Margy Ross

Ralph Kimball

Alan Alda is still widely known as "Hawkeye" Pierce from the hit television series "M*A*S*H" but he's also the long-time host of the PBS series "Scientific American Frontiers," in which he interviews research scientists. Discussing his 11-year hosting stint on National Public Radio recently, Alda described his approach for eliciting information from brilliant scientists. His tactics struck a chord with us because they're similar to the methods we advocate for collecting information from brilliant businesspeople at requirements interviews. Here are a few of Alda's practical techniques with our embellishments for DW/BI professionals.

Be curious, but not too smart. Skilled interviewers must be curious. Alda has a natural interest in science, but he warns of the "too-smart syndrome" where interviewers think they're nearly as well versed in the subject as interviewees:

"I found I wasn't asking good enough questions because I assumed I knew something. I would box them into a corner with a badly formed question, and they didn't know how to get out of it. Now, I let them take me through it step by step, and I listen. Then I say, 'Well, if that's true, then how could this be true?' or 'Tell me more about that.'"

Perhaps you've observed too-smart interviewers. Their questions tend to be long-winded, often eliciting blank stares or responses like "What was the question again?" Interviewers who try to impress others are missing the point. Ask simple, straightforward questions and you'll have a better chance of understanding complex concepts.

Know-it-all observers are also a potential problem. The observer's behind-the-scenes role during requirements interviews should be self-explanatory. Observers who would rather be in the limelight sometimes jump right in and start answering the questions on behalf of the intended interviewees. Once, while interviewing end users at a large manufacturer, a team had to have a "time out" after the first interview so they could suggest to one of the IT "observers" that he let the end users answer the questions.

The goal of a requirements interview is to ask questions in order to discover unknown frontiers. Think of it as a one-hour immersion to better understand what businesspeople do and why. How do they make decisions today, and how do they want to be making decisions in the future? First ask interviewees about their roles and responsibilities to get them engaged and focused on their spheres. From there, cover the following areas:

• What are the key business objectives?
• For each objective, ask about measures for success to learn more about key metrics and business dimensions.
• What roles do data and analysis play in achieving goals? Alternatively, how would better access and analysis benefit them?
• What are the current analysis challenges?

A good question to get the interview started is "How can people tell when you're doing a great job?" Marketing and salespeople especially like to answer this question.

Take a look at the sample questionnaires in The Data Warehouse Lifecycle Toolkit (see "Required Reading" at the end of this article ) to get a feeling for the process.

Be conversational. Alda believes "Scientific American Frontiers" is popular because of his conversational approach. Alda sets a tone that makes the interview more enjoyable for his subject:

"The conversational element ... makes it good. I'm trying to find out what she's doing, what it really means and how she really does it ... and she's trying to make me understand. She's not just giving a lecture. So it's a very personal interaction."

This is a profound subject, well known to office anthropologists. Thirty years ago at the Xerox Palo Alto Research Center, I (Ralph Kimball) was very influenced by the well-known researcher David Holtzmann, who said that the procedures in a typical office manual are worthless as indicators about how people acquire information and make decisions. What was needed, in Holtzmann's opinion, were the "shadow functions" that described what really took place. Perhaps the crucial source of information to make a decision came from informal conversations around the water cooler. The interviewer needs to get past the procedures manual and find these shadow functions.

For the DW/BI practitioner, being conversational means putting yourself into a business frame of mind. Learn the language of the business. Don't intimidate end users by asking "What do you want in a data warehouse?" End users aren't systems designers. Acronyms and IT vernacular don't belong in a business requirements interview. Business as a second language is a challenge for some of us: Not everyone is cut out to be a lead interviewer.

Tape recorders change meeting dynamics, so don't use them. Users are often uncomfortable being taped and they might want segments of the meeting to be "off the record," which makes for awkward transitions. If you rely on recorders, your written notes won't capture the interview's full content, so inevitably you'll have to listen to the recordings and take supplemental notes. This process is like watching reruns on television: It isn't very interesting, but consumes large chunks of time. It's better for the lead interviewer and an expert note-taker to be actively engaged in the session.

You're not setting the stage for conversation if you hand a list of data elements to interviewees and ask which ones are important. Ask open-ended questions to draw them out. Several techniques can help you establish a more conversational tone:

• Learn a bit about the business beforehand by reviewing the Web site or annual report to understand company-specific vocabulary and hot-button issues.
• Meet with interviewees on their own turf. Go to their offices or conference rooms, rather than IT meeting spaces.
• Prior to the interview, send out an announcement describing the high-level discussion topics and confirming the interview time and place. Don't attach a detailed questionnaire to this meeting notice. You can't achieve a conversational flow if you're reviewing questionnaire results-presuming anyone bothers to complete the survey.
• Interview questions prepared in advance are fallback devices, used only if uncomfortable lulls occur in conversations or to ensure key points are covered before ending sessions.
• Most good conversations tend to wander, so remember your session goals and steer conversations back on track if you stray too far from core issues. Stay at a relatively high level in the interview's early stages. Don't follow an early comment to a very low level of detail, only to run out of time and discover that you haven't discussed three other major areas of responsibility with important requirements for the DW/BI effort.