The big data problem - what's the point?

Written by Intern - 17 Nov 2017

In this ever-evolving digital age there are infinite ways to capture and utilise customer data – with a seemingly endless pool of metrics to fish from.

But how do you know where to cast your net to extract real value for your business? What are the metrics that matter to your audience, and how do you apply that insight to drive continued revenue and growth?

The ‘data deluge’

In academia there’s an understanding that you need to start at the beginning with data analysis – but it seems like marketing has missed the memo.

As marketers it’s easy to fall into the trap of gathering swathes of data just because we can – and then wonder what to do with it afterwards.

But jumping into the numbers without a purpose is like looking for a needle in a very large haystack. A time-consuming, inefficient and backwards way of completing the task – its lack of focus and intention inadvertently eliminating or distorting key insight.

As many data scientists have warned – it’s easy to get lost in the ‘data deluge’:

Large data sets, such as online customer review ratings, social network connections, and mappings of the human genome, promise rewarding insights but overwhelm past methods of analysis. The result is a “data deluge” (Hey and Trefethen, 2003) where current data sets can far exceed scientists’ capacity to understand them. Despite this difficulty, the rewards of understanding data are so promising that data analysis has been labelled the sexiest field of the next decade (Varian, 2009).” -

The challenge begins, then, not just in knowing how to read and interpret the data, but actually defining its purpose in the first place - and know what to do with it afterwards.

To do that, you need to look behind the numbers and focus on customer experience.

Always start with the customer journey

This is where Customer Experience Journey Mapping (CXJM) holds the key to effective capture, conditioning, analysis and use of data for future strategy.

CXJM is a process that puts the user at the centre, and should be the first thing you do before any data analysis. It switches the focus away from the numbers to look deeper into the needs and expectations of the customer.

It identifies all customer touchpoints with your brand (website, social, email, in store, word of mouth, mobile, etc.); reveals audience motivations, behaviours, desires and fears; and uncovers pain points and areas for optimisation (known as Moments of Truth) in their journey.

A moment of truth is a make-or-break moment between you and your customer. It’s your opportunity to streamline their journey and earn their true loyalty – or lose it forever.

Taking a CX approach to data analysis instead of jumping into the numbers blind means you can: 

1. Build a true picture of how your audience interacts with your brand, i.e.:

  • Use stakeholder workshops, customer focus groups and empathy mapping to get real life feedback from people who matter to your business.
  • Visualise each and every interaction a customer has with your business from first interaction through to brand loyalty and advocacy.
  • See what’s working well (and what’s not) for optimisation. 

2. Give purpose to your data gathering and know the questions that need answering, i.e.:

  • Not just where do customers drop off, buy why they do, and what pages did they visit beforehand.
  • What are people looking for but not finding?
  • Which pages need editing to meet customer needs and expectations?
  • Is there any reassuring messaging or changes you can make to your checkout process to make it simpler? 

3. Gain greater insight into problems that need solving, i.e. does the data confirm customer feedback and identify problem pages with:

  • High bounce rates.
  • High traffic but low conversion rates
  • Abandoned baskets.
  • i.e. which metrics best describe the problem to solve?

How do you apply the data?

So you’ve mapped out your customer experience, identified a problem, chosen the metrics that matter – what’s next? Now’s the time to carry that purpose you defined at the start through to the application of that data to impact your business. 

Ask yourself do you need to PREDICT, CATEGORISE, or REACT to the data?


To predict is to identify customers and prospects who are likely to buy certain products based on their demographic, characteristics and past purchase behaviour.

This is called predictive modelling, and it’s more than ‘just’ predicting the future – it’s using quantitative data such as a customer’s location, marital status, no. of children and car ownership to predict their likelihood of buying a product.  Or analysing past transactional information to predict other products they may be interested in.

Predictive modelling, when preceded with CXJM, is a valuable tool in making your marketing more personalised, relevant and tailored to your customers.

Let’s look at how a data scientist might work with the following example: 

The problem: Customers concerned about a harsh winter say they have trouble buying snow shovels once it starts to snow, as everywhere sells out quickly.

Question to ask (i.e. Moments of Truth):  How many snow shovels should I stock this winter?  When should I place a stock order to keep customers happy?

Metrics that matter:  History of sales, Unfulfilled orders, Weather data from previous season

Solution: Based on past data, create the rules that determine how many orders you’d expect for a given sequence in weather conditions.  Use your prediction model to order ahead of the weather this year, by feeding it the weather data and seeing how many shovels the model suggests you buy to stay ‘in stock’. 


To categorise is to use data segmentation and clustering to learn about your customers and recommend similar products based on their previous purchase history, behaviours and characteristics.

Basically a recommendation engine.  A continual cycle of learn-test-learn means all valuable insight into your customers’ behaviours, demographics, purchasing habits and more is fed back into your marketing activity and analytics configuration to further optimise customer experience with more tailored, personalised content.

You’d still need a data scientist to create and monitor these algorithms – but the rewards of personalisation are rife.  Let’s look at an example:

The problem:  Customers interested in photography are bewildered by the range of accessories available. They know they need to buy lenses, tripods and more but the market has several suppliers and new products launched all the time.

Questions to ask (i.e. Moments of Truth):  They need help, ideally from experienced photographers.  What can you suggest this visitor buys to help them enjoy their new hobby to the max?

Metrics that matter:  Customer data such as geo location, age, purchase history and current purchase interest

Solution:  Data clustering is the process of grouping customers together based on their data footprint.

As each purchase is made and each new customer signs up, different ‘data footprints’ are created and people are grouped into segments based on their locality, demographic and other characteristics (called ‘clusters’).  Close matches are grouped together, i.e. people from Guildford buying telescopic lenses will be in the same cluster, but then if one person buys the latest macro lens the algorithm will automatically move them to another.

With this automatic segmentation taking place it becomes possible to recommend similar or complimentary products to new customers after one purchase, using data available in nearby clusters.


To react is to use data to learn and react to customer expectations and needs.

This is where having an understanding of the customer journey comes into its own as you can visualise single interactions as part of their overall experience – and optimise that experience accordingly.  Let’s look at an example:

The problem:  Customers buying wine in bulk expect discount vouchers, so they’ll wait until they get one before buying.

Metrics that matter: Visitor recency and frequency, email offer prices, time of email delivery, open rates, product interest, cart abandonments

Solution:  Following a visit by a potential customer, record the details of the visit and generate a voucher value within a certain offer range.  Send an email containing the voucher and record the success of the response: Was the email opened, did they click through, did they purchase?  If the email was successful, reinforce this combination for the next time the same circumstances are encountered.

Over time, the optimum email voucher and delivery time will be decided through reinforcement learning.

Three key takeaways

As technology continues to grow, the big data problem will inevitably grow alongside it.  But the key to cracking it is two-fold – and it starts with customer experience:

1. Don’t jump into the numbers blind, understand your purpose before you start to give clarity to your data analysis.

2. Switch your focus to look behind the numbers at the customer journey. Behind each abandoned cart, bounced page, unfulfilled form, or incomplete action is a story to tell, and it’s your job (or ours, for many of our clients), to understand that story, uncover real ‘problems’ to solve and find solutions to optimise the customer experience.

3. Carry that purpose through to application of the data and decide whether to PREDICT, CATEGORISE or REACT

Putting CX first means you start at the beginning, give purpose to your data gathering and find it easier to extract real value for your business.

Need help with your analytics or customer journey mapping?  Find out more about how we can help with Customer Experience Journey Mapping and Analytics and Insight .