Finally: Data is on the Board Agenda

When data quality gets discussed, it is usually in the context of its impact on local operations, such as lists for direct mail, email or telemarketing. Errors and duplicates reduce performance of campaigns and create waste. Upstream from these areas of activity, less recognition exists about the potential costs and harm that can stem from poor data quality. Yet as companies increasingly move towards data-driven decision making, confidence in the underlying data sets is starting to emerge as a key concern. Unless the single view of the customer is clean and accurate, the company may draw up plans based on false assumptions about the size of opportunities or level of risk. Equally, predictive analytics can only be accurate if the analytical data set is not skewed or full of missing or inaccurate variables. That is why Richard Lees, chairman of Database Group, says: “We are seeing a shift. Marketing has tended to pigeonhole data quality as purely a concern in communications. When companies are making decisions about their product ranging, estate management or risk, however, the quality of their data takes on an extra dimension.”

In the context of multi-channel companies, this shift is even more marked. Companies that set up their dot.com operation as a standalone division have been rapidly integrating their e-commerce with their physical world operations. Ensuring that individual customers do not get identified as duplicates across each channel is essential to this process. “Most operations haven’t had a view of those multi-channel customers,” says Lees. “Yet customers are often researching online, then going offline to buy. You need to know about the journey they are taking.” A good quality single customer view is a fundamental requirement for companies in this scenario. While the focus of SCV projects is usually on the data asset which results from customer data integration, data quality has to be identified as a core and ongoing element of the programme. “There are only two ways to increase your number of customers – get more new ones or increase your rate of retention,” says Lees. He notes that since the size of the existing customer base is a function of attrition rate, companies are starting to realise how vital it is to be able to communicate accurately with those customers. “That is when data quality comes into play,” he says. This is often the first time that an organisation has thought about data quality as a standalone issue that needs to be resourced in its own right. That throws up a range of challenges, from introducing new processes to support the quality of data on the way in to investing in cleaning and validation technology. “Ownership of the data budget has to be decided. If the business has set out to use customer data across a range of strategic initiatives, then it will begin to value data and will therefore need to apply a budget to it,” he says. Return on investment into customer data quality is usually relatively easy to calculate, from reducing churn to increasing multiproduct holding. But data quality projects do not have to be limited to customer records – other data types can demonstrate even bigger returns and equally larger downside risks from errors. Lees points to the example of pricing data for a petrol retailer. Each 1p increase in the price of fuel at the pump can lead to a substantial increase in margin. So having accurate data on your own sales and competitor pricing is essential for the price elasticity modelling that is used to decide whether to increase or decrease pump prices.

Few organisations are in the position to start with a clean sheet and create their customer data assets to the right standard. Instead, there is usually a host of tactical activity to keep data clean at point of usage and, occasionally, at the point of entry. That does not add up to a coherent data strategy, however. What is different in the current climate is that companies are looking for a more solid solution that keeps data fit for purpose, rather than fixing it each time. SCV builds are one opportunity to make that happen.

Forecasting what the management information and analytical needs might be ahead of time is never easy for a client. That does not mean the data architecture should not be optimised for unforeseen requirements. Confidence is one of the elements missing from financial and housing markets at the moment, with the result that few transactions are taking place. Unless confidence exists within an organisation about the quality of its data, business users will not want to act on it. Just ask any sales team that has been passed hot leads, started to make contact and found that many of the phone numbers do not work. This confidence has to be built from the ground up, by ensuring that all data entering systems is validated for both content and format, as well as from the top down, by demonstrating that the board sees customer data as a critical business asset. Fail to get both aspects right and belief in the project evaporates.

The dictum that one bad apple spoils the whole barrel does not apply to a customer database. Every database, even those subject to the highest levels of data governance will have missing, invalid or incorrect fields and records. This does not stop the business from operating or making decisions. Equally, the most accurate data and forecasts may be presented to decision makers without them taking any notice.

In the financial services industry, regulatory pressures are making this concept ever more germane, since decision makers have to be able to demonstrate the grounds on which they made a particular deal. Accurate data is a fundamental when dealing with a SEC or FSA enquiry.

Having data as a visible asset is a very new concept to many businesses, so is putting in place personal and department-level data quality KPIs. Yet these are what connect the owners of data to the impacts on the business of what they are doing. That is why Database Group offers its Cognition process to kick off data quality projects. A team of senior people from customer services, account management, marketing and other functions work through a structured process to identify the critical key performance indicators in the business. These may be simple indicators, like revenue and margin, or more complex ones, like customer profile and lifetime value by identifying the KPIs that really matter to the business and how these are affected by data issues. This leads to a set of ranked actions that can be taken which are known to be achievable, without the risk of running into problems because of missing or flawed data. “The workshop process searches for the drivers of that revenue, for example getting customers to spend more, then it looks at what might create that change in behaviour, such as increased frequency of product use, pricing, or channel – things the business can influence,” says Lees. Some of these drivers may be easier to influence than others.

For example, there may be very little price elasticity in a highly competitive marketplace, but the same company could have a very low level of cross-selling that might be increased. Often drivers emerge that the business hadn’t recognised as central to its operating model. From this, a ranked list of priorities is developed for the business to take action on. At this stage, goals are identified which are achievable based on the resources and data available to avoid the risk of running into problems because vital information is flawed or missing. “Data quality almost always falls out as a key issue. That is the reason why we tie KPIs to the availability of data, in order to identify whether the data allows you to do something or not,” says Lees. For decision makers, knowing that an asset is available for exploitation is the corner stone of confidence.