Data Analysis Production Line : An Example from the Whitehead / MIT Center for Genomic Research

Finishing is the final phase of the gene sequencing process employed in the Human Genome Project. During the finishing process, analysts inspect and resolve ambiguous regions of DNA. While other phases of gene sequencing have evolved into high-volume, automated processes, finishing has remained manual, time-consuming, and costly. This thesis studies finishing as an example of a broader class of processes it terms data analysis production lines. Unlike traditional production lines, data analysis processes typically consist of complex, information-oriented tasks in which skilled analysts play a central role. Because task complexity and analyst skill level tend to vary, however, processes like finishing are often difficult to monitor, predict, and control. This thesis identifies a series of workflow policies and organizational changes designed to help managers regain control of data analysis production lines. Using insights gained from a six-month internship at the Whitehead/MIT Center for Genomic Research, it shows that process variability can be greatly reduced through disciplined workflow policies. Specific policy recommendations include: breaking complex tasks into simpler ones before assignment; limiting analysts' workloads; requiring analysts to process work in first-in-first-out order; and, matching task complexity to analyst skill level. A simple modeling framework serves as the basis for proving the efficacy of these policies. The thesis concludes with an analysis of the organizational dynamics at Whitehead. A new performance review and incentive system is proposed as a means of encouraging better teamwork and knowledge sharing between analysts. Thesis Supervisor: Stephen C. Graves Title: Professor of Management Thesis Supervisor: Roy E. Welsch Title: Professor of Management