Applying Theory of Constraints to Improve Microsoft XIT Maintenance Team Performance

Abstract – The case illustrates how changing conventional thinking with the Theory of Constraints (TOC) can empower managers, strengthen belief, and encourage correct actions. By applying TOC’s five‑step drum‑buffer‑rope (DBR) solution, the XIT maintenance engineering team increased productivity by 155% and cut lead times from five months to roughly two weeks.

Overview – The XIT team, part of Microsoft’s eight IT groups, maintained over 80 internal applications. In early 2005 the team was the worst‑performing unit, with a backlog five times its capacity and a five‑month average lead time for change requests.

Background – In September 2004 Dragos Dumitriu became the program manager for the team. The team consisted of three developers and three testers after offshore transfer to Hyderabad, handling 85 change requests per quarter but completing only 17. Estimation (ROM) consumed ~40% of capacity, and the SLA forced ROM completion within 48 hours, further straining resources.

Initial Situation (July 2004) – Change requests arrived daily, each requiring a ROM that took a full person‑day from the already overloaded staff. Prioritization meetings added additional wasted effort, and cost accounting treated each request as a separate budget line, obscuring true throughput.

Management Measures

Measure 1 – Buffer : Introduced a development buffer (≈7 days of work) and a rope limiting work‑in‑process to about 12 items, allowing the team to focus on a single task at a time and dramatically reducing lead time.

Measure 2 – Stop Estimation : Eliminated ROM estimation, freeing ~40% of capacity and instantly improving productivity.

Measure 3 – Resource Reallocation : Adjusted the developer‑to‑tester ratio to 2:1 (later 5 dev : 3 test), matching the optimal TSP/PSP staffing pattern and providing excess testing capacity.

Results (October 2005) – Throughput rose from 17 to 56 changes per quarter, average delivery time fell to 12‑14 working days, on‑time performance exceeded 90%, and cost per change dropped from $7,500 to $2,900. The backlog shrank from over 80 items to fewer than 10, and the team was no longer a bottleneck.

Conclusion – The TOC five‑step DBR approach proved highly effective for IT software maintenance without adding resources or altering engineering methods. It demonstrates that most constraints in software development stem from management and workflow, and that simple, disciplined process changes can yield >100% productivity gains without resorting to more complex TOC tools.