Getting Started and Finishing Well

A well-defined process for data warehouse projects brings business value and project success

Perhaps it seems obvious, but after I conquer these two obstacles, it's all downhill from there. In this column I would like to share with you the part that's not as obvious: a tested approach for starting, and then succeeding, with even the most complex data warehouse projects.

This simple approach, which I call the Management Decision-Making Process (MDMP), helps you decompose the difficult, messy tasks associated with the data warehouse and stay focused on the fundamental goals of the project. Using the MDMP helps ensure all the conditions are set for starting the project and gives you a consistent reference point for judging the project's ongoing success.

How It Works

The MDMP is the single most valuable piece of information to me when I develop proposals for, and implement, data warehouses. The MDMP, shown in Figure 1, is a circular, cause-and-effect model consisting of just four steps. I find it most effective to start with the "How's business?" step, progress completely around the model, and then visit the "How's business?" step one more time.

This approach makes sense from a business point of view. In my experience, the best business managers spend their time asking "How's business?", "Why is it like that?", and "What if I did something different to improve the business?" Then they make a decision and invest time and money changing the business. At that point, the question again becomes "How's business?"

Taking the perspective of a business manager, let's visit each of the MDMP's steps:

How's business? For an astute business manager, this should be the first and eternal question. The information applications that monitor the business should be able to answer it. These monitoring applications are typically the "green, amber, red" executive information systems - or reporting systems with some style of indicating when things are good, on target, or bad.

Why? Why is the business as it is? There's no point to noticing that business is bad if you can't find out why. The devil is in the details. You need detailed information you can search to find the cause of a warning signal. In addition to detailed data, you need levels of summarized data to allow for extensive analysis with a simple mechanism. That mechanism should let you drill through to the detailed transactions to see what specifically caused the problems.

What if? As soon as you know the "why," you should ask "what if?" What if the business were conducted differently? A valuable component of any data warehouse environment is its ability to provide some form of forecast for how an investment might turn out. I recommend providing a set of high, medium, and low investment scenarios to help management choose where to spend its scarce dollars.

Invest. With an idea of how to improve, the business manager invests some money in an initiative. The data warehouse lets the user monitor this investment very closely in order to finely tune the initiative early in the process, producing the best possible result. After the initiative is established and working, the eternal question comes back. That is:

How's business? (redux). How did the investment pay off? In the data warehouse, you can build up a corporate memory of initiatives you have tried and the results they produced. This becomes an incredibly valuable source of material for future business initiatives.

Getting Started

Data warehousing is all about helping management make decisions. It's that simple, and it's that complicated.

It's simple because "helping management make decisions" is an elegant expression of the goal of decision support and data warehousing.

Why is it complicated? Until now, the process the MDMP describes has not been well documented, well defined, or amenable to being broken down into its component parts, analyzed, systematized, and supported by computer systems. The simple model I present gives you the opportunity to define the process that you plan to support with a data warehouse.

Given this model, I recommend that when you start a data warehouse project, you and your business-manager clients approach the senior managers most invested in the company's future, present the MDMP as defined here, and then ask questions such as:

Do you have the information you need to:

• Measure, manage, and monitor the business on an ongoing basis? That is, can you answer the question, "How's business?"
• Make the big decisions?
• Find the causes behind the poor performance?
• Measure, manage, and monitor the decisions you have made in order to finetune them as they roll out?

How big a difference to the bottom line might it make if you had this information readily available?

If you determine you do not have enough information, that's when you know that you have the beginnings of a project! Time and again I have seen the light go on in a senior manager's eyes as I present the MDMP. You need to do your homework and develop your business case. You need to identify the decisions the managers face, the information they lack, how much of a difference improvements might make, and how your solution will support some of these decisions in the short term. By preparing well, you are building awareness of, and appreciation for, the value of a decisionsupport system.

Ensuring Success

There are no guarantees in life. Many authors have written on this topic in great detail. But looking at project success with the perspective of the MDMP, my additional guidance is this:

Rate This Article Rating: 1 (best) 2 3 4 5 (worst) Comments: Optional e-mail address:

• Find the people in the organization who really make decisions and who have staff working for them to put together the proposals for new business initiatives. These should be your target first users of the data warehouse.
• Think of the future. Don't develop anything that is closed to other areas of the business. Early in the project, talk with the business managers enough to discover a broad range of the major decisions they make. Ensure that the data warehouse design will support a wide range of decisions.
• Focus on collecting data that supports the decisionmaking process you aim to support. The fundamental business relationships and transactions against those relationships are the best data to collect. Most businesses have fundamental business processes that, when analyzed, provide guidance on future investments.
• Focus on developing fact tables that directly measure "How's business?" These tables should contain measures that people use to measure, manage, and monitor the business. Include such measures as revenue, profit, costs, movement of money and goods, delivery of services, and interactions with customers.
• When managers develop whatif scenarios to support a set of decisions, this data should be stored back into the data warehouse. As people start to use the data warehouse to make decisions, capture the forecasts that were developed.
• As the initiative is implemented, its effects should be derived from, and stored in, the data warehouse and made public. Use the results to show whether the investment paid off or not. Either way, the knowledge of what happened is very valuable corporate information.
• Collect the detailed transaction information. You may keep it for months, or for years, but you should definitely keep it long enough that investigations into the business can use this data in a thoughtful, retrospective way.
• Use star schemas (with conformed dimensions) to store the most detailed information. Doing so will provide a ready source of data that analysts can surf to find out why business is as it is. It will also serve as a solid foundation for generating multiple levels of summaries for monitor applications. At the higher levels, you may want to use summary fact tables or the multidimensional cubebased products.

Use It Wisely

I have used the MDMP on many successful projects to garner project approval and to give the project team a clear idea of the business value it needed to yield. Like any methodology, the MDMP is a way to decompose a complex task, not a rigid formula. The strength of the MDMP is its business and end-user focus. A data warehouse that meets the business and end-user needs of an organization will be a true decision-support system.

Guest columnist Peter Nolan (peternolan9@eircom.net) has worked in IT for 20 years, the last 10 of those specifically in data warehousing. He currently resides in Ireland and focuses his data warehousing practice in Europe.