Skip to content

Insights · Practice notes

How governed data systems are actually built.

Long-form notes on data engineering, decision systems and governance from the team that builds and operates SquareCampus. Frameworks and checklists you can apply with or without Fairhelm; no vendor pitches, no invented case studies.

Topics

  • Decision systems
  • Data engineering
  • Governance
  • Company
RSS feed
Decision systems

·10 min read

The KPI contract: definition, grain, owner, target, guardrail

Most dashboard disputes are not about the chart. They are about an undefined metric. Five fields that turn a KPI into a contract people can act on, and how to write them before anyone opens a charting tool.

Read the note

Decision systems

·10 min read

Exception ownership: the dashboard field nobody adds

A dashboard that shows what is wrong but not who owns fixing it is a report. Why exceptions need a named owner, an age and a next step, and how to design the surface so ownership survives staff changes.

Read the note

Data engineering

·8 min read

Batch or events? Choose data movement by recovery, not by trend

Streaming is not a maturity level and batch is not a compromise. A decision framework for picking how data moves, based on latency the decision actually needs, correctness, recovery and what the team can run at 2 a.m.

Read the note

Governance

·9 min read

Designing audit trails people can actually read

Most systems log everything and explain nothing. An audit trail is for a principal, a finance head or an auditor, not for a database. How to design one that answers who, what, when, why and under whose authority, in plain language.

Read the note

Company

·7 min read

Why Fairhelm is product-first and takes services by exception

A small technology company can be a product company or a services company, but pretending to be both usually means being neither. The reasoning behind Fairhelm's scope, what it says no to, and what a selective engineering engagement looks like.

Read the note

How to read these notes

Standards we hold ourselves to, written down before the case studies exist.

Fairhelm Systems is a young company. These notes describe the practice it applies to its own product and to the engineering work it takes on; they cite no customers because none are published. Worked examples are hypothetical and say so.

Vendor-neutral

Frameworks, checklists and decision tests that hold regardless of which tools an organisation runs.

Machine-readable

Every note has a Markdown alternate and is listed in fairhelmsystems.com/llms.txt, so agents read the same text a person does.

Checkable

Company facts derive from one canonical record. Nothing unverified is published, here or anywhere on the site.

Written to be acted on

Each note ends with what to do next: a checklist, a test, or a question to put to a vendor in writing.

Next move

Recognise your operating problem in one of these notes?

Bring the concrete version of it. Scope and success measure are agreed in writing before work begins.