Five reasons big data analytics can fail for customer service
Big data offers businesses an exponential increase in their ability to understand, predict and serve their customers.
But between the promise and the payoff, organizations have to navigate the perils of new technology adoption and implementation.
Everyone has heard of big data, but how many companies are actually implementing it?
Half of all respondents to an IBM survey said their organization already had more data than it could use effectively and identified improvement in information analytics as a top priority in their business.
Source: MIT Sloan Management Review, Winter 2011
Big data implementation lags behind awareness: between 10% and 25% of enterprises have begun big data implementation and about 70% have plans to implement it in the coming years. However, these figures don’t show how many companies have managed to create value from their big data efforts.
Big data is approached in different ways in different parts of the enterprise. As Dean Lane, Founder of the Office of the CIO, a San Francisco–based IT consultancy, points out:
IT leaders are more focused on the tactical aspects of big data like costs, project management, security, validation, and verification. The rest of the C-suite, on the other hand, is more concerned with deriving meaning from the data. They care about what it will reveal, what marketing campaigns it will be used for, and how to design a project around it.
The first place most enterprises are considering implementing big data is marketing.
Retailers see cross-selling or upselling opportunities, B2B pros look at the chance to do laser-focused persona-based marketing and sales people imagine picking up the phone and talking to a lead they have rich, deep information on.
All those hopes are well founded. But what about customer service? Big data offers a chance to move customer service to center stage.
And that’s where it belongs: siloed sales and marketing that rely on individual sales as their key metric for success neglect the effects of churn on business earnings.
Every year, bad customer service costs American businesses $84bn, and 78% of customers have walked away from a business because of poor customer service.
Slow responses and lack of social media engagement are among the most common causes, but overall the problem is simply stated: churn leaks money and bad customer service causes churn.
Big data offers clear opportunities to deliver rapid, personalized customer service that transforms business performance. In fact, for customer services, the promises of big data are in some ways even more beguiling than for marketing or for the enterprise as a whole.
Used correctly big data should deliver analytics that allow agents to have conversations guided by the kind of information about a customer a friend would have – preferences, history, interests, all at the agent’s fingertips.
But there’s a lot to go wrong between here and there. How can big data analytics projects fail customer services?
Big data has to be modelled appropriately in order to deliver value to customer services and ultimately to the customer. Modeling has changed radically to keep pace with big data demands, but adoption rates lag behind requirements.
Pre-big data models were often built on SQL, i.e., Search Query Language. This database-oriented modeling system let companies build useable structures for the data they had. But it was too rigid for big data, which is often collected in a raw state and requires more agility than SQL can provide.
The data we’re used to handling is mostly structured data, ideally suited to database storage and analysis. But big data is largely unstructured data – think Facebook posts, tweets and so on – and SQL is too rigid and restrictive to allow analysis of data this varied.
The nearest thing to a standard, enterprise-grade data model and analytics system for big data is Hadoop.
Created in the mid-2000s as a response to the increasing amount of data Google had to process, Hadoop relies on noSQL – for ‘not only SQL’ – together with a system for carrying out distributed computing across ‘nodes’ and code for planning those computing operations, called MapReduce.
Despite sounding ramshackle Hadoop is incredibly efficient and fast, but the very fact that all this background architecture is so well known points to the immaturity of the technology.
Source: Glenn Lockwood
Data models in common use face three challenges. Each of these can be a major stumbling block:
The greatest opportunity big data presents to customer service is the ability to accurately monitor performance and predict requirements in a way that just wasn’t possible before.
Surveys and tick-boxes gave an always-incomplete view of customer satisfaction. Big data can deliver greater depth and scope.
For instance, say you’re a retailer with a national presence. You can use a tool like Social Mention to identify the positivity and negativity of tweets based on sentiment and author influence. Then you can cross-reference that with internal data you collect from sales and customer care, segmented by region. This approach can yield genuinely useful insights.
If you find that customers whose customer service tickets take the longest to close are also most likely to become repeat customers, you have an interesting problem – you need to start listening to some calls and figure out what they value about lengthy customer care experiences.
If sales are down across the country for one product and complaints are up, the product might be the problem. If long ticket times go with lower revenues in a specific region, you might be looking at a training issue.
And all of these might never have come to light without access to big data. But the crucial word in that sentence is access. The data has to not only exist, but it also has to be accessible in such a way that people who can make business decisions can see and understand it.
They’re not usually going to be data scientists, so the task that lies before companies is to standardize data collection, modeling and analysis so insights are easier to get at for decision makers.
Without a system to arrange data in a comprehensible way, it’s worse than useless. That was true of the data we had on customers ten years ago: it’s exponentially truer of big data. It’s not a collection issue – you can have as much data as you want, enough to drown in. It’s an analytics issue, but the solution is a model that allows easy analysis.
Big data can create a huge range of opportunities for managers. Who wouldn’t look across all those new chances to accurately target customers with offers, deals and products tailored to their interests, and rub their hands with glee?
But it’s a temptation that has to be resisted if your big data efforts are not to actually sabotage your customer services.
Customers don’t want to be pursued by customer services. Ideally, customer services are like emergency services: unnoticed until they’re needed, then prompt and efficient. Big data can be used to make the customer available to you, but that does nothing to address the issues that irk customers.
When customers complain about unavailability they’re usually thinking of the difficulties they face in reaching someone in customer services who can actually help them.
Long phone waits, automated greetings and pushbutton menus, and a lack of integration across channels all add to this frustration. To address this, use big data insights as a roadmap to build a more integrated and efficient customer journey.
We can look to the example of Amazon; not only does Amazon collect gigantic quantities of data on its customers but it’s efficient at transferring the insights from that data to customer services, both on the phone and at every other interaction with the company.
Ovum’s Aphrodite Brinsmead advises:
Customers don’t want to have to leave a mobile application to then go to a community or chat to get technical assistance.
However, when a customer service agent is needed for a live interaction, then big data comes into its own: “Organizations should push context about the customer’s web history or previous questions to an agent in advance,” continues Brinsmead.
Within the organization big data insights face a distribution problem. In order to benefit from big data insights, managers actually need to see and understand them; restricting access to reports stifles efficacy and prevents full utilization of big data-derived insights.
Instead, as ARTYCS co-founder and CEO Laurent Fayet says, agile Big data Analytics needs to be both based on business use cases and made accessible to every level of the organization.
When big data is presented raw and unstructured – as the majority of it is – it’s unusable. The challenge of analyzing this unstructured data is massive: IDC estimate that only 0.5% of available unstructured data is even being analyzed.
When its insights are presented as statistical models they’re often beyond the capacity of managers to interpret and act upon.
What’s needed is simplification and mission orientation: at every stage away from the analytics team big data insights should be repurposed to be more user-friendly and actionable.
Big data doesn’t replace management skill: there have always been sources of data about processes and performance, and it has always been the task of management to extrapolate meaning from them and create strategies to improve.
Big data is no different in this respect: when organizations pass on statistical overviews or information without instructions to employees there is a failure of management and the result is a failure of teams at ground level to improve.
The classic example of this is Amazon. Amazon’s 152 million active customer accounts and 1.5 billion-item inventory generate enormous quantities of data, as well as $118,000 in revenue every minute.
Amazon has more data, going back further, than almost any retailer, but it also has an unrivalled track record of turning that data into useable insights: you’ve seen this in action, because Amazon has offered you ‘products you might also like.’
On the surface, it seems a simple thing but it’s backed by a huge data acquisition and analysis effort – and it yields Amazon 35% of its sales. The steps between data acquisition and actionable, customer-oriented insights are mostly to do with eliminating complexity.
As statistician Nate Silver observes,
Every day, 3 times per second, we produce the equivalent amount of data that the Library of Congress has in its entire print collection. Most of it is…irrelevant noise. So unless you have good techniques for filtering and processing the information, you’re going to get into trouble.
Big data is the next big thing. That’s not in doubt. But its transformative power has parallels in the past and new technologies typically carry with them two contradictory sets of expectations: that they will revolutionize life and make the impossible possible, and that they can be understood in terms of already-established tech. If they could, they wouldn’t be revolutionary.
Big data analytics has to deliver insights that can actually improve customer experience of your sales process, your website, your products – every contact with the business. And that’s possible. But initially it’s going to have to happen upstream, against a weight of inaccurate expectations about what big data is, what it can deliver and when.
That’s not to say that expectations of business advantage from big data are false: they’re not. As Matt Jachius says, the ability to proactively and predictively apply big data analytics to customers “is the basis of competitive advantage in the future… because you can provide a better experience.”
But while some managers will dismiss big data as smoke and mirrors because its claims sound too good to be true, others will err in the opposite direction and buy into the idea of massive predictive powers without the legwork and adaptation that will have to accompany their acquisition.
If you want your agents to open conversations with customers with a huge quantity of useful information about that customer at their fingertips, there’s a lot that has to happen first.
We’ve talked about data models and about how companies have to approach big data with strategies to prevent being entangled in complexity or simply drowned in raw signal – but reliance on analytics without checks can result in skewed results that actually prevent business success.
James Guzcza, Deloitte Consulting LLP’s Advanced Analytics & Modeling’s national predictive analytics lead, observes:
The business world is too complicated to make decisions based on gut feeling and professional judgment alone. Companies should consider shifting to more fact-based decision-making or risk being out maneuvered by their competitors who do.
Yet Guzcza also cautions that reliance on analytics alone can lead companies to make complicated decisions based on analytics alone when skilled judgement combined would provide better direction than either alone.
Traditional customer service has been about fielding complaints and questions.
That’s a flawed model for all kinds of reasons – the two most obvious are that it’s better to have a satisfied customer than to placate a dissatisfied one, and that only 4% of dissatisfied customers complain directly to the company involved.
Source: Understanding Customers by Ruby Newell-Legner
Instead big data offers the opportunity to develop predictive and proactive customer service. But that’s going to come at a massive cost in disruption to traditional business models.
First, executives will have to get used to using and developing new expertise: they’ll be valued more for their ability to discern the lessons offered by data and less for instinct.
Erik Brynjolfsson, Director at the MIT Centre for Digital Business, says,
Organizational judgment is in the midst of a fundamental change—from a reliance on a leader’s ‘gut instinct’ to increasingly data-based analytics.
Secondly, businesses will need to integrate horizontally; the separation between sales, marketing and customer service has always been artificial, and forward-looking people from UX designers to sales reps have been saying the same for years.
Finally, our whole conception of customer service will need to be upended. The future will look more like collaboratively creating and curating custom experiences, and as more data becomes more available to more people we’ll ultimately be collaborating with our customers to build them.
The exponential growth of data availability, powered by cheap memory, fast internet and cloud computing as much as by fresh data sources and new data models and analytics systems, is only the beginning.
The Internet of Things will generate data sets that will make the current deluge look like a leaky tap.
Customer service has the chance to reinvent itself at the heart of every business activity, using big data as fuel – but it’s easy to blow up on the launch pad.
We’ve seen how expectations within the company can stymie adoption and implementation, how complexity can overwhelm systems and how inappropriate analysis and modeling can deliver unusable results. Avoiding these pitfalls will place you ahead of the curve in the inevitable process of learning to live with and work with big data.