practice

Technology Standards

Constraining technology variety to reduce operational cost — with a waiver path, or the standard is broken quietly rather than formally.

standardsgovernancepaved-roadexceptionsvariety

Definition

A technology standard says what the organisation uses by default for a given purpose: this language, this database, this messaging system, this observability stack.

What standards actually buy

  • Reduced operational surface. Every technology needs expertise, patching, monitoring, upgrade paths and someone available at 3am. The number of things an organisation can be genuinely good at is small.
  • Mobility. Engineers can move between teams without learning a new stack.
  • Negotiating position with vendors, through consolidated spend.
  • Faster decisions. Most choices stop being debates.

The cost is local optimality: some team is doing something more expensively than they need to because the standard says so. That is usually a good trade, and it is a trade rather than a free win.

What makes standards work

A paved road, not a rulebook. The standard should be the easiest option — a template, a module, a pipeline that works out of the box. Compliance then happens because it is convenient, which is the only form that survives deadline pressure.

A fast, credible exception process. Standards without waivers are broken quietly rather than formally, and the exceptions become invisible — which is strictly worse than granting them, because you lose the ability to see and support what is actually running.

An exception should require a stated reason, an owner, and a review date, and should be granted in days rather than in a fortnightly committee.

A defined lifecycle. Adopt, trial, contain, retire — reviewed on a cadence, so the standard set evolves rather than freezing the estate on decisions made a decade ago.

What to standardise and what to leave alone

Standardise where variety costs operational effort and provides no benefit: observability, deployment, secrets management, identity, base images, network patterns.

Leave open where the workload genuinely differs and the team bears the consequences: a specialised data store for a specialised problem, a language suited to a particular domain.

Over-standardising the second category produces teams fighting the standard, which is where waivers and shadow infrastructure come from.

Failure scenarios

  • Standards with no waiver path, producing invisible exceptions.
  • Standards without a paved road, so compliance is extra work.
  • A standard set that never changes, freezing the estate.
  • Standards covering everything, so the exception rate is high and the standard is discredited.
  • Enforcement without support — mandated tooling with no team behind it.

Interview question

"How do you set a technology standard that teams follow without resenting?"