We conduct productivity benchmarks to diagnose issues, spotlight high-performing teams, and promote best practices. By leveraging data-driven insights, our model provides objective, scalable assessments of software engineering output, fostering transparent, constructive discussions to address challenges and optimize performance.

As engineering teams grow, relying on intuition and experience to assess productivity becomes unsustainable. Traditional metrics like story points or commit counts are flawed and easily manipulated. Our research at Stanford suggests a more effective approach: analyzing code downstream to uncover team dynamics and bottlenecks. While manual code reviews can yield accurate insights, they’re slow and costly. That’s why we developed a groundbreaking tool that evaluates code with precision, speed, and cost-efficiency—providing data-driven support to enhance developer productivity at scale.

This case study compares five software teams from three agencies working on similar projects, uncovering significant performance gaps. While Net Promoter Score (NPS) highlights client expectations, it fails to measure the true value teams deliver for their cost. By analyzing over 2 billion lines of code, we aim to provide transparent, data-driven insights to optimize team performance and outsourcing decisions.

Top software engineering teams measure both output and outcomes to perform at their best. Measuring only one can lead to underperformance in the other. Without this balance, teams risk either building the wrong things quickly (Quadrant 2) or the right things too slowly (Quadrant 3), which can cause them to lag behind competitors.

This case study highlights how teams in Poland, Brazil, and Romania outperformed the global average by ~50%, as measured by a Stanford-developed algorithm that evaluates team productivity through the impact of code commits. Unlike traditional metrics, this approach provides a more accurate and objective measure of team output, encouraging collaboration over competition. Our research empowers organizations to make data-driven decisions about team structure, methods, and tools for maximum productivity.

According to Google, Gemini 1.5 Pro outperforms GPT-4 in large-context and multimodal tasks, particularly excelling at finding specific details in huge datasets. For single "needle-in-a-haystack" searches, the model achieves >99.7% recall. In more complex tasks requiring multiple details, it outperforms GPT-4 but misses ~40% of relevant data, limiting certain applications.

Boosting your software team’s performance isn’t about hiring more engineers—it’s about optimizing how your current team works. Research at Stanford shows that most teams could move 30–60% faster with the right approach. Adding more people often slows things down, thanks to increased communication complexity, integration challenges, and ramp-up times.

A remote team in retail tech operated at 79% of their in-office productivity, highlighting how remote work affects performance differently across teams. Factors like team size and structure significantly influence outcomes, underscoring the need for data-driven insights. Using a Stanford-developed algorithm, we objectively measured productivity by analyzing the functional impact of source code changes.

A year post-layoffs, how does Twitter’s web traffic and monthly active users (MAUs) stack up? While most platforms saw global traffic gains, Twitter and Facebook declined. In the U.S., MAUs dropped or stayed flat, with Twitter and Facebook showing the largest decreases.

This case study highlights the flaws of using DORA metrics as productivity measures by comparing two "Elite" teams, where one achieves 2.4x more output despite similar deployment frequency. DORA metrics, while useful for tracking processes, fail to capture true productivity and can be easily gamed. Our Stanford-developed algorithm provides a more accurate measure by analyzing the functional impact of code changes, helping teams make data-driven decisions to improve performance.

Generative AI tools boosted productivity by ~15% in low-performing teams but only ~6% in high-performing ones, as shown in a 9-month study of five comparable software teams. This disparity suggests GenAI may provide greater value to teams with more room for improvement. Using a Stanford-developed algorithm, we objectively measured productivity based on the functional impact of code changes, offering a data-driven alternative to traditional metrics like Lines of Code and DORA.

In a high-performing development team, the distinction between productivity and roles is crucial to understand. A recent analysis of a top 5% Lead Engineer reveals they delivered 3.25x the output of a median developer, contributing 61% of the team's total output.

Our study explores when having a remote developer team makes sense, revealing surprising insights: while top performers thrive in remote environments, low performers struggle to stay productive.

Is relocating your developer team to a lower-cost region a viable strategy? One company successfully reorganized their ~90-person team by replacing 42 Bay Area engineers (averaging over $200k each) with 41 developers in Poland ($80k each) and 5 in India ($35k each), cutting their annual engineering spend from $12.3M to $7.4M.

This consumer tech team experienced a 30% drop in productivity during the first 9 months of remote work but rebounded to surpass pre-COVID levels by 2021. Our Stanford research highlights how remote work impacts productivity differently across teams, emphasizing the need for data-driven insights. Using advanced algorithms, we objectively analyzed developer output based on the functionality of their source code.