Skip to main content

Compare & Contrast: IT Services vs. Product Companies

Key Learnings for Enterprise Federation Design


Side-by-Side Comparison

DimensionIT Services (TCS, CTS, HCL, Accenture)Product Companies (Google, Amazon, Microsoft)
Who is the "customer" of the core?Internal delivery teams (but also their clients)Internal engineering teams
Mandate strengthSoft — client can overrideHard — company owns the outcome
Project lifespan1–3 years (engagement-bound)3–10+ years (product lifecycle)
Primary federation mechanismPractice / CoE / asset librariesPlatform team + tooling enforcement
Governance modelReview boards, advisory authorityAutomated gates, peer review culture
Technology diversityHigh (client-mandated stacks)Low-medium (internal standards enforceable)
Innovation loopSlow (client risk aversion)Fast (own the risk, own the reward)
Core funding modelCost center / overheadProduct / R&D investment
Success metricAsset reuse count, cost savingsDeveloper experience, delivery velocity
Biggest challengeMandate authorityScale and governance complexity

The Core Paradox

Both models face the same paradox from opposite directions:

IT Services: Too much autonomy (client mandates, project diversity) makes federation hard to enforce
Product Companies: Too much central control (platform mandates) creates bottlenecks and resentment

The answer for both is the same: federation by value, not by force.

Teams adopt the core because it makes them faster and safer, not because they have to.


7 Transferable Learnings

Learning 1: Governance Lives in the Toolchain, Not Documents

Both models succeed when compliance is automated into CI/CD pipelines, not enforced through manual review boards. Policy documents are read once and forgotten; a failed pipeline build is impossible to ignore.

IT services implication: Invest in shared CI/CD templates with embedded governance checks before anything else.
Product company implication: Treat every new standard as incomplete until it has an automated check.


Learning 2: Core Teams Must Be Product Teams

The single biggest structural mistake in both models is staffing the core team like an IT operations team instead of a product team. Core teams need:

  • A product manager (not just a technical architect)
  • A roadmap with priorities driven by consumer feedback
  • An adoption metric as the primary success measure

Learning 3: Inner Source is the Bridge

The healthiest federation models have a formal inner-source model — project teams can contribute to core components, with core team review. This:

  • Reduces the bottleneck on the core team
  • Builds project team ownership and trust in core components
  • Surfaces real-world needs into the core roadmap
  • Creates a talent pipeline into the core team

Both Google (monorepo + open contribution) and Accenture (Technology Strategy Boards) use versions of this.


Learning 4: Federation Must Have a Thin Core and a Rich Ecosystem

The worst federation models try to own everything centrally. The best models own a thin, stable foundation and create a rich ecosystem of domain-specific patterns built on top.

THIN CORE (centrally owned, mandatory)
→ DOMAIN LAYER (owned by domain experts, governed by core rules)
→ PROJECT LAYER (fully autonomous within boundaries)

Amazon's API-first model is the clearest expression of this: mandate the interface contract, not the implementation.


Learning 5: Legacy Systems Need an On-Ramp, Not a Mandate

Both models fail when they ignore existing systems. The federation layer must design explicit legacy on-ramps:

  • Adapter/wrapper patterns for legacy-to-core integration
  • Strangler-fig migration paths
  • "Federate partially" options — a legacy system can comply with security and logging mandates without full re-platforming

Learning 6: Funding Is the Root Cause of Most Failures

The model that works: Core team funded as a strategic investment (R&D / product budget), with its own P&L thinking and adoption targets.

The model that fails: Core team funded as overhead, recharged back to project teams, with headcount tied to delivery rather than adoption.

Accenture's productization of internal platforms (selling them to clients) is an elegant solution — it creates commercial accountability without the overhead model's failure modes.


Learning 7: Rotate Engineers Between Core and Project Teams

The knowledge gap between core team and project teams is corrosive. Both models benefit from structured rotation:

  • Core engineers spend 1–2 quarters on delivery teams annually
  • High-performing project engineers rotate into core for a quarter
  • This builds empathy, shared language, and realistic platform design

Google's 20% time and Amazon's "builder" culture both create informal versions of this rotation.


The Spectrum: Where Should Your Federation Sit?

Not every organization needs the same model. Place yourself on this spectrum:

LOOSE FEDERATION TIGHT FEDERATION
(IT Services extreme) (Product Company extreme)
|---------------------------------------------------|
CoP + Asset Library → Platform Team → Monorepo + Toolchain Enforcement
(Soft governance) (Mixed mandate) (Hard governance)

Most enterprises should aim for Platform Team with Mixed Mandate — strong enough to enforce security and compliance, light enough to let project teams move fast.


The Ultimate Takeaway

The federation layer is not a technology problem. It is an organizational design problem that technology can either help or hinder.

The organizations that get this right share one trait: they treat the federation layer as a living product with customers, a roadmap, feedback loops, and success metrics — not as a policy document, a shared GitHub repo, or an architecture committee.

Build the governance first. Let the platform grow from the governance. Not the other way around.