Editor's note: Amy Starer is vice president, GENESYS Sampling Systems, a full-service sampling company based in Fort Washington, Pa. This article originally appeared in the company's newsletter.

The value and the necessity of business-to-business research has become increasingly apparent to the corporate world. This has caused explosive growth in the demand for this type of research. To capitalize on the new opportunities presented by this growth, many research organizations find themselves venturing into unfamiliar, uncharted and somewhat confusing waters.

It's not surprising, then, that our company's customer service staff is encountering more and more researchers who harbor misconceptions or a lack of basic knowledge about the sampling aspects of business-to-business research. These misconceptions and myths, like stories of sea monsters and mermaids, can seem humorous to those who have already sailed these seas, but they can be extremely dangerous to the less experienced. We would like to take a few of these myths and examine the reality hiding underneath.

Myth #1 - The supplier that provided the list also compiled the list. This may sound silly, but you would be amazed at the suppliers who claim to be generating sample from a proprietary data source when they are actually just remarketing someone else's database. For example, at this time there are primarily two companies that compile yellow page information for the entire United States: Database America and American Business Lists. Everyone who markets a yellow page list uses basic information that most likely started with one of these two companies. Many sampling products that are misrepresented in this way will have glitzy names -- be cautious.

Myth #2 - Standard Industrial Classification (SIC) codes are, in fact, standard. SIC codes classify companies based on their line of business (i.e., doctors, lawyers, restaurants, tool and die companies) and are a valuable tool for targeting purposes. The first four digits of the SIC code (called the basic four) are administered by the United States Government's Office of Management and Budget, and are standard. However, many marketing information companies further classify businesses and industries into smaller, more specific subgroups (i.e., medical specialty or restaurant type). This is accomplished by adding more digits onto the basic four-digit SIC code. For example, Dun & Bradstreet Information Services has developed an enhanced SIC coding system by adding four more digits to the basic four. These additional digits are proprietary and non-standard. Other compilers start with the same, basic four digits but the additional digits they add may represent something very different than those in the Dun & Bradstreet scheme.

Because this additional coding is not standard and varies from company to company, great caution should always be used when defining a sample based on what is being called an SIC code. It is very important that you ascertain exactly what type of SIC code is being used, especially when your project calls for a sample definition that will supposedly replicate an earlier study.

Myth #3 - Record-level data is always accurate. Much of the data contained in the sample record is accurate and up to date. However, some data is simply modeled using other available information. And sometimes, data that was once accurate (or perhaps even exact) has since changed. This, of course, presents difficulties when one is targeting a sample based on this data.

For example, most business databases have a field related to the number of employees at that location. Often, this number does not represent reality and is an approximation based (or modeled) on other information the database compiler has obtained about the company. Further, even if the number represents what was once an exact employee count, that number usually changes too quickly to be relied on completely. How many employees has your company gained or lost in the past year?

Myth #4 - Yellow page-based databases have limited targeting capabilities. This myth seems to be perpetuated because it is assumed that the lack of information available in the basic yellow page listings (i.e., company name, address, phone number and yellow page header) translates into a lack of useful targeting information in the resultant databases. However, many compilers take the basic yellow page information and merge other data with it. Often this data can be used successfully to target a business sample. It's important to keep in mind that the accuracy of this data can vary depending on the supplier and how the information was obtained (see myth #3). As always, the usefulness of this type of information for targeting purposes depends greatly on the particular specifications of the study.

Myth #5 - Targeting samples based on the headquarters/branch designation is as simple as it sounds. It's not a trivial task to gather the data necessary to accurately represent the relationships between the various business entities within all of the companies and corporations in American commerce. There is no easy way to determine, on a large scale, which business entity is a parent company or a corporate headquarters. How do you find out which locations are subsidiaries, which are branches and which are considered independent, though related, businesses? This information can only be gathered by contacting the firms directly or as a by-product of other data acquisition activities which are typically proprietary.

Researchers must be aware that, depending on the source, this information can be very accurate or very inaccurate. For example, Dun & Bradstreet gathers this information for millions of businesses as a routine part of its credit services operations. The information is highly reliable. However, list compilers who do not have access to this type of information, and who don't want to incur the expense of contacting all the firms on their list, may attempt to approximate this information through other, less effective means (such as name matches).

It is situations like this that most remind us that these lists are not created with market researchers (who are most concerned with representation) in mind. If list compilers come across headquarters/branch-type information, they may include it in their list, but they don't make an effort to gather this data for every record. This means that a record not marked as a headquarters location may, in fact, be one. In these cases, the lack of a designation is not the same as a negative designation, it only indicates an absence of data.

In conclusion, it cannot be emphasized enough that there isn't one ideal set of attributes that makes a list perfect for all studies. The very qualities that make a particular list or supplier the best source of sample for one study may spell disaster for another. The best defense against this happening is to ask questions. If your supplier is remarketing someone else's data, make sure they know enough about the product to be able to answer all your questions. Taking the time up front to effectively evaluate the various options available to you will lead to a more representative, cost effective survey and, ultimately, a happier client.