Data integration: putting the pieces together
By James Lawson.
3 Aug 2011: The sheer volume of data and the minutely detailed work required to build a Single Customer View (SCV) database can be daunting, particularly when bringing together online and offline data. Though every project is different, bearing some simple planning rules in mind, identifying possible quick wins and avoiding the temptation to take the “big bang” route should all help avoid a stalled project. Let’s hear what the experts have to say.
Take the smart route
“It’s surprising how few companies know the rules and structure of their databases,” says Martin Dawson, business development manager at Abacus 360. “We often get vague explanations and people say, ‘you’re the data experts, you’ll work it out’. It can turn a project into a massive exercise in trial and error.”
It’s essential to have some kind of document that lists the data structure, tables, variables, rules, codes and so forth employed in any databases a company employs. Without this information, the pain levels of any SCV project will rise significantly.
A detailed data audit is invaluable too, especially if the database documentation is poor or missing altogether. The first evaluation is not a job for external migration providers; far better to have a solid audit in hand before entering the tender process at all. Along with a detailed requirements specification, prospective suppliers will be able to use the audit to quote accurately on time and cost, and this information will help with many other parts of the project. That’s if an external provider is to carry out the project of course.
“Assuming all SCV business objectives have been accurately captured, you would expect an externally-developed SCV to offer payback on investment quicker than an internal development,” says Karl-Magnus Wadsack, Strategic Consultant at Equifax. “External SCV solution providers have the experience to ensure that all of the common pitfalls of building an accurate and robust SCV are avoided.”
Being clear at the outset about project goals helps avoid trouble later on. Why is the work being carried out? What systems will it support and what variables will be needed within these systems? Discussing expectations in detail both before any engagement and also in planning meetings with a supplier is a far better approach than periodically asking for new features as the project (very slowly) progresses.
“If the initial scope is too broad, the time to deliver can be high, and the project may fail to deliver its objectives early enough,” says David Barker, Head of CDI, Europe at Acxiom. “However if the scope is too limited, it will not meet the needs of all stakeholders, who will continue to use other systems and applications with potentially a different view of the customer.”
Given that the concept of a customer-centric database clashes head on with the power structures inherent in traditional product-based job and departmental roles, it’s best to try to see the crash coming and avoid it. For example, a new customer value management system that only offers high margin products to the best prospects – so reducing low margin product sales – will not go down well with those whose job depends on maximising sales of those low margin products. Setting out to define and agree “rules of customer engagement” early on can help defuse potential conflicts.
“Getting the organisation to put aside traditional business and departmental delineations and buy into the idea of data integration and master data management can be challenging,” says Penny Hutton, Strategy and Planning Director at Eclipse Marketing. “There is no textbook ‘best way’ to achieve this, it’s a case of getting stakeholders to look beyond immediate departmental goals and understand the wider organisation and customer service benefits offered by sharing available data.”
This political aspect of an SCV is often overlooked but can completely derail a project. According to Hutton, “Stakeholders can feel that a move to a centralised system will be a retrograde step for their business area and that they need to ‘protect their data’.”
She advises that companies should make great efforts to understand those concerns which might be completely rational. For example, they may think that the integrity of marketing permission data could be damaged by the move to a central SCV. In this case, evangelising about the improved data quality the system will offer and the bottom line financial benefits to each department as well as the whole business could help overcome these concerns.
Definitions are another key part of early-stage discussions. What constitutes a customer, enquirer, prospect and so on for this particular project? What defines an accurate match such that two records can confidently be combined? Should customer data be stored at individual or household level?
“Agreeing the keying protocol and rules is often managed by a few technical people,” says Marie Myles, Director of Consulting at Experian Marketing Information Services. “It is necessary to make sure that business users are involved in this as they understand the data from an actionable point of view.”
Striving for clarity on the goals of an SCV should mean that further planning details will emerge, and so more questions. If running customer analytics is important, then what sorts of analyses will they be? Which variables will they require over what time period? A company may already have all the data on its source systems or might have to start collecting certain items in a hurry – or it could simply buy in the variables it needs.
“Decisions need to be made on what to use to build the single customer view,” says Myles. “Is it done all in-house using internal data or do you use external software with the matching logic already built and access wider reference data? This decision comes down to scale, skills in-house, depth of internal data and budgets.”
Running staged implementations is now accepted as the best way to drive early value from an expensive system while lowering the risks of total failure. Again, how staged a project can be when it comes to data is a moot point. Contact data for the principal channels will be essential, as are suppressions and basic demographics like sex and date of birth, but which other variables should make the first cut will again be down to the goals for the system.
“Contact information is an obvious priority for a mailing database, as is hard data: what they have spent with you on what and when,” says Dawson. “Other data depends very much on the business and its sector. What do you usually select on for campaigns? Channel-specific suppression data is crucial, otherwise it will cost you a fortune in customer services.”
As an example, an initial SCV to support customer cross-selling will not require expensive external datasets if the internal data is rich enough and reasonably complete, while the variables used to drive RFM and LTV scoring might well make the first cut. Stage two of the project might involve adding in proprietary lifestyle data codes against the customer records to help with profiling and to better link information derived from customers to prospecting work.
“There are always opportunities to deliver the solution in phases, to ensure that quick wins can be achieved without integrating data from all sources of customer information within the organisation,” says Barker. “These will vary in each project, but a focus on improving the data quality in the SCV will usually deliver savings to the business units using it through reduced waste in duplicated activities, marketing spend and so forth.”
Linking online behaviour to an identified individual is a classic online challenge that also applies when building a single customer view, as is dealing with the sheer volume of data that web servers and web analytics generate. For example, raw clickstream data relating to an anonymous website visitor is potentially useful, but does it really belong in the SCV? Aggregating this data or holding a six-month history elsewhere and then bringing it into the SCV when the person “reveals” themselves may be a better option.
“In many ways, bringing in online data is one of the easiest parts of building a SCV,” says Daniel Cross, Director of Strategy at Lateral Group. “It tends to be clean because it’s machine-generated.”
Because of its value in revealing trends and relevant customer behaviour, Cross recommends holding all email clickstream data at individual level despite the high data volumes involved. Data items like time of clickthrough, where the person subsequently navigates to and the propositions involved all feed into honing subsequent email targeting as well as informing numerous other activities. Matching email clickstream files is usually simpler than web-derived data as there will be a valid email address as well as consistent, predictable data.
“We are now finding new players in the market that tend to use email address as the main piece of data that identifies an individual,” notes Myles. “Email addresses are more volatile – people have more than one email address and change these more frequently than they move. Ultimately it depends on the business' priorities, requirement for address and demographic data and approach to data capture.”
Extracting email clickstream data from an ESP can be problematic so it’s best to clearly define in advance who owns what at the end of the contract period. Thinking ahead is the key again here: the website must be set up properly to generate usable tracking data, while some analytics tools like Google Analytics can be tougher to extract information from.
“Google Analytics is great and free,” notes Cross. “But it’s harder to move the data into a SCV. There are ways around it as long as you consider it in advance. You need to get an idea of what you want to track, how you intend to do it and what you are going to use the data for.”
Again, the desired analytics will define which data types are needed and their format. How far back in time analysts want to look will affect how often web log files are deleted and purged, for example. Likewise, which variables are required (and in what form) to explore any “halo effect” linkage such as that between the PPC keyword used and the subsequent LTV of the buyer?
A final consideration is the ongoing cookie permission debate. The UK Government has set a date of May 2012 for companies to work out a compliant way of obtaining permission to set tracking cookies; any radical change here may affect the data a company collects on an individual and how can be linked.
Invest in planning
Other solid advice includes building in contingency to the budget to cover any unexpected work. Also periodically “fixing” usable database builds means it’s possible to roll back from a problem to a working version of the database. Likewise, testing a new system thoroughly alongside the old system means that the old system is still there as backup. This period of parallel running should be used to train up staff – a remarkably large and detailed task on a large system – and to go through every possible permutation of system use before the old applications are switched off for good.
“If you don’t know where you are starting from, how do you find your destination,” concludes Dawson. “It delays everything if there is no-one around to check and answer questions. A good provider will hold your hand throughout a project but the client needs to invest time and energy too.”
20 May 2013: In the third of our special series of articles profiling the database marketing industry, we catch up with Hopewiser, Alchemetrics, Callcredit and greenstone data solutions.
22 May 2013: Insurance provider Ageas has appointed Mark Hanson as Marketing Director for its two over 50s insurance brands, RIAS and Castle Cover.
20 May 2013: Comet Global Consulting, the customer interaction expert, has secured an additional two-year deal with Sky, to further develop and support the company’s customer relationship management and customer interaction systems and operations.
21 May 2013: — Experian has launched a new online credit risk and marketing portal putting a wealth of vital business information at the fingertips of UK SMEs.
8 May 2013: Hector Vass, former Systems Architect in the Customer Lifecycle and Optimisation team at the Lloyds Banking Group, has joined marketing analytics agency Metametrics as Chief Technology Officer.
8 May 2013: The Institute of Fundraising’s Special Interest Group for Insight in Fundraising has announced the shortlist of finalists for the Insight in Fundraising Awards 2013.
8 May 2013: ghd, the UK hair-straightening brand, has seen conversion rates soar as a result of improving the shopping experience on mobile devices using data gathered through Celebrus Technologies’ software.
7 May 2030: Maileva, a European leader in hybrid direct-mail solutions, and Selectabase, a supplier of marketing data, have joined forces to offer marketers what they call "a unique, easy-to-use and cost-effective direct-marketing solution".
7 May 2013: Really Simple Systems, one of the world’s largest providers of hosted CRM systems, has announced availability of an interface to Sage One – Sage's online accounting system.
15 Apr 2013: Following the launch of the Smartdepart suppression file last month, data and targeting specialist Wilmington Millennium has now launched Smartlink, a new goneaway database for direct marketers.
15 Apr 2013: Celebrus Technologies has deepened its partnership with Teradata through its new ability to feed individual-level digital channel data directly into Teradata Real-Time Interaction Manager.
15 Apr 2013: Data8, the data enrichment specialist, has added to its suite of real time data validation applications with the introduction of bank account validation.
15 Apr 2013: The consolidated revenue of Weborama, the European leader in the management, distribution and tracking of targeted online advertising, grew by 5% in the first quarter of 2013.
5 Apr 2013: Kwik-Fit Group has appointed customer engagement agency Lateral to handle the automotive company’s customer eCRM programme.
5 Apr 2013: It is predicted that, over the next year, the human race will produce 4 Zetabytes of new data, which represents about 1 quintillion new objects. Are marketers ready for this deluge and is it an opportunity or a threat, asks the IDM?