Implementing a Data Warehouse: Results of Dr. Payton’s Action Research with Solectron

In a 2001 project with Solectron, Dr. Fay Payton and a handful of students tracked data warehouse implementation for 8 months. During the last months (March-May 2002), the authors conducted an interview with members of the Solectron management team to determine how our research had caused “change” to result in the organization – as we sought to resolve the “problems” noted from our initial interactions and field experiments. To this end, we offer lessons learned to the field as action researchers and to practice. We have structured these lessons learned as “Propositions” that we believe future researchers need to study in more detail, and which align with our re-specified model. Results of this study have been submitted in a manuscript to the MIS Quarterly Journal.

Proposition 1: a) Pre-implementation metrics and b) post-implementation metrics that specifically identify the criteria for success that must be used as a means of tracking the data warehouse implementation project.
Performance measures and metrics are needed prior to data warehouse vendor selection. These metrics should be tightly linked to users pre-implementation requirements associated with their business processes. The performance metrics our team provided helped Solectron with its monitoring of data warehousing vendor. Solectron, however, indicated that having these methods earlier in its vendor selection process – prior to deployment – would have been valuable, in general. In particular, the organization would have managed contract negotiations and assessed vendor selection using more quantitative criteria. Thus, vendors would have to demonstrate performance earlier in the implementation process. As one Solectron manager rationalizes: “we hope to use these metrics to communicate with the [warehousing] vendor on how they can improve the services provided to us and we can in turn pass improvements in productivity and efficiency to our business partners and customers.” As shown in the model, metrics should be limited to a few key performance indicators, as the project progresses, to enable the organization to track and monitor its technology investment and perceived benefits.

On-going post-implementation metrics are also required to ensure success. As of the second quarter of 2002, the Global Data Warehouse Manager at Solectron continues to advocate the performance measures and metrics tools to evaluate performance. While graphics plots assist the practitioners to visualize key problem areas and communicate concerns to upper management of the vendor, challenges remain with the data warehouse’s success measures. In particular, problems point to the data warehouse vendor and its inability to improve performance associated with runtime in Solectron’s global operations centers and management of the extract-transfer-load (ETL) process. Further, the lack of additional resources continued to impact the implementation process as well as the warehouse’s performance.

Proposition 2: Defining organizational barriers and team skill requirements prior to implementation is more likely to lead to data warehouse implementation success.
Outsourcing critical, strategic application can be met with amplified implementation challenges. This is particularly the case when a dedicated internal IT staff is nonexistent. We also offer that the quality of the outsourcing partnership has largely been affected – as the degree of trust and conflict have been challenged (Lee and Kim, 1999). As shown in this action research, outsourcing to a data warehouse vendor without some degree of internal technical skills as a building block is a mistake. Admittedly, our action research team learned a great deal in the process of this project.

To increase the IT skills of the development team, Solectron has hired two data warehouse architects, one of which is responsible for performance tracking utilizing the measures and graphical tools described earlier. Further, upper management has allocated additional monetary resources to support the data warehouse implementation which was described as a “blessing in these economic times and a signal that the project is now mission critical for the entire organization”. More organizational communications endorsing the data warehouse are now occurring as a directive from upper management; these efforts are targeting senior management that the project has global impact for Solectron and its supplier relations. In particular, sourcing and order fulfillment capabilities are vital to the strategic application of the data warehouse. The sourcing function includes supply base management, controlling total cost, creating and exchanging long-term value, and creation of value partnerships with suppliers. The order fulfillment function focuses on plant management, OEM relationships, distribution, configuration of products to customer orders, and tying in to OEM’s customer order systems to identify configurations and post-manufacturing support. Optimization of these capabilities is anticipated to result in enhanced systems and data quality – thereby yielding real net benefits associated with the data warehouse implementation.

Proposition 3: A phased-in approach which addresses data integrity and system quality issues as they arise is preferable to a direct cutover approach that assumes the problems can be handled in a single batch mode.
Our action research case suggests that a phased-in approach has a number of advantages to a direct cutover approach. In our scenario, Solectron was deploying a global data warehouse and an ERP application, along with a myriad of internally developed source systems at multiple international locations. This combination of technology platforms complicated Solectron’s strategic vision and the sustainability of its core competencies. The politics of resource allocation,monetary, time and human, among these projects tended to impact organizational, project and system implementation success and, ultimately, systems and data quality. One interviewee spoke to the firm’s issue with the ERP application by stating: “…the robustness of EDI remains and it (is) a low cost option. Implementations of ERPs and data warehouses are costly and have tons of implicit consequences to the organization….such as process reengineering and cost overruns…. Understanding how the ERP views data versus how Solectron views data is a challenge. Often, we have had to rethink our processes to (fit) the ERP rather than pay for additional customization from (ERP vendor)”.

One of the most important but overlooked elements of data warehouse success is data quality and integration. The task of data integration requires that organizations begin data requirements and definitions processes early in the process. Once the data warehouse has been implemented without well-defined data elements, organizational resources (e.g., time, human and financial resources) can become depleted – thereby distracting from the strategic vision and core competencies of the firm. As one Solectron manager explained, “We had to make sure what the data elements were first, and to some degree, we are still looking at data elements. The warehouse hosts a large number of tables, and this was time consuming.” Planned goals and deliverables with pre-implementation metrics are fundamental as suggested by our model.

In the phased-in approach, the data requirements of each department should be carefully identified. Use of the information requirements determination (with a user-centered approach) methodologies are highly recommended. The “freshness” of the data required is an important input for designing the data warehouse at this point and stands to influence overall data quality.

Ideally, the historical data stored on the older system should used to populate the new data warehouse. Throughout this process, the data must be cleaned, validated, and reformatted to support the data warehouse structure. That is, the structure of the data warehouse must be aligned with users’ requirements. Data warehouses can be structured in sundry ways. Some methods of structuring the data include:

• Subject-oriented — organized around customers, parts, suppliers, etc.
• Integrated — consistent naming and formatting across databases
• Time variant — providing an historical, “snapshot” view

• Enterprise-wide — recognizes interdependent, “process” relationships across the organization