Data fracking approach
Data is frequently and rightly referred to as one of the most valuable assets an organisation has. Which is why it is frustrating to see it so often wasted and squandered by organisational structures and attitudes that make it impossible to access, interpret and exploit. It is all too often buried in data silos which are seen as too difficult or expensive to mine.
That’s why at bigdog we’ve developed a data fracking approach. Just like real-world fracking, it’s disruptive; it’s a challenge to the established ways of thinking and working. And just like real-world fracking, it has explosive potential to uncover value that to date has not been exploited because of perceptions of cost and difficulty. bigdog’s data fracking approach is lean, low-cost compared with traditional models, and can explode silos where data is trapped, exploiting what is revealed.
"At bigdog we’ve developed a data fracking approach. Just like real-world fracking, it’s disruptive; it’s a challenge to the established ways of thinking and working."
Organisations have a tendency to create and maintain discrete pots of data, in separate databases, which don’t (and often can’t) talk to each other. And keeping data in separate silos gives us only a partial view of it, which often leads to a diminished customer experience and diminished sales and revenues for us.
These may be physical silos, and they may exist for a variety of reasons. They may arise as a result of the way an organisation has grown, for example through acquisition, where existing customer databases have been retained instead of being replaced with an integrated database. They may arise because organisations maintain separate sales and service databases (leading us, for example, to be constantly called by the sales teams of double glazing companies who have already installed all our windows!). Physical silos can occur when a bank uses and brands a white-label credit card from a partner provider, so that bank data exists internally to the bank, whilst card data sits separately with the partner. And they can occur for data protection reasons (and in this case, rather than trying to ‘get round’ the legislation, our creativity should be focused on gaining permission from customers to share their data in order to offer them an improved service).
Data silos can also be ‘philosophical’ rather than physical. A good example of this is the tendency of many large organisations to look at their data in terms of product line silos: all the data may well be sitting in one place, but the organisation (or the individual within it) is looking only at personal loans data, or telephony data, and has no interest in whether or not the customer also holds a current account or a broadband package. We also come up against the concept of ‘product line possessiveness’ in dealing with philosophical silos - the idea that a customer is a ‘loan’ customer and must be jealously protected from invasive contact by the current account or the mortgage sales teams, leading to internal turf wars within the organisation. These are often disguised as frameworks relating to data governance and contact frequencies, with contact windows being bid for on a revenue-only basis and taking no account of customer needs, propensities or motivations.
There are two fundamental problems with data silos:
- Organisations can be missing important information about their best – or potentially best – customers if they are only looking at a subset of the data they hold about those customers. Or they can fail to understand significant elements of their customers’ behaviours. This can lead to badly targeted communications and, of course, lost revenues through missed sales opportunities.
As an example of this, the membership data of a holiday organisation was held on one system, with details of the insurance they had taken out to cover their holidays in a separate database. Bringing the two together allowed us to see correlations between booking patterns and lead times, and the type of insurance product selected (single trip or annual), which informed a behavioural segmentation within the membership – leading to better segmented and targeted communications to the members.
- The customer experience can be deeply flawed as a result of data silos. To the organisation, there may be a logical or a business perspective in keeping white-label credit card data separate from main banking, or in seeing someone as a loan customer (rather than a customer of the bank). But of course, for the customer, their perspective is very different: they only see one organisation, one brand, and to them that’s a brand acting in a very disjointed manner. Customers know when they are valuable to a brand: it’s extremely galling for them when the brand doesn’t see that.
If an insurance company, for example, only looks at customers within product lines, the customer in each case is only a ‘single product’ customer, as a result of the way the company thinks. The customer, of course, may have several products and may be irritated to be only offered a standard discount or incentive when considering taking another product from that insurer; they expect to be rewarded for their loyalty, but the company has organisationally made itself blind to that loyalty.
"Data fracking obviates the need for large-scale systems development or the need to change the way the organisation works, before the financial value of the approach is proved."
The vital task is to challenge these silos, where they exist, for the good of the brand’s revenues and reputation. We need to make the case for bringing disparate pots of data together, whether this is by physically creating a unified database with a single customer view or by challenging the product-centric approach to customer communications and replacing it with a customer-focused approach.
Of course there are obstacles to doing this, primarily vested organisational structures and data management systems and logistics, and there will be howls of “That’s not the way we do things!” and “But the cost will be prohibitive!” It will be essential to gain high-level support within the organisation, and to do this we will need to demonstrate the value in dismantling the data silos in a way which shows the increased revenues and increased customer satisfaction this will deliver.
The way to do this is by data fracking: carrying out small, disruptive tests and pieces of analysis which can identify seams or pockets of data value where traditional wisdom (or traditional less-than-wise thinking) believes that no value exists. Just like fracking in its traditional sense, data fracking is a small-scale and/or localised tool to identify a much bigger opportunity.
The data fracking approach involves:
- Taking samples from each of the data silos which exist and merging them into one temporary database. In the case of the philosophical silos, the data will all come from one place in any case; for physical silos we simply need to agree a model for matching and merging the different data sets.
- Create analysis records for each customer - for example, by creating summary fields for number of products held and overall value, flags to indicate individual products, and so on. Use these analysis records to carry out multivariate profiling and segmentation modelling to identify what high-value customers look like and how they behave. For example, which product is the most valuable acquisition vehicle for long-term profitable engagement.
- As with real-world fracking, which uses additives or catalysts to create reactions, data fracking can involve external data sources to augment the data and enhance its value, or it can develop creative hybrid values as they are identified in the data modelling, which help connect or reflect data in new ways.
- Match the outputs of this analysis back to the existing business targeting models to identify the disparities – which of the high-value customers (or potentially high-value) has the traditional approach been missing.
- Convert this disparity into lost revenue based on something such as a previous communications plan: what would we have additionally gained by including the ‘missed’ customers.
- If the findings of the step above suggest that there is indeed value in bringing the data out of its silos and integrating it, the next step is to develop a test communications plan to run longitudinally alongside the normal communications plan for a period of, say, six months or a year. Ring fence a small (but statistically significant) group of customers to be included in the test plan, with a comparable control group in the main plan. At the end of the period, compare the test and control groups not only in terms of response rates and revenues but also in terms of brand engagement and customer satisfaction metrics.
- Use the learnings from the analysis and testing to support a business plan to justify any technical or organisational changes required to tear down the data silos and replace them with an integrated, customer-centred approach.
Most of the above is not rocket science: it’s a very traditional data-driven approach to marketing. But what is different is the determination to frack: to model samples of the data manually in order to prove the worth of integrating it.
And, of course, data fracking need not only be a temporary means to the end of proving the value of larger scale data reorganisation. It can also be a semi-permanent, agile and, again, lower-cost tool which can focus on singular tasks or campaigns. As above, the approach would establish a fairly large analysis platform, offline from the main silos, import data, analyse, export an actionable data file. But as a longer-term platform, it can be fast, cheap and accurate like the Zipcar solution that avoids massive capital expense. Massive capital expense that potentially locks in another round of obsolescence five years down the line. Fracking becomes a permanent way of thinking about data while your underlying systems architecture continues on its glacial evolution cycle.