Overview / Resources · B06
Reference shelf
Short guides for the people inside organisations who end up responsible for data without having asked to be.
General information. Nothing here is legal advice or a statement of compliance.
Writing validation rules people agree with
Automation fails when nobody agreed what correct data looks like. Start with the people who use the output. Ask what has gone wrong before: missing codes, negative quantities, dates in the wrong century. Turn each answer into a rule with a name, a test and an action (reject, flag or correct). Write them in a table everyone can read, and keep it next to the job.
Two questions about recovery
Before buying any backup arrangement, answer two questions as a business, not a technical team. First: if everything since a point in time vanished, how far back could you tolerate? A day? An hour? Second: how long could you work without this data before real harm? The answers set the backup frequency and the restore design. Everything else is detail.
Asking where data really lives
Ask any provider, including us:
- Which legal entity holds the data?
- In which country and region is it stored, including backups?
- Who can access it, and how is that recorded?
- How would we get all of it back, in what format, how quickly?
- What happens to copies when we leave?
When a spreadsheet becomes a system
Signs it is time to move a process out of a spreadsheet: several people edit it, other processes depend on it, it takes more than a few minutes to recalculate, or nobody is sure which version is current. None of these mean you need anything elaborate — often a small hosted table and a scheduled job is enough.
Glossary
- ETL
- Extract, transform, load: taking data from a source, reshaping it and putting it somewhere new.
- Validation rule
- A written test that data must pass, such as “every row has an account code”.
- Checksum
- A short fingerprint of a file, used to confirm it arrived unchanged.
- Reconciliation
- Counting records out of one system and into another to prove nothing was lost.
- Retention schedule
- How long each kind of data is kept, and what happens afterwards.
- RPO
- Recovery point objective: the most data, measured in time, you could afford to lose.
- RTO
- Recovery time objective: how quickly service must be restored.
- Runbook
- Step-by-step instructions for a routine or emergency task.
- Environment record
- Our term for the document describing where and how a hosted arrangement is set up.