The #1 Platform Engineering Problem You've Never Heard Of: Platform Decay Abby Bangser Principal Engineer Daniel Bryant Head of Product Marketing

tl;dr ● Everyone is building a platform; do it intentionally ● Platform decay is an issue regardless of organisational maturity ● We’ll cover the identification, explanation, and fixes for: ○ Lack of adoption ○ Poor ROI ○ Never-ending migrations ● Treat your platform as a product ○ Everything-as-a service ○ Process automation ○ Fleet management

Setting the scene

Gartner: What is platform engineering? "Platform engineering improves developer experience and productivity by providing self-service capabilities with automated infrastructure operations. It is trending because of its promise to optimise the developer experience and accelerate product teams' delivery customer value." https://www.gartner.com/en/articles/what-is-platform-engineering

to fall into ruin: the city's decaying neighborhoods to decline from a sound or prosperous condition: a decaying empire

How decay presents for internal platforms ● Growing maintenance leading to reduced ROI ● Slow adoption ● Internal competing implementations ● Never-ending migrations ● Narrow solutions leave tech behind

Let’s dive deeper into the signs of platform decay

Lack of adoption

https://www.linkedin.com/posts/markwoneill_kubecon-activity-7176562437108682753-RgCi

Field of (broken) dreams: If you build it, they may not come How to identify this? ● Low adoption/usage ● High cost to recruit/onboard new users ○ Unique feature requests ○ Narrow use cases showing interest ● Poor developer/devex survey data https://www.reddit.com/r/devops/comments/stuep4/weve_spent_months_building_this_platform_devs/

Field of (broken) dreams: If you build it, they may not come Lack of adoption (or declining adoption) is a common sign of platform decay Why? ● Developers’ needs not understood ○ Lack of product ownership skills and design thinking ○ Platform not offered “as-a-service” ● Ongoing support not a first class concern ○ Does not scale with application usage ○ Lacks day 2 management operations Steps you can take to stop it ● Apply product mindset including ○ Survey developers to identify biggest blockers and pain points (and fix them) ○ Iterate with small releases to meet needs ● Assign a dedicated “enabling” team to onboard teams onto the platform ● Market the value provided to existing users

Low platform ROI

https://www.linkedin.com/posts/markwoneill_kubecon-activity-7176562437108682753-RgCi

Platform money pit: The dangers of high cost and low impact How to identify this? ● Outcomes are unclear ○ Lack of platform roadmap ○ No deliverables for 6+ months ○ Unattached to company OKRs ● Developers using “shadow platforms” ● Multiple stakeholders seeking funding to create platforms https://www.reddit.com/r/devops/comments/13dbccy/am_i_a_bad_platform_engineer_or_am_i_just_a_bad/

Why? ● Project rather than a product implementations ● Scoping too low level, trying to create an “internal AWS/GCP/Azure” ● Big bang releases resulting in a big bang with expensive cleanup ● No clear ownership, budget, roadmap alignment with company and user needs Platform money pit: The dangers of high cost and low impact Platform decay can manifest as low ROI, often caused by high platform maintenance costs Steps you can take to stop it ● Apply product mindset including ○ Identify, scope, and deliver a platform-based fix for one problem/team ○ Focus on common solutions, allow edge cases to be off platform (for now) ● Hold a centralized platform group accountable for extensibility and standardization ● Leverage 3rd party tools where you can and build only differentiated components

Never-ending platform migration

https://twitter.com/morgantepell/status/1782517898017390739

Avoiding multiple migrations: Delivering value incrementally How to identify this? ● Multiple (large-scale) on-going platform migrations ● Platform team spends more than 50% of time on migration-related work ● Users are impacted by each migration ○ Unclear which generation of a solution to use ○ Documentation contains more than two methods of accomplishing any one platform-related task https://www.gartner.com/smarterwithgartner/6-ways-cloud-migration-costs-go-off-the-rails

Why? ● Project mindset funds only big bang solutions / migrations ● Lack of focus on defining a platform API and providing X-as-a-service ● Solutions are not built for day 2 or scaling realities which then require a rebuild ● New engineering leaders incentivised to “start afresh” when joining org/dept Avoiding multiple migrations: Delivering value incrementally Managing multiple platform migrations is often a trailing indicator of platform decay Steps you can take to stop it ● Apply a product mindset including ○ Continuous evolution over single release ○ Plan value driven migrations rather than be forced to complete EOL driven migrations ● Enable and hold accountable a centralized platform group ○ Measure and report on user impact ○ Ensure fleet management during design ○ Day 2 operations as a first class concern

Wrapping up

Conclusion ● Everyone is building a platform; do it intentionally ● Platform decay is an issue regardless of organisational maturity ● We have covered the identification, explanation, and fixes for: ○ Lack of adoption ○ Poor ROI ○ Never-ending migrations ● Treat your platform as a product ○ Everything-as-a service ○ Process automation ○ Fleet management

Thank you: Additional information Platform Engineering: Orchestrating Applications, Platforms, and Infrastructure (bit.ly/3UNxQei ) How to Take the First Step Towards Self Serve Internal Developer Platforms (bit.ly/4bLMzMD ) Platform Orchestrators: The Missing Middle of Internal Developer Platforms? (bit.ly/3URBoML)