Estimated reading time: 5 minutes

New DX benchmarking unveils the three ingredients that improve the bottom line for engineering, as the impact of AI coding assistants begins to be felt.

Engineering has long been a black box. But new research from developer experience specialists DX reveals how certain investments – especially in AI tools and vendor services – correlate with higher revenue per engineer (RpE), offering a new window into how that department contributes to the bottom line.

RpE, a measure of a company’s revenue over the number of engineers, is a secondary metric of DX Core 4, a new productivity framework that combines elements of the popular DORA and SPACE frameworks, with business-related metrics. 

DX analyzed revenue per engineer alongside research and development as a percent of revenue across 300 software-as-a-service companies. They found that the median revenue per engineer is $892,000. The top-quartile benchmark is $1.5m. The interesting part is what drives that gap.

Data suggests the top performers are investing in modern technologies like AI assistants and often resemble startups past the growth stage, without the corporate bureaucracy of large enterprises.

R&D and vendor tools drive efficiency

The DX study found that investing more in research and development (R&D), which for the sake of the study includes people costs, contractor costs, and non-people costs (vendor tools and services), directly correlates to higher revenue per engineer. But not all investments have equal returns.

Companies that spend a higher percentage of their total budget on vendor tools and services (non-people costs) have a higher revenue per engineer. The findings indicate that using third-party tools is more efficient than adding engineering staff or contractors to reinvent the wheel for common functions. 

That said, the report doesn’t assess the total ROI of these tools – only their correlation with higher revenue per engineer. A deeper analysis of the actual dollar cost of tools and impact would be needed to determine whether this approach is truly more efficient overall.

“Many companies still build everything in-house, but this data suggests that might not be the best idea,” says Brook Perry, head of marketing at DX. Building in-house might seem cheaper at first, but it can lead to higher long-term maintenance costs compared to buying a well-supported third-party tool.

AI assistants boost developer output

One such investment that is likely moving the needle is in AI tooling. “In theory, AI and automation should drive up RpE by accelerating trivial tasks for developers, so developers can reallocate time to tasks with more business impact,” says the report.

For instance, a large fintech company saw GitHub Copilot users deliver 24% more pull requests than non-Copilot users. Multiplied across a company with 1,000 or more engineers, those gains become significant.

Other studies portray AI’s impact in terms of time savings. A Google study found that AI significantly reduced the time developers spent on complex tasks, estimating a 21% decrease in task completion time. A survey of users at ANZ Bank found it led to completing tasks 42% faster.

DX data has found the most time-saving use cases for AI code assistants to be stack trace analysis for debugging, refactoring existing code, writing logic inside loops, test case generation, and learning new development techniques. While not directly correlated, these time savings may have a positive effect on RpE.

Young and lean = better returns

Unsurprising to most, the largest organizations show a low revenue per engineer. Companies with over 1,500 engineers are 18% less efficient than companies with 251-500 engineers.

Efficiency grows until org size passes 500 engineers, at which point it declines and then levels off, indicating the point at which inefficient corporate structures begin to hinder innovation.

Companies founded after 2010 are 27% more efficient than their predecessors. They’re typically built on modern tech stacks and processes, giving them an edge in productivity.

High-growth ≠ highly efficient

Moving too fast can be detrimental, however. Companies with a year-over-year (YoY) growth rate of 71% or more have the worst median RpE, at $462,000. The sweet spot is a steady 11-20% YoY growth, at the highest, $1.29m RpE.

Another leading indicator is size. RpE at companies with fewer than 50 engineers is highly variable, demonstrating the volatility of early-stage startups. When revenue passes $500m, efficiency tends to stabilize.

“Many people expect fast-growing companies to sacrifice efficiency, and this was very apparent in the data,” says Perry. If you’re growing 100% year over year and hiring like crazy, you’re likely not operating in the most efficient way, she suggests.

For now, it seems stable, mid-size companies have the edge – past their initial high-growth rates, with a base-level of efficiency, and using modern tools without as much red tape.

Will AI change future efficiency?

As the quality of AI-generated code improves, the status quo for engineering output in high-growth companies could change drastically. 

While AI can spin up proof-of-concept in seconds, it can also introduce technical debt and debugging toil.

According to Perry, we’ll need more data to track how or if revenue per engineer differs for fast-paced AI-based companies over time. It may flip the model entirely.

Thankfully for developers, not many leaders are using AI investments to justify slimming their workforce, says Perry. The focus is on how to recapture and gain more capacity per engineer instead of using high RpE to justify layoffs.

RpE as a north star metric

While RpE is a lagging metric, certain actions can nudge it forward. For instance, there’s more evidence that investing in AI coding assistants can reduce the waste and friction developers face, impacting productivity.

DX also recommends regularly surveying developers, since doing so can expose pain points that impact key metrics, as well as allocating time to reduce technical debt, and accelerating feedback loops in the build-test cycle.

The general push towards operating more efficiently has many interested in evaluating engineering output. RpE is one of the many metrics that provides a window into organizational efficiency when paired with other metrics.

“Increasing developer productivity is about gaining leverage and driving higher ROI per developer,” says Abi Noda, cofounder and CEO at DX. “So what better way to measure this than by revenue?” It’s a great tool to frame the value of engineering investments, he adds.

That said, it’s hard to say what exact activity directly improves revenue per engineer. It’s more like a north star metric that a combination of many factors move over time, says Perry. As such, while checking in on RpE once or twice a year is helpful, she cautions against setting exact targets around individual projects.

Get articles like this in your inbox

Choose your LeadDev newsletters to subscribe to.