This year, the DORA report explored the topic of platform engineering. Overall, the impact of a platform is positive:
Individuals were 8% more productive and teams performed 10% better when using an internal developer platform. Beyond productivity, we also see gains when a platform is used in an organization’s overall performance, with an increase of 6%.
Nevertheless, it wasn’t all good:
Throughput and change stability saw decreases of 8% and 14%, respectively.
Here are some ideas that might help reverse those negative effects.
Keep it Lean
If we are to treat platforms as a product, one of the first questions to look into is:
“how much platform do we need?”
Do we really need to automate the set up of ALL THE THINGS? Are all the processes we have in place really necessary to deliver software in good quality? One of the hypothesis of the DORA report about the decrease of throughput is:
… when an internal developer platform is being used to build and deliver software, there is usually an increase in the number of “handoffs” between systems and implicitly teams.
Gradually reducing the platform services to a minimum can be a useful experiment to test the hypothesis.
Promote resilient services
Now, that you have your minimum set of services, imagine that every time you run them they take a long, long time to provide a result. Imagine that they break on a whim, 60% of the time. Would you consider skipping them, just to save yourself the emotional burden of trying to figure out what has happened?
A hypothesis for the reduction of change stability in the report is:
The platform isn’t effective at ensuring the quality of changes and/or deployments to production.
Unlike what is suggested in the report, that teams are not using a platform service because they don’t know about it, I would suggest that they skip it because of its quality. Making the services more robust could help test this assumption.
Make the services delightful
Things are looking better! You easily get services from your platform that are reliably there for you in a timely fashion. Now you are all set to get some feedback from them. You gather their bookmarks in one place. There are 34 important findings that need your attention spread across all of these services. About 20 of them are not even valid issues the rest of them you have no hints how to fix. Why don’t just ignore the lot of them?
Such a scenario can have a negative affect on both the throughput, stop to figure out what to fix as well as the stability, ignore the issues completely. Services that provide addressable feedback at the places that a user expects to find it, could help alleviate the cons found by the report.
My hypothesis for the cons of platform usage, is the we are not taking full advantage of platform factors that have a positive impact on productivity.
A user-centered approach that enables developer independence through self-service and workflows that can be completed autonomously.
Many times we focus on the “self-service” part by streamlining provisioning, but we don’t spend enough time about the “workflows” part.
An idea that I have seen work, is to add some entry criteria to the services we introduce in our platforms like:
- Set up: Is it really necessary to have it?
- Run: Is it reliable enough?
- Use: Is it easy to understand its feedback?
Maybe the DORA report 2025 will have some insights if that would work!
More Questions Than Answers 