/ Analytics

Hello From The Other Side

Synergy between market researchers and data scientists…

In an era where everyone is jumping on the Big Data and unstructured data analytics bandwagon, it is surprising to see how many market research studies (especially those involving customer satisfaction and loyalty) ignore basic data analytics. By this we don’t mean no analytics were done, it is just the absence of looking beyond what was asked on the questionnaire.

A case to be made against reporting ONLY a ‘single measure’

Almost all companies generate at least one customer metric through operational or strategic metrics: the Net Promoter Score®. This score was invented by Dr. Friedrich Reichheld and popularised by Bain and Co. over the past decade. It is intuitive (measuring whether a customer is likely to recommend your brand, product or service to friends, family, and/or colleagues) and easy to calculate (NPS® = % of people willing to promote – a 9 or 10 on a 0-10 point scale – minus % of people not willing to promote – scoring 0 to 6 on the same scale).

The score can be interpreted relative to the company’s past NPS performance, to other players in the industry (through local providers such as the South African Customer Satisfaction Index (http://www.sacsi.co.za) or internationally through providers such as Satmetrix (http://www.satmetrix.com)

However, what constitutes a good NPS? How can the NPS be used to drive action? Does the company even benefit in real terms from an increased NPS?

To answer the first question companies often establish whether they improved on the previous NPS. Alternatively, the company’s NPS could be viewed in comparison to the industry.

If the question was measured on the back of a service engagement, there typically is not much to link it with. So to answer the second question companies typically turn to theory and look to do two things: improve overall satisfaction and decrease complaint rates. Easy, right?

In our perspective, this is like trying to solve a business problem with your eyes closed, with your hands tied behind your back, and being spun around twenty times. What if the 50% promoters are all your most valuable clients? What if the 20% detractors have got the most influence in the social sphere? What if the promoters link to a specific product, channel, or region? What if the behaviours of the promoters and the detractors do not differ?

The key to making NPS work for the business is to understand the why behind the number; and this why can be determined from more than just respondent feedback. Very often, companies sit on a gold deposit of existing business data that can inform this why.

Business Linkage Analysis – Providing Real Value

To answer the underlying questions stemming from any number (estimate) generated from research surveys, companies must link it back to other sources of data and analyse holistically. By adding financial information, customer behaviours, value segments and the like to the data, companies can extend a cross sectional view (from surveys) into a longitudinal analysis that determines cause and effect.

Unfortunately, this process does not start after the research has been done, but should be an integrated approach between the research and the business intelligence (BI) teams from the start. Furthermore, the maintenance and updating of integrated data systems should remain a priority for business after initial implementation in order to facilitate quick access to relevant information. This quick access to relevant BI is a differentiator in the business world (more about this in our next post). Unfortunately, this rarely happens, and it is the reason why companies find it hard to prove the Return on Investment on customer experience programmes (34% of leading companies find it difficult to tie customer experience to business outcomes – HBR Analytic Services 2014).

For example, each research project starts (or should start) with a business question, problem, or objective. This enables the research team to choose the right methodology, developing the right instrument (questionnaire), and ring-fence the right population groups to study. Too often, the only interaction with the BI team is to get the contact details of those customers they need to survey. And even the research outcomes are not shared with the BI team to establish to what degree findings can be linked to existing BI.

Ironically, a data analytics project starts off in exactly the same manner as the research counterpart. From a business question, problem or objective, the most fitting data sets are sourced and combined. These data sets are then cleaned, structured for analysis, and mined to deliver answers on what is happening in relation to the business question.

Unfortunately, too seldom would there be interaction between analysts and customer experience teams to understand why customers behave the way they do (from what is highlighted in the modelling process).

The Synergy Process

Our suggested process (called - very fittingly - the Synergy Process, based on the thinking described in the book by Stephen Covey ‘The 3rd Alternative’) combines the typical processes to deliver a synergistic approach and to leverage on the benefits of BI in the market research space, as well as leverage from the knowledge gained in market research projects for model building purposes in the BI space.

Big Data The Synergy Process

Figure 1: The Synergy Process

The main steps in this process are explained here in short:

  • The process starts off with a business need, problem or question. This is the ‘why’ (as described by Simon Sinek).
  • The Research and BI (and other relevant) teams collaborate to determine critical objectives.
  • Understand the data you have and how it relates to the objectives.
  • Augment knowledge gaps in the existing data with Research.
  • Having completed the Research, the new data should be linked to the internal data to enhance the insights in terms of the what, how and why of consumer behaviour.
  • Involve experienced field experts and overlay research insights on the existing data. This will deepen understanding the ‘why’ research insights and added to the model building process.
  • Action your insights. Influence the need originally identified. This can be facilitated using a model to predict, forecast or classify some objective. As written in Prediction Impact Inc.’s paper ‘Seven Reasons You Need Predictive Analytics Today’, “Predictive analytics provides abundant opportunities for enterprise evolution. It is specifically designed to generate conclusive action imperatives.”
  • Lastly, with research and models alike, it has to drive measurable action to improve and evolve business. That is why it is important from the start to have a business objective that has significant value (top level buy-in), and will introduce significant benefits when evolved, improved, or made more efficient. In short, real Return on Investment.

With the advent of Big Data, Internet of Things, and Social Data, it will become imperative to be able to link research objectives to real world data to drive real business value and actual change within business.

Can you afford not to?