Within the grand scheme of software program improvement, platform engineering is a comparatively new self-discipline. As such, platform engineering groups are nonetheless determining greatest practices and messing up alongside the best way.
In a chat at PlatformCon 2025 final week, Camille Fournier, CTO of Open Athena and co-author (alongside Ian Nowland) of the ebook “Platform Engineering: A Information for Technical, Product, and Individuals Leaders,” explored frequent errors she sees groups making and presents recommendation on tips on how to keep away from them.
“We predict that platform engineering is the subsequent logical evolution that’s wanted by the expertise business to essentially deal with plenty of the underlying complexity that we’re seeing in the present day, particularly in massive expertise organizations,” she mentioned. “We predict it is a crucial matter, however we additionally assume it’s a really laborious factor to do. We’ve seen lots of people attempt to battle to construct out profitable platform groups, and so we wrote this ebook as an try to assist individuals who had been fighting platform engineering to do a greater job.”
RELATED CONTENT: Constructing a tradition that may drive platform engineering success
A typical mistake individuals make is just not placing the appropriate individuals on the crew, comparable to solely together with software program engineers or solely together with operations. Platform engineering groups want a mixture of individuals with completely different abilities, together with software program engineers, DevOps, SREs, infrastructure engineers, and programs engineers.
Software program engineering is a core a part of platform engineering, since you want to have the ability to write significant software program so as to handle complexity. “Past automation and past operations — each of that are extraordinarily necessary — you need to be prepared to construct new software program merchandise,” Fournier mentioned. “You need to be prepared to construct self-service interfaces and enhanced APIs and safety and high quality guardrails, however you want software program engineers on these groups when you’re going to essentially be capable of create the form of complexity discount that issues.”
Alternatively, in case your platform crew is just software program engineers, that introduces an entire different set of issues. Software program engineers could not need to take into consideration operations. They need to construct frameworks, they need to construct a library, they need to construct a blueprint, she defined.
“There isn’t any lasting worth when you do not need operational possession … If you wish to have a platform crew that isn’t going to get defunded, you higher be working some issues that individuals truly depend upon … You’ll construct higher software program when you run it and preserve it in manufacturing. However the large value of that is upkeep, it’s operations, it’s upgrades. You want individuals with these system abilities.”
Not having a product strategy is one other mistake platform groups make, as this results in constructing in options that customers aren’t truly utilizing. Platform groups must be working with their finish customers to know how they are going to use the platform.
“You’ve acquired to have that buyer empathy in your platform crew that really cares in regards to the individuals which are going to make use of this software program and will get their enter on what you’re constructing, so that you simply’re constructing one thing that really meets their wants and calls for, and never simply what you assume is true,” she mentioned.
There are two main failure factors generally seen when constructing the platform, Fournier identified. One is that the platform crew builds what they assume their customers want, and the other downside is listening an excessive amount of to customers and implementing each single function they want.
“If you find yourself on this function manufacturing unit, you find yourself constructing these kind of Rube Goldberg architectures that themselves create the identical issues that you simply acquired within the first place,” Fournier mentioned. “Upon getting a Rube Goldberg structure, it’s laborious to construct one thing that your prospects can extra simply plug into and use. It’s laborious to evolve. You develop into increasingly of a bottleneck.”
In keeping with Fournier, when you can mix software program engineering abilities, operational abilities, and a product focus, that’s an ideal baseline for constructing out a platform group.
One other main mistake is constructing a v2. What she means by that is that generally platform groups will discover themselves in a state of affairs the place they have already got a system, however they will’t actually incrementally change it, so that they go and construct a wholly new system.
Issues come up as a result of irrespective of the way you assume customers are utilizing your system, you possibly can’t actually know for certain. Odds are, there’s some crew or particular person counting on some a part of it, and transferring on to one thing else will lead to reliability points. Subsequently, constructing a V2 is a excessive threat operation.
One other method through which it’s a excessive threat operation depends upon the best way your crew is about up. She referred to Simon Wardley’s pioneers, settlers, and city planners idea. The pioneers are those doing actually progressive work, who’re snug with threat.
“They discover one thing which may work, after which if they’re profitable, they’re adopted by people who find themselves extra like settlers who’re snug with some ambiguity, they usually wish to form of take one thing that’s messy and clear it up and make it a bit bit extra steady and scalable, after which over time you get the true city planners who need to make this method actually environment friendly and are very snug on this kind of massive system that has numerous completely different trade-offs for effectivity and progress.”
A V2 of a undertaking is usually began by a pioneer, however platform groups are often not made up of pioneers; profitable platform groups sometimes encompass settlers and city planners.
Even when a platform crew managed to think about a brand new progressive factor, there’s the problem of migrations. Fournier mentioned there’s truly a giant alternative for platform engineering groups to determine methods to make migrations much less painful.
“If everyone on this room takes away one factor, assume very laborious about how one can make migrations a lot simpler in your prospects,” she mentioned.