How Clean Is Your Database?

Take the test: Maintenance benchmarks revealed

July 2005 By Hallie Mummert And Irene Cherkassky

multipage

2. How often should you update your files for householding information?

a.) monthly
b.) quarterly
c.) once a year
d.) as new records are added

3. How often should you conduct NCOA processing on your database?

a.) monthly
b.) quarterly
c.) every six months
d.) annually

4. How often should you clean the phone numbers on your database?

a.) quarterly
b.) twice a year
c.) annually
d.) every other year

5. How often should you update your e-mail list addresses?

a.) daily
b.) monthly
c.) every six months
d.) once a year

6. How often should you clean your e-mail addresses?

a.) daily
b.) monthly
c.) quarterly
d.) once a year

7. How often should you process opt-out requests?

a.) daily
b.) every 10 days
c.) once a month
d.) once a year

8. How often should you score/rebuild your models?

a.) quarterly
b.) yearly
c.) every two to three years
d.) every five years

9. How often should you test your models?

a.) monthly
b.) quarterly
c.) annually
d.) seasonally

10. How often should you perform data enhancements to your customer file?

a.) monthly
b.) quarterly
c.) annually
d.) every other year

11. How often should you perform a data integrity audit?

a.) monthly
b.) quarterly
c.) annually
d.) every other year

12. How often should you perform a full data asset audit?

a.) annually
b.) every other year
c.) every five years
d.) every 10 years

The Answers

1. How often should you standardize the addresses on your database?
Answer: c.) once or twice a year

Standardization, along with ZIP code append/correction, is a maintenance process that greatly affects the efficiency of other list hygiene tools and merge/purge efforts. It involves assessing the accuracy of individual address elements, such as city, state, country and postal code. If telesales representatives or order entry systems are not capturing address information according to the format guidelines you’ve established, you could have erroneous, and possibly meaningless, address information in some of your records.

Some standardization is completed during NCOA processing, but not enough to consider yourself done with address analysis, says Lynne Mysliwiec, vice president of analytics, Epsilon, a relationship marketing company. Most marketers would be well served by standardizing their address records every six months to annually, she states, with marketers who house more than 6 million customer records standardizing their files on a quarterly basis.

2. How often should you update your files for householding information?
Answer: d.) as new records are added

“Do householding when you add records to your database,” says Scott Cone, vice president of professional services for database marketing services firm Merkle. “Somewhere between daily and monthly, depending on the cadence of your communication and interaction with your customers and prospects.” For example, if you are a cataloger that mails every two to six weeks, updating your database on a monthly basis may be sufficient.

The most basic benefit of householding your data as it comes in is avoiding unintentional multiple mailings to the same household from the get-go—saving you valuable resources, and your customers the annoyance of receiving too many identical mail pieces.

Householding is just one element of a total database refreshing process that should be done on a quarterly basis at the minimum, adds Tom Young, executive vice president of database marketing services for direct marketing services provider KnowledgeBase Marketing Inc. This includes CASS-certifying your data, which keeps your records compliant with USPS regulations and makes you eligible for postal discounts; running your data through NCOA; deduping your database and assigning your households and residences. “That kind of total refresh of your database should occur at least quarterly,” Young says. “And if you have a significant direct mail program, you probably want to get there monthly.” You’ll be rewarded with more accurate data and more effective models.

If you’re not keeping to the bare minimum of cycling your database records through the above steps on a quarterly basis, you may be losing as much as 3 percent to 5 percent revenue annually, according to Young.

“A lot of people have a database that fluctuates wildly from being very clean and then over time it gets dirtier and the householding isn’t as good, and two or four months later they do hygiene and it’s clean again,” describes Cone. “You want to get out of that kind of cycle. You want to have confidence in your database at all times.” 

3. How often should you conduct NCOA processing on your database?
Answer: d.) at least annually

As a general rule, says Mysliwiec, you should match your housefile against the NCOA database at least once a year; B-to-B companies should do this at a minimum of every other year.

Optimally, however, the rule is that the larger your market, the more often you need to update your file with change-of-address data, she explains. Also, you can take into consideration whether your customer base skews older or younger; in general, younger audiences are more mobile and change residences more often than older audiences.

Mysliwiec urges marketers to consider recent move statistics when determining how often they need to perform NCOA updating. Between 2002 and 2003, she notes, 40 million out of 280 million people in the United States moved; that’s 14.3 percent of the U.S. population, roughly 3.33 million people per month.

It doesn’t take much thought to realize that not staying on top of address changes for your customers results in undeliverable-as-addressed mail, lower postage discounts and flawed models that use geographic data—which adds up to lost revenue and wasted budget investments.

4. How often should you clean the phone numbers on your database?
Answer: c.) annually

Direct marketers who conduct telemarketing campaigns should process their phone number files against the NPA/NXX database once a year, says Mysliwiec. This file helps marketers assign new exchanges to old area codes. Not only does this process help keep customer phone numbers up-to-date, but facilitates better matches on any telemarketing suppression files.

Mysliwiec adds that companies that geo-code their telemarketing lists probably only need to recode their list every couple of years, since the U.S. Census Bureau doesn’t reassign geo-codes very often.

5. How often should you update your e-mail list addresses?
Answer: a.) daily

“Studies indicate that about 30 percent of e-mail addresses will change over a given year,” says Reggie Brady, president of Reggie Brady Marketing Solutions, a direct and e-mail marketing consultancy. In the face of such daunting statistics, it’s essential to update your e-mail lists as often as possible. Adele Ghantous, account director for Young & Rubicam Brands says, “When a customer advises you of a change in their e-mail address, you need to act on it immediately and make the change in your database to make sure your next mailing reaches them.” If your mailing cycle is more extended—once every week or once every two weeks—the update should happen before the launch of your next communication. To supplement these efforts, Ghantous recommends using an ECOA service. The volume of your e-mail efforts and your customer base will determine how often you may want to use such a service.

Regardless of how often you send e-mails, it is best practice to include a link in every e-mail communication that allows recipients to access a preference center where they can update their information. Brady also recommends sending a confirmation e-mail that will both reinforce the relationship with the consumer, as well as confirm that you have a working e-mail address. If your addresses are out of date or incorrect, you’ll generate a higher number of bouncebacks. Too many bouncebacks, and ISPs will consider that you’re spamming their customers, and you may be blacklisted.

Beyond reputation, having outdated addresses also means lost revenue. Brady says, “If a company can get somebody to buy via direct mail and also via online or e-mail, that customer is going to spend three to seven times more with that company than a single-channel customer will.” One more incentive to keep your e-mail addresses as fresh as possible.

6. How often should you clean your e-mail addresses?
Answer: a.) daily

Optimally, cleaning e-mail address lists should happen every day. Ghantous recommends cleaning addresses at the collection point, as soon as a customer opts into your list. “Applying hygiene to e-mail addresses is necessary not only when people register with you the first time, but also when they change their e-mail address, as this is another instance when an address can be misspelled or have typos,” she adds.

One way to cut down on incorrectly input addresses is to have subscribers enter their address twice. Ghantous adds, “You may also want to partner with a service provider that offers e-mail address hygiene services and updates their hygiene rules regularly.” As well, supplement daily hygiene efforts with a periodic general cleanup. Pull a list of the invalid addresses on a monthly or bimonthly basis, and run the hygiene rules against them.

Failure to clean addresses as quickly as possible hinders your ability to reach your customers. As with outdated e-mail addresses, a dirty e-mail list means lots of bouncebacks, and the risk of your reputation being sullied with the ISPs. “What I wouldn’t recommend is waiting until the end of the year, or doing a once yearly spring clean,” says Laura McFarlane, global digital lead, Microsoft account, for advertising and media services firm Young & Rubicam Brands. The bottom line is that following hygiene best practices increases deliverability rates.

7. How often should you process opt-out requests?
Answer: a.) daily

According to current Can Spam Act regulations, marketers have 10 days to process e-mail opt-out requests. However, best practices call for a more immediate response. “An opt-out should be applied to your database immediately,” Young says. “If you have a touchpoint with consumers where they’re allowed to opt-out, in most cases that should be a daily process.”

Variables such as how well your e-mail infrastructure is integrated with the systems that capture opt-out information, as well as the frequency with which you send your communications, will affect this benchmark. McFarlane says, “If, for example, you’re a business and you send out a quarterly newsletter, then obviously your need to update your list may not be as great as someone who’s sending out very frequent communication, say on a weekly basis.” The key is to process the requests before your next transmission, whether that’s once a month, or once a day. By doing so, you’ll ensure you are reaching people who want to receive your communication.

If you don’t take care of your opt-out requests in a timely fashion, there is a good chance customers will flag your communications as spam, which in turn can lead to being blacklisted by ISPs altogether. “There are still people who are doing that [opt-out processing] monthly or the bare minimum, to almost meet industry standards,” Cone warns. His advice is not to put your company’s reputation at risk and process those requests as soon as possible.

However, e-mail opt-outs are only part of the story. Young says, “The other process that needs to go along with that, to create the rest of the suppression information that you need, should be occurring on a weekly basis.” This includes checking your records against the National Do-Not-Call Registry, wireless suppression lists to avoid calling land-line numbers ported to wireless numbers and the DMA Mail Preference Service file. “For instance, updates to the Do-Not-Call list come out weekly,” says Young. “While the government says you have 30 to 60 days to apply that, you really ought to be refreshing your database with those permissions on a weekly basis.” Compliance helps you avoid fines and penalties, and even more importantly, maintains your brand’s good image.

8. How often should you score/rebuild your models?
Answer: c.) every two to three years

The frequency with which you refine your models typically depends on the shelf-life of the data variables used to create them, and the business assumptions made about what the models will predict. For example, your assumptions about the average age range of your best customers could change over time.

But, ideally, the effective lifetime of most models is 18 to 24 months, says Mysliwiec. “If you don’t re-estimate every two to three years, then you’re missing out on [performance] opportunities.”

Other factors that could affect the validity of your models include gaining/losing a competitor, introduction of a new selling channel and changes in the way your data provider collects data elements.

This benchmark holds true for both B-to-C and B-to-B marketers, says Mysliwiec.

9. How often should you test your models?
Answers: b.) quarterly & d.) seasonally

About 85 percent of companies are pretty stable, asserts Mysliwiec, so model testing can be performed periodically, say quarterly. If your company is in a rapidly changing industry or you are trying to model to improve online business performance, you might need more frequent testing to accommodate more of the continual change associated with such environments.

Direct marketers who have a highly seasonal business will want a model that reflects this selling period, and thus must test this model seasonally, too.

10. How often should you perform data enhancements to your customer file?
Answer: b.) quarterly

For most B-to-C marketers with a customer base of less than 10 million, primary data additions—such as age, marital status, years at present address, income, presence of children, etc.—should be kept current. These data sources should be updated and replenished at least quarterly, says Mysliwiec.

On the B-to-B side, she adds, core firmographic and demographic data needs to be updated annually.

All other data sources should be evaluated for their ROI potential before appending at all. Mysliwiec advises marketers build a test overlay for each model with such additional data elements to see if the response generated would yield enough revenue to pay for the cost to purchase, apply and maintain the data. “Every field in your database should return business value. Even if you get useless data for free, it still costs you in the end because it takes up storage, maintenance time and analysis,” she explains.

One additional criterion: If your company is in heavy acquisition mode, be sure to update your prospecting file with enhancement data on a quarterly basis, too.

11. How often should you perform a data integrity audit?
Answer: all of the above

Putting an optimal timeframe on a data integrity audit is the kind of question that begs the answer, “it depends,” but some benchmarks do exist.

The basic idea behind conducting a data integrity audit is to review each database field to determine if it contains the information you expect it to, e.g., the first name field does not contain a title like ‘Mr.’ At the minimum, this kind of audit can be performed every two years, says Mysliwiec. But, the better practice is to conduct it monthly or quarterly, she stresses, especially if myriad people and/or departments enter information into the database regularly.

Two other kinds of integrity audits are limited scope and referential. A limited scope audit only takes place when a structural change is made to a data table; you’ll want to review the data table and any related data processes to see how they might be affected, Mysliwiec explains. A referential integrity audit is conducted when you’re trying to link two types of data assets, such as purchase history to customer records, and occurs on an as-needed basis.

12. How often should you perform a full data asset audit?
Answer: c.) every five years

The most appropriate time to conduct a full data asset audit is when you rebuild or substantially change your database, which generally occurs every four or five years for most mid-size businesses, says Mysliwiec. Usually, the catalyst for the change is the migration to new software or other system tools.

While you might not perform a full audit of your database on a frequent basis, you should strive to review the meta data, or documentation, pertaining to your marketing data activities every other year—that is, says Mysliwiec, if you even create and store this documentation. The goal behind this extra maintenance step is to ensure that every person who makes decisions about the information going into the applications coming out of the database understand the origins and maintenance of the data on file.

For example, Mysliwiec points out, you should know what business rules were created, who created them, why they were created and how these rules affect the data. If at some point MM/DD/YY changed to YY/MM/DD, when did it happen, why and what does it mean for the company going forward.

• MORE ON
• DATABASE & CRM