Skip to main content
Catalog
T021
Technology

API Versioning Graveyard

MEDIUM(75%)
·
February 2026
·
4 sources
T021Technology
75% confidence

What people believe

API versioning maintains backward compatibility and keeps consumers happy.

What actually happens
+400%API versions maintained
+25%Engineering time on version maintenance
-75%Consumer migration rate
-30%Feature velocity
4 sources · 3 falsifiability criteria
Context

Teams version their APIs to maintain backward compatibility — v1, v2, v3. Each version promises to keep existing integrations working while new features ship on the latest version. The intent is good: don't break your consumers. But API versions rarely get deprecated. Consumers don't migrate because migration has cost and no user-facing benefit. The API team now maintains 3-5 versions simultaneously. Bug fixes must be backported. Security patches must be applied to all versions. New features must work across version boundaries. The engineering cost of maintaining the graveyard of old API versions grows linearly while the team size stays flat. Eventually, the oldest versions become unmaintainable — the engineers who built them have left, the code is undocumented, and the consumers who depend on them are the ones least likely to migrate.

Hypothesis

What people believe

API versioning maintains backward compatibility and keeps consumers happy.

Actual Chain
Old versions never get deprecated(3-5 versions maintained simultaneously)
Consumers have no incentive to migrate
Deprecation announcements ignored for years
Largest consumers are on oldest versions and hardest to move
Maintenance cost grows linearly with versions(Each version requires bug fixes, security patches, monitoring)
Bug fixes backported across all active versions
Security patches must cover all versions
Testing matrix multiplies with each version
Innovation velocity decreases(New features constrained by old version compatibility)
Data model changes blocked by v1 assumptions
Engineering time spent on compatibility, not features
Impact
MetricBeforeAfterDelta
API versions maintained13-5 simultaneously+400%
Engineering time on version maintenance0%20-30% of API team+25%
Consumer migration rateExpected 80%/yrActual 10-20%/yr-75%
Feature velocityBaseline-30% due to version constraints-30%
Navigation

Don't If

  • You're creating a new API version for every breaking change instead of batching them
  • You have no deprecation timeline or enforcement mechanism

If You Must

  • 1.Set hard deprecation dates with contractual enforcement
  • 2.Provide automated migration tools for consumers
  • 3.Use additive-only changes to avoid versioning when possible
  • 4.Charge premium support rates for deprecated API versions

Alternatives

  • Additive API evolutionAdd new fields/endpoints without breaking existing ones — no versions needed
  • GraphQLClients request exactly what they need — schema evolution without versioning
  • Consumer-driven contractsTest against actual consumer usage, not assumed compatibility
Falsifiability

This analysis is wrong if:

  • API consumers consistently migrate to new versions within the announced deprecation timeline
  • Maintaining multiple API versions costs less than 10% of API team engineering time
  • API versioning doesn't constrain feature development velocity compared to unversioned APIs
Sources
  1. 1.
    Stripe API Versioning Strategy

    Stripe maintains 100+ API versions through automated compatibility layers

  2. 2.
    Google API Design Guide: Versioning

    Google's approach to API versioning and the challenges of deprecation at scale

  3. 3.
    Thoughtworks: API Versioning Anti-Patterns

    Common mistakes in API versioning that lead to maintenance burden

  4. 4.
    Postman State of APIs Report

    Survey data showing API version deprecation timelines are rarely met

Related

This is a mirror — it shows what's already true.

Want to surface the hidden consequences of your engineering decisions?

Try Lagbase