How big should Developer Productivity teams be?
Benchmarking data on the percentage of engineering headcount that companies allocate to centralized developer productivity teams.
This is the first issue of a new series bringing you insights and trends on developer productivity. This article provides benchmarks on how much companies spend on centralized developer productivity and enablement efforts.
Improving developer productivity requires a two-pronged approach. Cross-cutting issues such as development infrastructure are well-tackled by centralized teams. Meanwhile, local team issues such as code quality and collaboration are best solved via local team members and leaders.
At DX, we frequently get this question: what percentage of headcount should be allocated toward centralized productivity teams? It's possible to learn about one-off examples, such as Etsy's 20% of headcount investment. But, to truly understand how a company's investment compares to industry leaders, more data is needed.
To tackle this, our team spent the past several months analyzing data from DX customers and non-customers who were willing to share information confidentially. For each company, we gathered the following data points:
Total engineering headcount
Headcount allocated toward centralized productivity teams
Breakdown of the what these different teams are focused on
We found that, on average, the companies we studied with under 1,000 engineers allocate 18.9% of their headcount toward centralized productivity teams, with a range of between 8% and 37%.
The full distribution of data points is shown in the chart below, and breakdowns by other segments such as funding stage are available here.
While initially analyzing the data, we noticed that the average allocation decreased to 17.8% when we included companies with more than 1,000 engineers. We were curious, therefore, to better understand how allocation differs based on organization size.
As shown in the chart below, we found that larger organizations tended to allocate a smaller percentage of headcount, although these smaller percentages still represented a higher absolute headcount than smaller companies.
We have a couple of speculations as to the cause of this.
First, there are diminishing returns in increasing centralized team headcount thanks to leverage. For example: it doesn’t require a team ten times larger to provide CI/CD tooling or curated best practices to 500 developers versus 5,000 developers.
Second, as organizations grow, centralized efforts may be less effective than localized efforts. Atlassian, for example, directs all developers across the organization to spend 10% of their time fixing productivity impediments. Time spent on productivity efforts by local product and platform teams were not captured in our data set.
In our study, we encountered the challenge of defining exactly what a “centralized productivity team” is. Many companies structure their productivity teams under similar names - such as "Infrastructure", "Platform", or "Developer Productivity". When we dug deeper into how these groups were organized, however, we found significant variance in the naming and scope of the sub-teams.
By iteratively coding our data, we were able to identify eight types of teams which we included in our definition for this study:
The data we’ve shared can help organizations benchmark and compare their investments. But the data also shows that no two companies are exactly alike. Thus, the question remains: what’s the best way to determine the optimal allocation for your company? To provide guidance on this, we asked several leaders to offer their perspectives.
Will Larson, CTO at Carta and author of the upcoming book The Engineering Executive’s Primer:
“Executives certainly make staffing decisions about Platform and Infrastructure teams, but I think they do something even more important: define the company’s engineering strategy. Many companies take a permissive stance towards adopting a variety of technologies, which creates a diffuse approach to internal teams who are supporting too broad a technology landscape. At both Carta and Calm, my most valuable contribution to developer productivity has been constraining allowed technologies, which allows us to concentrate our investment–however large it happens to be–as densely as possible.”
Mike Fisher, former CTO of Etsy and author of the newsletter
:“The way that I suggest approaching this challenge of headcount allocation for internal-facing teams is by focusing on specific goals and utilizing a feedback loop to ensure you are achieving the expected benefits. Examples of feedback loops include qualitative surveys of engineers and return on investment (ROI) metrics. For either one of these you should establish a baseline, implement a small change in resource allocation, and then monitor the results to see if the changes had the desired effect. In this way you are ensuring the investment you are making is having a positive result.”
Rajeev Rajan, CTO of Atlassian:
"Overall, I believe 10-20% is the right amount of total investment in developer productivity. Beyond empowered teams, this can also be local projects that tie back to the central developer productivity team. At Atlassian, we wanted to empower developers across our engineering team to fix the local problems that impact their productivity. We started by simply asking our engineers: how satisfied are you in your day-to-day, and what are the top 5 things getting in your way? In addition to our central developer productivity team, we empower every engineering team to allocate at least 10% of their time toward developer productivity and put it directly on their team roadmaps. Importantly, we backed this up with a company-wide OKR to keep us accountable to this goal. This is an ongoing journey, but we’re seeing great progress already.”
Earlier this year, Gartner reported that 78% of surveyed organizations had a formal DevEx initiative either established or planned. Given the rise of DevEx and Platform engineering, we predict that allocation toward productivity teams will increase in the coming months and years.
Want to see more data? You can download the full report here.
That’s it for this week! Thanks for reading.
-Abi