Hitmetrix - User behavior analytics & recording

Customer ID: The Key to Your Data Mart

How do you identify a customer? It seems like an easy question. But to answer this, you first must define a customer. So first you need to consider what kind of business you are in.

If you are in a business-to-business environment, is a customer simply a person at a single business location? Or is it a little more complicated? Is a customer an account that you have set up for billing at a single location that multiple people can use at that company? Or further still, is a customer really the combination of multiple accounts at multiple locations that are legally bound into a contract for discounting purposes?

What if you don’t have any account number (prospective customer)? If a person with decision-making authority leaves a company, but someone else steps into the same position and has the same authority, have you lost a customer or prospective customer?

OK, the point has been made for BTB, but what about business-to-consumer? This is a little more straightforward, isn’t it?

Consider this about consumers. Everyone is identified by Social Security numbers, driver’s license numbers, phone numbers, addresses, passport numbers, e-mail addresses, credit card numbers, credit report numbers, various company account numbers and dates of birth, just to name a few identifiers. How many of these are distributable, unique numbers that do not change? Your answer: Social Security number? Is that your final answer? Sorry, there isn’t a million dollars at the end of this quiz and no easy solution.

Consider this: Besides being a sensitive number that many people are understandably less willing to freely distribute, Social Security numbers are not even necessarily unique, and everyone isn’t required to have one. So the pursuit for the perfect unique customer key goes on.

I recently worked on a major cross-channel marketing initiative for a large transportation company. In this campaign we worked with or maintained six different unique numbers associated with a customer or prospect (account number, site location number, contact number, unique outbound number, vendor-assigned number and response number).

Each was a legitimate number that needed to be included or maintained for specific purposes. Account numbers identified customers with billing accounts. Site numbers pulled together multiple account numbers at a given physical address. Contact numbers identified people associated with account numbers. Outbound numbers allowed for each mail piece (customers or prospects) to be given a unique number. Vendor numbers gave prospect records identity associated with a vendor collecting inbound responses before these responses were officially processed, validated and assigned response numbers. Finally, response number recognized each official processed response.

Not every campaign you work on will be this complicated. However, as you embrace the true concepts of customer relationship marketing, direct marketing and database marketing, and attempt to apply these notions across multiple points of customer contact, across functional systems and with the use of outside vendors and data, you inevitably will discover that multiple customer and prospect identification schemes are present and must be dealt with.

Here are a few lessons that will help you avoid common pitfalls that have wrought havoc on others:

• If you are building a data mart, make sure to allocate money to research and design for an effective identification scheme. Look inside and outside your organization for identification data items that may be used at some point. Carefully investigate the advantages and disadvantages of each data element.

• Beware of identifiers that do not apply to all customers or prospects. Ask the question: “Is the identification item always populated (not null)?”

• Ask the question: “Is the number really unique (e.g., is it possibly reused over time)?” Depending on the amount of historical data you store in your data mart, this may not show up as a problem now but may show up later as historical data limits are approached.

• Does the data item have enough unique number capacity to accommodate future growth in your customer and prospect base?

• Is there intelligence built into the number that could cause problems? Did you realize that even Social Security numbers have built-in intelligence? For example, the first three digits identify the geographic office that assigned the number. Do not rely on items with intelligence as sole keys.

• Design a numbering system that assigns new, sufficiently long, nonintelligent, unique identifiers to customers and prospects. Do not start the numbering at 1, but instead if the item is 10 digits, start at 1000000001. This way, if other systems read this as a number, the item will remain 10 digits.

• If you cannot generate new numbers, consider using more than one data item to uniquely identify a customer. For instance, the customer’s account number plus his registration date.

• If you cannot generate new prospect numbers, consider using a compiled prospect database instead of lists. One of the major advantages of compiled sources is that they maintain unique numbers for you, which allows you to deduplicate and track prospects effectively and more accurately.

• Before assigning any new unique numbers, obtain help from professionals who understand and can effectively execute a robust matching process to minimize duplicates and combine different levels of customer data.

• Devise a system for maintaining matching methods and householding over time. Plan to have an orderly way to retire customers and their records from your data mart without sacrificing the integrity of your identification schemes.

Assume that the world of customer and prospect data maintenance will be complex. If you are not faced with it now, at some time you will be asked to import and export data to other systems and to other vendors. Anticipating who all of these are upfront will be next to impossible, but study the ones you do know about.

The key is to place critical emphasis on the discovery and design associated with customer and prospect identification methods. Spend time carefully matching and deduplicating data before loading it into your data mart. Understand the different levels of customer data that will be needed to enable customer-oriented action from your data.

For example, even if you do not initially plan to load prospect data, consider how you will accomplish this and how this data eventually will interact with your existing customer data. Further, suppose you are not marketing to households today and, therefore, do not think that you need a householding approach. Still, ask yourself, “Is it possible that households will become a more important marketing level for us in the future?” Vigilantly searching and planning for potential data identification issues will pay huge future dividends.

Related Posts