March 2004

By Robert Rich, Ascential Software Corporation

CRM projects don't always turn out the way they're supposed to, despite the best intentions of every department and individual involved. These implementations fail for many reasons: outdated business processes and models, faulty project management and inadequate software applications, but the most insidious risk factor for CRM failure is the quality of the data underlying the initiative.
Data quality problems often stem from minor, almost undetectable roots. It might be the four lines of name and address information intended to print mail labels in your legacy system that you need to load into your new CRM system, or it could be the fixed fields in the old order entry system that truncate values or cause the data entry operator to type into the next field on the screen (regardless of what it's supposed to be used for). Maybe it's the creative way the service organization used the status or comments field to capture all that valuable information about your customers that is supposed to be the centerpiece of the new application. Simply put, your CRM system can be only as good as your data; success depends on your organization's ability to ensure that data are available and reliable.
Consider this scenario: you're a vendor of electronic components and you've sold your products into multiple divisions of a $40 billion manufacturer, but have only 20 customer records for the company. They're all apparently from one division, and half of them are duplicates. Some records show the same name but different addresses and, even worse, you have no context for the relationship history with contacts in the other organizations using your products. This disjointed view will foul your CRM system, tarnish your professional image, cripple upsell/cross-sell opportunities and threaten the entire customer relationship. This is just one of a hundred ways bad data can make you look foolish.
Consider another example.You're a financial services company with millions of customers. You want to target high net worth clients across all investments, such as equities, cash, CDs, household accounts and lines of credit. When the data have come together from different systems, however, many e-mail addresses or ZIP codes are missing, and 20 percent of the fields in the CRM system now have conflicting or missing data. How do you execute an effective direct marketing campaign when you are hitting some customers five times and missing others entirely?
Or, if Jack Jones phones in and your customer service representative (CSR) pulls up three records for him, how can the CSR readily determine which information is correct and deliver a positive experience that leverages past history to sell Jack something new? Does the CSR know how profitable the customer on the phone is, how he or she has responded to upsell offers in the past, or whether your current promotion brings the customer value?
Is your call center trying to sell your best customers products they already have? With customer information becoming obsolete at a rate of two percent each month (25 percent each year), are you spending $500,000 or $1,000,000 annually sending mailings to addresses long since abandoned
Continued...
'Data Quality' continued from page 36
or to people who don't work there anymore? Have discrete locations had more than one catalog shipped in the last big season? Do your client-facing apps help you determine how many unique customers you really have, and which 10 percent are most profitable? What is the true lifetime value of a customer across all purchases, and what is the cost to your company if the best five percent of your customers are misunderstood?
Do you have a common way of identifying customers at their point of contact? Does the Web site use a different customer database than point-of-sale? The sales force? The support center?
The bottom line is that fixing data problems can go a long way toward recouping your CRM investment and saving your customer relationships. With that in mind, here are 10 ways to shore up your data, and make you the kind of CRM statistic you want to be: the hero who made CRM pay off quickly.
Know before you go. Before connecting pipes haphazardly and flooding your new CRM apps with raw data, profile all source systems to capture data content, quality and dependencies. 'Reconnaissance' software tools can automate this tedious, error-prone process, and they cut as much as 90 percent of the time off manual profiling in the initial deployment. Plus, detecting and correcting problems early dramatically reduces the costs and risks of failure down the road. Best to look at each source of data individually to reconcile how a particular business function operates. That puts you in a great position to harmonize data across many sources.
Build in data quality. First, standardize all relevant data (e.g., customer names, contacts, addresses, relationships, product information) to ensure consistency. Often that means taking the time to get into the bowels of the current systems to understand how customer-related data are captured in the database. That's also a good time to get agreement on 'simple' things like 'How do we define customer?' Then, match records across sources with statistical logic to eliminate duplicates and create relationships for a 360-degree view of all customer activity. Consistent procedures, processes and business rules around data quality must apply to batch data cleansing and to all new data entering the company in real-time through Web transactions and other mechanisms.
Weave inside data with external reference information. Anticipate marrying historic data, such as customers' transaction histories, with real-time information, such as order status and trouble tickets. Add enrichment information like financials and demographics about people and businesses that deepen your 360-degree view. That will put you in a position to better determine who is most likely to buy which product, respond favorably to an offer, cancel a service or churn.
Share common meta data. With data from separate sources often having conflicting meanings, it's critical to share meta data across end-to-end integration processes to have one enterprise data 'dictionary' for all users. Make it possible for your organization to agree on what your master data mean and store the information about the data and related processes in one place where everyone can get at the information.
Abandon hand-coding. Tools exist to automatically extract data from disparate sources, re-purpose and transform them into the required form and load the information into the proper target application. That makes the process for manipulating data understandable, consistent and repeatable. Manual coding of these steps drains time, money, morale and the opportunity to focus limited resources on something more valuable. Moreover, hand-coding limits your ability to quickly respond to changes in the business.
Architect for 'right-time.' Take the time to understand how your organization 'touches' your customers and prospects. Find out where and how new customer accounts are set up. Use a data integration engine with one set of business rules for both bulk and real-time processing, so that all customer data are managed consistently, regardless of channel or source. Apply one set of business rules consistently wherever customer related information is processed.
Implement a highly scalable foundation. Give yourself room to grow and the flexibility to adapt to changing business requirements. Managing massive amounts of customer data in short time intervals keeps your new CRM system up-to-date with the rest of the organization. Use parallel processing to accommodate growing data volumes and allow integration tasks to scale without requiring costly re-writes.
Ensure interoperability. Since a number of tools are required to integrate and homogenize data from disparate sources, teams should avoid the additional burden of integrating the integration tools. Instead, give consideration to suites that offer data quality, profiling, integration, meta data management and other critical functions on one interoperable platform, so you can integrate the enterprise.
Embrace open standards. With enterprises needing to do more with less when integrating best-of-breed applications with legacy systems, use of Web services, J2EE and other open standards makes it much faster and easier to connect multiple data sources with CRM systems and to extend data integration capabilities throughout the enterprise. Find your enterprise architect in the systems department and see how your CRM application leverages the corporate architecture.
'Futureproof' your investment. Bringing together data and standardizing and re-purposing them is an ongoing process. Your CRM needs tomorrow will be vastly different from those of today. Get religious about data quality. Put appropriate customer data management programs in place to maintain the integrity of data over time, taking into account inevitable growth in the variety and volume of online connections with customers and partners. That way, you can transform your data into a driver of profits, not a CRM killer.

Robert Rich is senior product manager for ProfileStage and QualityStage for Ascential Software Corporation (www.ascential.com), a provider of enterprise data integration.

If you are interested in purchasing reprints of this article (in either print or HTML format), please visit Reprint Management Services online at www.reprintbuyer.com or contact a representative via e-mail at reprints@tmcnet.com or by phone at 800-290-5460.

For information and subscriptions, visit www.TMCnet.com or call 203-852-6800.

[Return To The March 2004 Table Of Contents]