Responsible systems

Clear ownership, traceable changes and proportionate access.

Stewardship

Give a dataset an accountable owner

A dataset can be technically available while nobody knows who can explain its meaning or approve a change. Ownership creates a route for questions and decisions. It should describe a responsibility that survives staff changes, not only a person's name in a spreadsheet.

Stewardship
Classify data by the handling it needs

A classification label is useful when it changes how data is stored, accessed or shared. A large vocabulary without practical consequences creates paperwork. Start with a small set of categories tied to actual handling decisions and understandable examples.

Stewardship
Grant the access needed for the task

Access should match a clear task and scope. A person who only needs to read a report usually does not need to modify its source tables. Smaller permissions reduce the consequence of mistakes and make reviews easier to understand.

Stewardship
Collect the fields the purpose actually needs

An analysis often begins with a convenient export containing more fields than the task requires. Reducing that input simplifies handling and limits exposure. The question is not only whether a field can be collected, but whether it changes the intended result.

Stewardship
Record where information came from

Provenance describes the origin and processing history of information. It can help a reader assess reliability and understand transformations. Recording a source alone does not prove that its contents are correct; it makes the claim inspectable.

Stewardship

Operations

Log the change that matters

An audit trail should explain who performed a significant action, what changed and whether it succeeded. It is not the same as verbose application debugging. Useful audit events support accountability while avoiding unnecessary capture of sensitive payloads.

Operations
Make a proposed change easy to inspect

A reviewer needs to understand the resulting behaviour, not reconstruct a long conversation. A reviewable change describes the current problem, the proposed outcome and how the result was checked. Smaller, coherent changes are easier to assess and reverse.

Operations
Prove that a backup can become a service

A successful backup job shows that a copy was produced. It does not establish that the service can be restored from that copy. Recovery depends on configuration, credentials, software versions and the order in which components become available.

Operations
Preserve facts while an incident is unfolding

During an incident, observations, hypotheses and actions easily become mixed together. A short chronological record keeps them distinct. This helps responders coordinate and prevents an early guess from becoming the accepted explanation without evidence.

Operations
Keep changes connected to their purpose

Configuration drift becomes difficult to explain when settings change without a record of purpose. A useful history connects each significant version to the problem it addressed and the evidence that justified it. It should also identify which version is active now.

Operations

Search the library