Not needing customer service is the best example of customer service.
Big data analytics helps you tune into your customers’ expectations, frustrations and demands, so you can constantly evolve your business platform, providing a superior and smooth experience to your customers.
However there is a giant elephant in the room: the probability of failure of big data analytics. You won’t hear people talking about it for the fear of appearing incompetent or ignorant, but believe me, it is a common occurrence with many problems and roadblocks on the way.
Let’s examine a few reasons why big data analytics fail and some instances where they succeed when it comes to customer service intelligence, so that you can avoid these mistakes and improve your chances of success in retention as well as acquisition.
Data models
Bear with me while I spew some mumbo-jumbo.
The concept of data models is very complex. In a nutshell, it manages its constituent elements and their mutual relationships. A database model, in turn, is a logical data model, which determines the structure of a computerized (usually) database, and specifies how data can be added, stored, organized, accessed and edited. Common database models include hierarchical, relational, and semantic.
In your organization, the decision of which big data or automation tools you’ll adopt and deploy is often dependent on the data model that is on offer, whether you realize it or not.
Suppose you are a retail ecommerce website selling t-shirts and such. Using your analytics tools, you could easily profile shoppers who interested in product X – things like the source of visit, age, location, etc. Let’s say you find that the majority of your customers are millennials.
Now you want to know what kind of t-shirts millennials like. So you decide to do product A/B testing based on the relational model i.e. the relation between a demographic (millennials) and their consumer behavior (t-shirt preferences).
See where I am going? At the end of the day, unless you make sure your customer service team sounds more authentic, and totally gets the marketing channels favored by millennials, you won’t see any results from either your data analytics or testing.
Again, it all boils down to the data model. Even a business that makes extensive use of analytics data can go wrong. For example, Google Analytics follows the “last Interaction” attribution model by default when it comes to tracking conversions. As per this model, a product purchase is credited to the last channel that your site visitor interacted with before making the purchase.
For instance, Tim finds your website through organic search, sees some cool t-shirts and forgets all about it. His second interaction comes through a tweet about your blog post. During both these visits, he has liked your products and maybe registered and added a couple to his wish-list but hasn’t taken any action. His final purchase comes a few weeks later when he really needs to buy a t-shirt, googles “hip tshirts” again and clicks through the first ad he sees (yours).
Google Analytics will attribute the success of this conversion to paid search, based on its last touch attribution model. Consequently, you might feel compelled to increase your AdWords budget. This is how you can go wrong.
You must use a lot of different data models to make an informed decision. Blindly trusting one data model can prove to be an expensive mistake.
Unavailability
The Columbia Business School and the New York American Marketing Association surveyed over 250 corporate decision-makers in marketing – director-level at large companies. This is what they found:
- 51% of the respondents said that a lack of sharing customer data within their own organization was a big challenge to overcome.
- Nearly half weren’t using data to personalize their communications.
- Almost a third did not know which high-value customers to focus their marketing on.
- 39% said their company’s data collection methods weren’t well-timed.
There are countless tools – from Hadoop to Kyvos – that help enterprises collect and analyze big data. However, you must remember these are just tools. They will give you valuable insights on your data, but that doesn’t guarantee changes at the ground level.
Most of the times, analysis reports are seen only by the select few in the upper echelons while the team that actually connects with the customers on a regular basis is left out.
For instance, I once ordered a pair of jeans only to find them a size too small. I wanted to get a bigger size, so I left a message on the retailer’s website, for which I got a support ticket. But I didn’t hear from them, so I emailed the whole thing again to the customer care ID and waited another 24 hours before giving them a call. I was told to hang up (my and wait for someone to get in touch with me, which of course never materialized. So I took the jeans to their store (luckily for me, they have one in my city), repeated the whole story to the manager and got my jeans exchanged. To cut a long story short, had the details of my order and issue been available to all employees on the shop floor, warehouse as well as customer service departments, we all would have saved a lot of time.
Although this is my personal experience, I’m sure you will identify with this story. It illustrates how important it is to make real time data analytics more available to everyone, right down to your customer service team.
Expectations
As with everything, CIOs and managers want everything yesterday! They want to see results from big data in as early as 3 to 6 months. Most CMOs and CIOs go as far as to calculate ROI within the first year.
The truth is that it takes you far longer than three months to even make sense of the overwhelming amount of data analytics today’s tool present you, let alone glean insights from them. Then you go on to draw up plans on which metrics you’d like to monitor and meet, based on your business goals, and proceed to implement it over the next six months to a year depending on the scope of the task at hand.
Even then, you can never be a 100% sure that you have made the right changes, so you keep tweaking your data analytics models; the question of ROI doesn’t arise this early in the game.
So how do you work around this problem?
- Find the pain points in your customer service.
- Define metrics for improvement. The success of measuring ROI in big data analytics depends on how well you pinpoint metrics that precisely gauge its success.
- Set realistic short and long term goals.
- Keep updating your big data models as and when you get more relevant data or insights.
For instance, XO Communications’ ultimate aim was to model their customer base and use that data to deduce whether a customer was happy or not. However, this was a long term goal and it would have been impossible to define metrics or determine success based on this goal alone. So, they broke it down into a short term goal of identifying “high risk” customers who could possibly switch to another carrier, contacting them in time and convincing them to stay. (Another win for customer retention!)
XO then converted this goal into KPI form – they aimed at reducing customer query times by up to 90 percent (from 7 to 10 days to less than a day). This was a realistic and measurable objective and they found they saved up to 5 million in revenue in just 30 days by solving this pain point.
They scaled up and changed their models several times after that and their annual savings shot up from $11 million in the first year to $15 million with a subsequent optimized data model.
Disruption
One of the most important ingredients in the recipe for big data success is disruption. If you keep trying to milk the same old data warehouse, team, IT infrastructure and tools, you are headed for big-time big data failure.
To ensure success, you need to be able to do a 360-degree pivot at the snap of a finger – hire experienced data scientists, do not be afraid to use newer tools, encourage disruptive thinking and most importantly, be prepared for implementing changes at all times.
The biggest case in point here would be Groupon’s failure to capture China’s budding market. Encouraged by their success in Europe, Groupon duplicated the same approach in China – a high volume, low touch, cold-call approach (read, mass email marketing). However they didn’t take market differentiators like Chinese culture into the big data equation and thus failed to please their Chinese customers.
Southwest Airlines, on the other hand, used data intelligence quite wisely. They were always good at data analytics and have a track record of successfully using it to improve customer service several times. Some time back, they announced the deployment of speech analytics in order to extract information out of live-recorded interactions between customers and service personnel, in an effort to dig deeper for customer insights. No surprises that they are amongst the top 3 airlines for excellent customer service.
Tesla too created disruption by using data to understand their vehicles’ security issues and recruited hackers to break into their car’s security control unit, a preventive step ahead of their plans to collect more data from their connected cars. This is a great example of how companies today think inside and outside the box.
Over to You
As you saw, there are several ways you can fumble and fail to achieve results from big data for your customer service. If you’re a business decision maker, here are a few sources of big data learning you must make a habit of revisiting regularly. This will help you to gain industry insights on everything related to big data or customer service.
- Clickz’s “Analyzing Customer Data” section
- Inside Big Data’s concise insights on big data strategy
- This huge list active blogs on big data, data science, data mining, machine learning and analytics
- This Quora question, where everyone from data novices to entrepreneurs have shared their favorite resources for big data analytics
Have you tried using big data analytics for improving customer service or are planning to in the near future? What are the other business areas you wish to improve with big data analytics?
I’d love to hear about any news, case studies, experiences and insights on everything related to big data that you might have to share. See you