A spirited debate has ignited in Silicon Valley following a viral series of tweets by Stanford researcher Yegor Denisov-Blanche. His analysis of the work of over 50,000 engineers across hundreds of companies revealed that approximately 9.5% of programmers effectively contribute nothing while continuing to draw salaries. Denisov-Blanche has coined the term “phantom engineers” to describe such employees.
Denisov-Blanche’s team was granted access to the internal code repositories of various companies, and over a two-year period, they conducted an analysis using an algorithm designed to evaluate employee performance. The algorithm assessed the complexity of assigned tasks, the time spent writing code, its quality, and structure. The study concluded that the productivity of “phantom engineers” was less than 10% of the average level.
According to Denisov-Blanche, companies could save billions of dollars by identifying and dismissing such employees. He emphasized that retaining “phantom engineers” wastes resources, burdens teams, and hampers progress. However, the research findings have yet to be published in peer-reviewed journals and are currently based solely on graphs shared via social media.
The growing interest in identifying unproductive employees aligns with the rise of “overemployment”—a phenomenon where individuals secretly hold multiple jobs simultaneously. In response, companies are increasingly adopting workplace monitoring tools to detect inefficiencies. One notable example involves a system that logs user inactivity on a computer for more than 30 seconds and evaluates “productivity” based on desk behavior.
Denisov-Blanche noted that in the past, companies often accepted the inefficiency of 10–15% of engineers as an unavoidable cost of doing business. However, the situation has shifted. The algorithm developed by his team provides highly accurate assessments of code quality and writing time while eliminating the possibility of metric manipulation (e.g., superficial changes to repositories).
Critics argue that evaluating engineers solely on their coding output fails to account for their broader contributions, such as design or management roles. Denisov-Blanche countered that companies clarify job responsibilities to exclude employees whose work extends beyond writing code from the analysis. Nevertheless, the research has already attracted significant interest from organizations.
Amid a tightened labor market and widespread layoffs, engineers are finding it increasingly difficult to secure new positions with comparable conditions. This shift has heightened scrutiny of workforce efficiency and accelerated the adoption of advanced technologies to optimize workflows. Denisov-Blanche predicts that in the future, programmers’ productivity will be assessed similarly to salespeople—based on tangible results rather than indirect metrics.
