Many organisations have a lot of data present but lack the internal structure to actually perform the necessary data structuring and analyses. They are sitting on a goldmine, but forget to take out the shovels and start mining. That's a shame because, with a number of simple tips and tools and a bit of commitment, organisations can easily unlock the value of their data.
First, let's sketch the current situation within many organisations that do not have economies of scale to appoint a dedicated data analysis officer. First of all, nobody is responsible for all data together. Every dataset has its owner. Sometimes it's even worse and the ownership of data isn't taken care of at all. This means that potentially there is a lot of data and thus possibly a lot of useful knowledge. The only problem is, no one is responsible and even if someone sees the value of gathering the data and trying to make more out of it, they aren't given the time needed to do so.
Secondly, whether they see it or not, most of the time organisations actually have a lot of data at their disposal. For example, order-info, lost sales and customer contact information. Each of these datasets is often kept meticulously by their owners because of the need to gather, maintain and save this information for their daily operations. The thing is, the datasets are available, but not used for forecasting and analysis.
Finally, while a lot of data is available it is generally spread out within the organisation. Every department usually has their own spreadsheets or software packages. Meaning that combining the different datasets seems like a difficult job.
Solution
So, what should organisations do? The key is to find the right person within the organisation and give him or her the time, resources and proper responsibility for a data-mining project.
Who is the right person for the job? Every organisation has one: someone who is data-savvy. With Excel skills and just enough database knowledge to perform a few basic analyses. Since the true value lies in combining the different datasets available, give him or her the tools to retrieve data from the available sources and build a clean, well-prepared dataset. Once those elements are in place, there are two routes the organisation can take. One is examining broad strokes and the other is to try to formulate specific hypotheses. If executed properly, both these routes can be interchanged and, in the end, provide a steady basis to build on.
The first route; a number of basic analyses:
- Percentile scores
Dividing the customers into ten tiers will give insight into customer segments, possibly undiscovered previously. The tiers can be based on any characteristic of choice; e.g. no. of products bought or average revenue/cost. After the organisations have been assigned their specific tier the follow-up analysis could be: what does the organisation make, on average, for each percentile? - Customer segment insights
If the organisation has a definition of the most important customer segments, has this definition ever been validated? In other words: is it correct, is it still up to date? Data analysis can provide an answer to that question. Starting from the initial criteria on which the customer segments were divided, the organisation should try to reproduce the same customer segments using the data. If this exercise is successful, the customer segments have been validated. If not, previously unknown customer segments have just been discovered. This is what's called a win-win; the analysis will provide valuable information in either way.Once the customer segments have been (re)defined or validated, the information can then be compared to the overall costs made. The result of this analysis should give insight into a subset of customers that make more (or less) money than previously assumed. Armed with this information, the organisation can then recalibrate the service offering, pricing or revenue structure to increase operating revenue or adjust service levels and the corresponding costs.You'll soon notice that even within customer segments, there are more and less valuable customers side by side. What makes those differ from each other? - Use of service offering
While the focus is often on generating more revenue, organisations often have a lot of hidden cost within their status quo. Specific service offerings can be over- or underutilised, making them far less effective than assumed. Gaining an insight into the internal operations can be of great value then. Added benefit is that generally internal operations already have a large paper trail making it the ideal candidate to investigate further.So, the organisation should evaluate the current usage of its service offering(s)? Is the current use still in line with the initial business case? What is the status of late payments, the use of specific (costly) channels or returns and/or warranty claims? All of these services provided to end customers have a specific price tag. While services are an essential element of a customer-oriented organisation, the pricing of these services should always be balanced against the revenue they generate. Even a superficial analysis of these costs could reveal a completely different customer value, compared to the organisation thinks it knows.The insight that some organisations actually cost the organisation money can be confronting, but having the insight is better than not having it at all. Because in this way, action can be undertaken to resolve the situation. Possible comparison to verify the results: compare this to the report provided by the organisation's business controller.
Hypotheses
Another approach to valorise the organisation's data, is by working with specific hypotheses. A benefit of a hypothesis is that it is fully customizable to any specific situation. A first example is to analyse the 'after-claim behaviour' of customers. A complaint often is a moment of truth in the customer journey and a tipping point for future customer value. Does the claim-handling process turn customers into more loyal followers or even fans? Or does it increase the odds that the customer will leave?
As a second example, organisations should try to gain an insight into the remaining length of the customer relationship. In other words; how long will someone stay after a specific moment, a purchase, or contact with a service employee? Additionally, by selling extra products/services on top of the main offering, the customer value will increase. By staying longer, buying more at higher prices. While this seems pretty straightforward, as always, it is easier said than done. An example of this strategy can be observed by the media organisations in the Netherlands at the moment with quad play offerings.
Case example
Business product supplier, a wholesale chain with a broad product portfolio to both large and small customers. First, 8 different datasets had to be tied together (sales data, financial data, logistics info and the internal sales department sales data. Next, the service offering was evaluated. The firm has three types: online sales, email and phone, ranging from none to very labour intensive. Based on the full dataset, a percentile evaluation was made based on revenue capacity. Furthermore, the costs per service offering were evaluated. This way, an average cost and revenue per customer could be defined. The analysis helped in evaluating the different customer segments the organisation had previously defined and made clear that the definition was ready for an update. By redefining the customer segments and consequently modifying the pricing of their service offering, most customers are now profitably, whereas this was not the case earlier. Moreover, they could now provide truly competitive pricing to their most valuable customers.
Finally
So, with the approaches, the hypotheses and the case example, any organisation should be able to start out with data analysis. Please keep in mind that; every time, the question should be: what am I looking for, and what do I see? Even if in the end it appears that it was, in fact, a correlation and not causation, a valuable lesson will be learned, that something did not work. Document the failure and go on. It is important to evaluate continuously, and to have a number of goals before the search has been started.
