more at home with data than people, the voice of a real-life customer can be both effective and intimidating. “One managing director had addressed the Stock Exchange two weeks before we met,” Scott adds, “but he was terrified to meet a housewife in Hull.”


Who is the B2B customer?

In the consumer world, measuring the voice of the customer is straightforward, because it is the customer who makes the purchase or uses the service. In the B2B world, it is not always so easy, as the internal buyer or procurement department often does not make decisions in isolation. Many B2B marketers are starting to listen more to the people in the company who use the products they sell – not just the ones who sign the cheques – realising that they often have a significant (if informal) influence on how the product is bought or used.

When she joined SAP as chief marketing officer, Alicia Tillman realised the business was “at a crossroads”. The brand was all about technology, so had limited communication with, or knowledge of, the people on the front line who used it every day. “I’m no longer marketing just to the decision-maker,” Tillman says. “The end users will tell us what they need to be successful. So our method has to become more consumer-like, because technology has given people a voice. Every company talks about voice of customer, but now every role within our company is focused on finding that voice.”

This also means reorienting the company to interview and survey its users – not just rely on operational data – and to make rapid product updates based on the feedback.

“There has been a tremendous push towards data, but if we believe that the only way to make a decision is based on the data, we will not win the hearts and minds of our customers,” Tillman adds.

Feefo feedback

“It started out that we just helped people collect reviews for their brand,” says Neil Mcilroy, head of product innovation at Feefo. “But our product has evolved; now clients are using it to interrogate verbatim responses that they collect from customers as well.”

Nothing shows the voice of the customer in its rawest form than online review sites, but the problem of fake reviews has made some of these tools unreliable as a basis for VoC analysis – and it is hard to act on reviews unless you know something about the reviewer. Feefo claims to avoid the fakeness problem by ensuring that reviewers are validated customers, and by preventing clients from cherry-picking who is invited to review or whether those reviews are displayed.

It is also expanding its offering. “We now deliver campaign management tools and reporting tools, which help measure that customer life-cycle and journey that they’re providing for their customers,” Mcilroy says. This means giving scripts to the client to integrate Feefo’s tools with its transactional data.

To use the reporting tools, Feefo captures details of the transaction – for example, whether it was online or offline – or customer profile data to allow them to target campaigns. Now you’re as likely to get a request from Feefo to give the retailer you purchased from an NPS as a review. Customers such as Iceland are using this information to estimate the RoI of improvements to service.

It has also been building sentiment analysis tools to examine the reviews it has collected. “The product – up until the past 18 months or so – was a lot of numbers, benchmarking, performance reports, but all very quantitative. What we’ve tried to do is take all that qualitative feedback they collect on a case-by-case basis from customers and give them the tools to take action and get insights from that information at scale,” Mcilroy says.


A force for change by Watermelon

A common problem we see with organisations that have implemented a voice of the customer programme is that they can quickly become a report on ‘business as usual’. It ceases to be a force for change, with employees feeling the programme is giving no new insights they can learn from or use.

It’s vital to engage employees and communicate the rationale and aim of the programme. To effectively embed it and gain commercial benefit, employees must believe in – and continually use – the programme.

So it’s important to build an evolution plan into the implementation and design phases to address current business needs, as well as being flexible in longer-term objectives. Input and resource is required to expand the programme so it not only continues to change with the business, but also shapes tactical and strategic planning.

At Watermelon, we recommend the following steps:

  • Support the rollout of a programme with a detailed and phased approach to engagement. Don’t try to launch too much too soon. This can be overwhelming and confusing for colleagues who are unfamiliar with the data and tools at their disposal
  • Don’t just leave it in place and let it tick over. Schedule performance reviews of VoC to challenge whether new measures, touchpoints and channels need to be added, or others changed or dropped. Map the changes required and be sure everyone is aware of the plan and is working towards it
  • When re-scoping, focus on innovating rather than just more numbers – whether that is how you are engaging with customers to capture data, or how data is shared and visualised across the business
  • Communicate the benefits of VoC to the business and customers: ‘This is what we have been told, these are the steps put in place to address it, and this is the change you can expect to see.’ This way, you ensure colleagues continue to be invested in the programme and customers see the value in responding
  • It is important to track key metrics, but always look for the drivers behind this and any emerging items. Text analytics can be a powerful tool in discovering where new issues and opportunities lie
  • Any VoC programme should help to uncover root causes of problems and service improvements, but make sure there is a positive, developmental aspect, too. Sharing best practice and learnings across teams and departments helps to develop colleagues, promote collaboration and improve the business.

This article was first published in the Issue 25 of Impact.