Features
Reporting
This page is about how a figure is produced and defended: where a metric is defined, how you get from a chart to the records underneath it, what happens when the data is incomplete, and how a report reaches somebody on a schedule. If your question is what to put in front of a board, the Solutions page is the better starting point.
- Quality and compliance
- Executive and finance
- Biomedical engineering
What is the reporting engine?
Definition
The Rydya reporting engine
The reporting engine is a registry of versioned metric definitions, each computed from the operational records rather than from a separate copy, each drillable to the rows that produced it, and each rendered into reports that can be scheduled, delivered and exported without losing that lineage.
The word that matters is drillable. A number you cannot trace is not evidence, it is a claim, and the difference becomes obvious the first time somebody senior disagrees with it. If the answer to "where did 82 per cent come from" is a fortnight of investigation, the report was decoration.
The second is versioned. Metric definitions change, because organisations learn. What must not happen is a definition changing silently, so that a trend line shows an improvement which is really a redefinition.
Why reports lose the room
Because a single indefensible number discredits the whole pack, and a pack assembled without lineage as a requirement usually contains at least one.
The failure is social before it is technical. A report goes to a board. Somebody who knows the estate spots a figure that cannot be right. Nobody can explain how it was calculated in the meeting, so the item is deferred. Every subsequent report from that source is read with suspicion, no matter how good it is, and the department loses the argument it was actually winning.
The most common cause is silent gap-filling. A device with no data becomes a zero, a missing month is interpolated, an incomplete return is quietly excluded. Each of these is individually defensible and collectively fatal, because the report now presents an estimate with the confidence of a measurement.
The second cause is the parallel truth. Reporting is built on an extract, the extract drifts, and now the report and the workspace disagree. Somebody notices, and the question stops being "what does this tell us" and becomes "which system is right", which is a question nobody in the room can settle.
How a number is produced and defended
Defined once, computed from the live records, traceable to rows, and honest about what is missing.
- 1
The metric is defined in a registry
Once, with a version, so the same words mean the same thing in every report. Compliance measured one way on a dashboard and another way in a pack is how two true reports contradict each other.
- 2
It computes from the operational records
The same rows the workspace shows, not a nightly extract living its own life. There is no second version of the truth to reconcile, because there is no second copy.
- 3
It drills down to the rows underneath
From a figure to the records that produced it. This is the feature that survives a hostile question, and it is the one most often missing.
- 4
Missing data is stated as missing
Not zero, not interpolated, not quietly dropped. Incomplete data never renders as compliant, because a report that hides its own gaps is worse than no report.
- 5
Delivery and export keep the lineage
Scheduled runs, real CSV, XLSX, HTML and JSON outputs, and secure links that expire. Exports are permission-checked and audited on download rather than being a URL anyone can forward.
What the engine refuses to do
Every refusal here is a place where a helpful default would produce a confident wrong answer.
It will not fill a gap
A device with no data is reported as a device with no data. Zero is a measurement, and using it for absence is the most common way a compliance figure ends up flattering and false.
It will not let missing data read as compliant
The absence of a failed test is not a passed test. This is the single most consequential rule in the engine, and it is enforced rather than left to a report author's judgement.
It will not invent money
An unpriced downtime hour stays unpriced. Original amounts and currencies are immutable, and conversions live separately with their rate id, source and timestamp, so a figure quoted last year still reproduces exactly.
It will not leak across scope
A report runs with the permissions and location scope of the person who ran it, enforced at the server and at the database row. A scheduled report does not become a way to receive data you could not otherwise see.
When a report is worth building, and who should own it
When somebody will make a decision on it. Reports that exist to demonstrate activity are a tax that departments pay in the hours they most need back.
The honest test for any recurring report is: what would happen if it stopped. If the answer is nothing, it has been running as a ritual, and a scheduled ritual costs a real person real hours every month. Most estates are carrying several, and the cost is invisible precisely because it is spread thin.
The reports worth automating are the ones somebody chases. Those have a named audience and a decision attached, which also means they have someone to complain when a figure looks wrong, which is exactly the pressure that keeps a report honest. Ownership by the person who needs the answer beats ownership by the person who knows the tool.
Questions
Can we trace a figure back to the records behind it?
Yes. Every metric drills down to the rows that produced it, and that traceability is the property the rest of the engine is built around. A number that cannot be traced is a claim rather than evidence, and the difference becomes very obvious the first time somebody senior disagrees with it in a meeting.
What happens when the data is incomplete?
It is reported as incomplete. Missing data is never rendered as zero, interpolated, or quietly excluded, and it never reads as compliant, because the absence of a failed test is not a passed test. A report that hides its own gaps presents an estimate with the confidence of a measurement, which is how a pack loses a board permanently.
Does reporting run off a separate copy of the data?
No. Metrics compute from the same operational records the workspace shows, so there is no extract to drift and no second version of the truth. When a report and a workspace disagree, the discussion stops being about the estate and becomes about which system is right, which is a question nobody in the room can settle.
Can a scheduled report send someone data they should not see?
No. Reports run with the permissions and location scope of the person who ran them, enforced on the server and at the database row by row level security. Exports are permission-checked and audited on download, and secure links expire, so a delivery is not a way around access control.
See it on your equipment
Live in an afternoon, useful the same week. A person replies, usually within one working day.
Contact us