Hanging on a wall in the office of Simon Kaffel, the data and analysis director of marketing strategy at BSkyB, is a huge and complex diagram, which has an idiosyncrasy to it only disciples of marketing are likely to begin to grasp. “It is a map of whereabouts data is in our operation and contains all the sources of information in the data warehouse,” says Kaffel. It is said only two people really understand the map and Kaffel admits it gives him nightmares.
There is also a critical point about the way the diagram has been drawn – it is not centred on the customer. For that, Sky pulls data from the data warehouse into its Olive system, managed by SkyIQ, to create a single customer view (SCV). “We don’t have to touch the data warehouse itself,” notes Kaffel.
What this approach underlines is the distinction between having data on the customer and having a view of the customer. For years, organisations have pursued the goal of pulling every instance of customer data into a single place to provide a 360-degree view. Sky has tackled that by pulling information out of a very diverse range of sources into a separate system, rather than trying to build the integrated view in its operational data warehouse.
Kaffel emphasises that neither the sources nor the SCV are static. “It is evolving month to month. It is down to the business to decide how to leverage as much as it can out of the data asset. The technology needed to do that has moved on,” he says.
Far more complex
The operational data warehouse was originally built at the start of the last decade. “Ten years in data management technology is a lifetime,” he says. “The amount of data you are able to process, the complexity of extraction, transformation and loading processes and the data itself are all far more complex now.”
As a business, Sky has also moved on – ten years ago it did not offer gambling, for example – so the needs and sources of data continue to expand. Collating that information to provide each business function with the data required is the very heart of what Kaffel’s team does. It has also helped to drive the media company’s success.
The pragmatist in me says not all data is created equal
“We’re not standing still. It is not as if the single customer view is set in stone and cannot change. We are constantly moving data in, building models and trying to meet the demands of a business that is changing all the time,” he says.
When it comes to getting a view of the customer and what makes them valuable, you need to look at everything available. No single item of data stands alone – the combination is what gives it power.
“I do believe in bringing in as much information as possible to the system,” says James Morgan, head of the customer intelligence centre (CIC) at O2. Even so, this is subject to limits based on business rules. “Some of the data, such as users of smaller gaming apps, may only be on a few hundred customers. Building links to those third parties, getting legal agreements, running data transfers would have a cost that is not warranted. We do have to draw a line at some point.”
Bigger vs value
The opposite side of that argument is where very large data sets exist that require significant processing power, but where the value to the business is not necessarily clear. “We have to be careful about trying to put anything into the data warehouse where the value is not there. We can’t know the value of every bit of information, so we have to have guidelines,” he says.
CIC has to balance the role it has in enabling business ideas by providing accurate data with its other function as gatekeeper to the data and cost setter. A clear change management process exists that is followed when a new data feed has to be created.
“We cost up proposals and may say, ’do you really want to spend this on a new data feed, or do you just want to know who has got a particular product?’,” says Morgan. New product launches are a classic example of where the existing single view of the customer is likely to be stretched and careful management is required.
Morgan says a phased approach is often the most effective. On day one, what the line of business needs to know is which customers have bought a product and whether they are active. In a highly dynamic and innovative sector such as telecoms, there are high numbers of product launches, not all of which are successful in the long term. By limiting the initial management information provided, costs are also constrained.
If a product establishes itself, a much stronger business case can be made to establish this data as a permanent part of the SCV.
What this approach underlines is the distinction between having data on the customer and having a view of the customer
“The risk is to over-invest in something you don’t know will take off. It is also possible to get a product to market without any information, but if you do that, you have to be certain that the risk premium is small,” says Morgan. For an organisation that is oriented around customer-centred decision making, that would be an exception, rather than a rule.
The concept of value-driven data integration has grown in stature. Data managers have always resisted the pressure to simply keep adding variables to a database, but it has often been a struggle to explain why. By putting an onus onto the business function to prove that value will be returned, a greater degree of precision and purpose is brought to bear on the SCV.
With data volumes rising dramatically, filtering requests in this way is becoming an ever more important part of the task. Asked about whether this value can always be demonstrated, Stephen Boyle, database marketing manager at Reader’s Digest, answers, “the data analyst in me says yes. But the pragmatist says that not all data is created equal and it can be hard to justify being integrated across all data sources.”
He says the critical step is to identify which data items are fundamental for the effective management of a business process. “The key point is to make judgements on how that data can work across different channels. That, of course, is easy to say, but harder in practice,” says Boyle.
Multichannel marketing’s relentless expansion of its touchpoints is undoubtedly putting pressure on existing templates for single views of the customer. Whether the SCV can be adapted to include new data from these will depend on a range of factors, such as the underlying data model, the matching points available in new data to allow it to be integrated to a customer file, or the completeness of the information.
Stewart Robbins, head of customer insight at E.on Energy, says: “Personally, I think you have to make decisions about what variables to include. For example, if you were a mobile network looking at the number of calls a customer had received, that is in the operational data, but to pull a list of all of them into the SCV won’t support your marketing activity in the right way.”
Aggregating that data into a code or flag showing if that call activity has declined or risen in a given period immediately transforms this data into useful insight and avoids overloading the marketing database.
“Processing power does cost money, but you could decide as a marketing-led organisation that the SCV is worth investing in if it drives revenue,” he notes.
Political issues also need to be borne in mind when specifying the scope of an SCV. Marketers have always struggled with the issue of who “owns” a data item they might want to include in a customer database. Where the company operates in a network or matrix model, this extends the problem.
“Our business partners want to know what we are doing with their data,” says a senior marketer with one financial services provider. “In some countries, such as France and Spain, we could do anything we want to, whereas in others, such as the UK, Germany or Sweden, they certainly want to know exactly what data we are taking and what is being done with it.”
His company provides products on a white-label basis through a wide range of business partners and affiliates. It then manages customer relationships and cross-sell activity.
“That can make it complicated to decide the ROI when you go to a partner and say you want to include some information in the marketing database,” the marketer notes.
Single view of the customer is a vision, more than a reality. But it is now possible to get something very close to the ideal, provided the right decisions are made. All it takes is a combination of careful planning and belief.
Data and analysis director,
marketing strategy, BSkyB
There is always the danger that as a business, you want as much information as possible in your single customer view (SCV). But it has got to be relevant and follow data protection principles. For Sky, it is really about understanding a subscriber’s engagement with the business – the more services they take up, the more loyal they are and the more engaged they become with the brand. The challenge is building a SCV that is as scalable as possible. But if you have the extraction, transformation and loading routines set up
correctly and have the right level of data, you can build a very effective view of the
Database marketing manager,
Clearly, there is a large and possibly expanding amount of data generated from digital and mobile channels. I’m not sure there is a clearview of what is the key data to collect – we seem to be mesmerised by the amount and ability to store this volume, but less sure how to make this data work in practice. I also think there is a challenge in ensuring we get best se of this new data allied to the existing data sources, typically generated from offline activity. In many ways, this is the next big leap forward, but there is a danger that we end up swamped with data in the process.
Head of customer intelligence
For us, it is not about trying to put all the original information into a centralised solution. I would shy away from the idea of an SCV. We look at it as a customer-centric information source that we can exploit appropriately to add value. A single view of the customer doesn’t work for every channel, but each channel needs to have access to the whole customer picture to understand their value to the business. That doesn’t mean creating a single, centralised version that suppresses other views. It is about creating careful links and understanding the formal joins as opposed to inferred data.
What is a single customer view?
Organisations capture data on customers in a wide range of operating systems across the business. All of this data may be stored in a
central data warehouse, but it only becomes a single customer view when each piece of information relating to the individual customer is brought together under their name, rather than an account number, product or other view used by the business.
How is the data brought together in a single customer view?
Data has to be taken from operating systems or the data warehouse using extraction, transformation and loading routines. These identify relevant items of data, standardise them and push them into the single customer view. Records then need to be matched together using a range of different techniques, from identifying common elements (surname, address) to giving customers a unique identifying number.
57% Proportion of companies that say they have not yet built a single customer view which integrates all pieces of data they hold on a customer into a centralised repository.
32% Percentage of companies that hold more than 5 million customer records in their marketing database, with 22% holding between 1 million and 5 million records.
76% Companies saying that the single customer view is extremely important to their organisation. Another 21% say it is quite important.
21% Level of extreme satisfaction with the quality of information held on the SCV, with 55% saying they are quite satisfied with the data quality.
brand in the spotlight – Q&A
DOMESTIC & GENERAL
Data Strategy (DS): What range of data sources do you deal with?
Dean Turley, director of database marketing, Domestic & General (DT): Our business is still very traditional – online and digital channels are increasing significantly, but we still only make less than 5% of our transactions that way. The profile of our customers is very traditional: Middle
England, older consumers. That gives us much less of a data challenge. Where we do tend to use the web is as a response channel to traditional direct marketing, particularly direct mail. We are seeing an increasing number of our prospects and customers going to the web to take up an offer.
DS: So has Domestic & General built a single customer view (SCV)?
DT: We do have an integrated view of the customer. We have a transactional database that all our telephone operators and sales agents go into every day. Off the back of that, we create an SCV that is integrated in near real-time, 24/7. There is a permanent extract, transform, and load process between the two constantly pulling data. We also hold a full contact history in that customer database.
DS: What are the challenges you face integrating customer data?
DT: We have a lot of overlapping customer cycles with a lot of information coming into the operating systems from different sources.
We can talk to one customer many different times about the same appliance, because our core product is an appliance warranty. We want to know about that person and their washing machine, tumble dryer, dishwasher and may learn about each one several times over.
DS: How does the business see a benefit from its SCV?
DT: Every offer is pre-rated before it goes out so we can identify 100% of responders. That doesn’t create much of a data challenge to track marketing response. The key benefit for us is that at the point where a customer calls to take up an offer, we can be 100% certain what price and offer has been made to them. We are able to track back that response to the initial marketing strategy so we can be certain about the marketing effectiveness.
DS: What is your biggest data challenge?
DT: Our biggest challenge is the volume of data. We have a customer database in the tens of millions and data on more than 200 million appliances. Our system has been around a while. It is very good at the highvolume, high-frequency type of marketing it was originally built for. The challenge is to make it more flexible to allow us to do things differently. The demands for data are changing. There are certain things we have a very good handle on, such as the drivers of marketing performance, campaign performance, channel performance. We have less information about our customers – often all we know about an individual is an appliance they have.
top tips you need to know
- The single view of the customer (SVC) should operate separately from the data warehouse to avoid conflicting business needs and cut the risk of changing a piece of original data.
- SCV should be designed to be flexible and allow extra items of data to be included in the future – these are usually impossible to forecast when the SCV is first designed. Not every item of data about a customer needs to be included in the SCV. Many pieces of information can be aggregated or banded to avoid overloading the system and reducing its performance.
- If someone asks for information to be added to the SCV, ask for a business case that will justify the investment to include that data based on a positive return.
- Be prepared for political battles about ’ownership’ of the customer or a piece of customer data – SCV is often seen as a threat by individual functions.