Today, competitiveness = speed + accuracy. If an organization is going to move effectively to keep pace with market trends and stay competitive, it needs to quickly gather and analyze relevant facts from its value chain and use that information to make better decisions. Moreover, it needs to make this happen as cheaply as possible, especially if it is budget-constrained.

Easier said than done. To succeed in today's competitive environment, management needs real-time access to relevant information that must often be aggregated from across the business landscape. Often, the available data cannot answer management's questions such as, "What is the availability and use of a company's assets and the profitability of its activities?" Typically, getting specific answers requires making a formal request to the IT department. Then, the IT department has to search for data through disparate sources and consolidate the information into one view. The final challenge is in applying the proper analytics to the data to deliver meaningful insight. Thus, a request of this nature could take weeks or months to complete. By then, the market and business conditions have changed, and the information is irrelevant.

What can be done? The answer is easier for organizations with bigger budgets. They can implement performance management systems supported by a corporate data warehouse. How-ever, what about budget-constrained companies? What about companies that have independent business units and distributed computing environments or those that have multiple types of transaction systems (some of which may be manual), each with different data formats and platforms? What about organizations driven by mergers and acquisitions that must regularly deal with complex data integration issues?

As market pressures accelerate business change, these organizations face a dilemma. They are struggling to obtain meaningful information on a consistent basis, yet they cannot afford to make large investments or multiyear commitments to enterprise-wide business intelligence programs. These organizations need a radically different approach to data delivery.

There is hope. Budget- constrained and information-starved organizations can follow several key principles to optimize their data delivery programs. They can: 1) structure their information requirements in a manner that helps manage the project; 2) standardize the data and the granularity of the data extracted across all sources by subject area; and 3) distribute data extraction to the source systems. Organizations with large and complex data integration issues can benefit greatly because results can be delivered quickly, at low cost, before the business changes.

The Challenges

Here is a common scenario, one that will resonate with the experience of many organizations. Consider a global organization that has multiple manufacturing plants. Some plants have an enterprise resource planning system; however, each system is implemented differently. Data and data granularity differ across locations. Obtaining basic revenue and cost data by customer, product, market and location requires monthly, resource-intensive, manual consolidation of spreadsheets from each location. Corporate only has access to summary data, and it usually arrives several weeks late. Common order entry does not exist. Worst of all, it isn't anyone's job to manage the data from an enterprise-wide perspective. Is it the job of corporate IT? Maintaining and supporting each plant's enterprise resource planning system is not its core competency or responsibility. The business is focused on manufacturing and selling product. What about plant IT? Maintaining and supporting its own enterprise resource planning system is its job, not consolidating and aggregating data from the other plants for performance reporting.

Regardless, management needs a consolidated view of daily sales and orders across all plants to support their decision making. If this can happen, there will be a better understanding of sales price, customer segmentation, customer service costs and product mix. If the data can be consolidated and delivered via one "trusted" source, this will reduce manual reporting processes and data inconsistencies.

The goal sounds wonderful; but for many organizations, it's an impossible dream.

The Approach

What can be done? Many years of experience as a consultant led me to consider the three optimization principles introduced earlier.

Principle 1: Structure the information requirements to manage data delivery. Consider once again the hypothetical manufacturing organization. For this company to understand its business, it must deliver consolidated daily order and sales data from across all plants. Program management can prioritize the delivery of sales and order key performance indicators (KPIs) into one or more time-phased releases with the help of a matrix of KPIs by dimension, such as the one shown in the adjacent chart. I use this diagram to help identify and prioritize the KPIs and understand how to tackle the daunting task of data delivery.

As you can see in the chart, the KPIs are the outputs (e.g., gross sales, net sales, fixed costs, variable costs and margin figures). The dimensions are the characteristics by which the values of the outputs are constrained (e.g., monthly sales by product, customer and market). In this example, sales data should be addressed in the first release. Aggregated costs will follow in the second release and so on. Each release should be managed as a single project in which more specific KPIs or dimensions are reprioritized with future requirements and are scheduled for delivery in subsequent releases. An effective way of maximizing value, early and often, is to reprioritize requirements and reschedule releases as business needs or market conditions change.

Additional detail should be added to the matrix by identifying each attribute of each dimension in the matrix; the product dimension, for example, should describe all attributes in the product hierarchy such as cost of sales, expenses, etc. When the column and the row intersect, it represents a deliverable or the delivery of data for a specific purpose. By color-coding these cells, I can easily describe scope and requirements, track delivery progress, indicate availability (or lack thereof) of data, rate data quality, identify future data requirements and define security requirements. This multidimensional view also provides a vision for analytical reporting. Thus, the matrix is a powerful communication document.

The matrix also highlights a number of functional challenges, which may include the allocation of plant fixed costs, variable costs, variances and corporate expenses, and application of country-specific accounting rules. In addition, attributing the dimensions will raise questions about how I combine or roll up data from across the plants. It may not be possible to deliver all levels of the data at once. Even rolling up sales and order data by time can create challenges because each location may have different accounting periods. These are just a few of the challenges that may arise and, thus, indicate where scope will need to be managed.

Principle 2: Standardize the data and granularity of the data by subject area. The second optimization principle is to define the data elements at the granularity required to calculate the KPIs. Then, organize them logically by subject area such as "customer" or "sales." This method of organization is called a standard record layout – or SRL for short. A good example for the manufacturing scenario we're using here is the "product" SRL, which would include the hierarchy, characteristics and standard cost information for each product from each plant. The advantages of this approach are:

  • It is independent of data source implementation.
  • If plants merge, split operations, drop or increase product lines, the information requirements to manage the business often remain the same.
  • It supports merger and acquisition activity during and after the due-diligence phase.
  • It handles different frequency requirements (daily, monthly and quarterly, if necessary). Some plants may run their order entry and financials on spreadsheets; because of the low volume of sales, it may only be necessary (possible or practical) to obtain sales and order data once per month.
  • It accommodates different granularities of data. Not every attribute of the product hierarchy may be available at each plant. This difference in granularity is often a consequence of merger and acquisition activity.
  • If the organization divests a plant, the interface from that plant can simply be turned off.

Principle 3: Distribute data extraction to the source systems. The next step is to treat each plant as if it were a new acquisition. The objective of this principle is to get the raw data into the business users' hands so they can interpret and cleanse the data. Then the next step is to get the corrections to the local IT resources that know how to make changes to the data to improve its quality. When the data comes in initially, it may be in poor condition, including missing fields, overused fields (multiple meanings and interpretations), invalid values (billion-dollar orders, negative values, alpha characters in numeric fields) and so on.

The major challenges (and the sometimes exorbitant costs) in building a data warehouse are data extraction and cleansing. Costs can be shifted by having each plant perform data extraction in parallel using the SRL. The assumption is that plant IT resources know their systems better than any other area of the company and can quickly address data and application-specific anomalies.

By distributing data extraction activities across all plants, the project achieves its first economy of scale. A second economy of scale is realized by distributing quality improvement initiatives to the product line, sales and order process managers at each of the plants. These resources typically have the authority and budget to force the necessary business process changes (for example, improved data entry edits) and set policies for the correct application of business rules in each plant's systems.

Delivering the Data

Complex, distributed data management challenges require a different data delivery approach that produces results quickly. This approach uses small, iterative cycles to deliver results in a relatively short time to meet business demands for management information, while not sacrificing data quality or compromising the best principles of program management and design.

There are many benefits to this approach, including:

  • A better framework for centralizing program management.
  • Centralizing information quality management and factoring in distribution of data extraction and data quality improvement activities can have significant cost advantages, such as reducing inconsistent, error-prone and multitudinous forms of data cleansing across functional groups.
  • Reduced risk. Management can schedule resources and costs in much smaller amounts such that the marginal cost of delivering the next increment or release is minimal. Additionally, as data quality improves, so does decision quality, which reduces business risk.
  • Avoiding many of the problems inherent in traditional development, including long lead times for data extraction and cleansing, which reduces data availability.
  • Enabling users to quickly alter data requests to perform more sophisticated analysis.

Many organizations may need some guidance to implement the approach discussed here. I hope this article gives some hope to companies. Even without a big financial and time commitment, you can act now to leverage information assets to improve your business.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access