On this page
An analytical result has little value if it cannot be understood by the people who need to act on it. Yet sharing a result creates its own risk: the more polished and portable the presentation becomes, the easier it is for the evidence behind it to disappear.
The problem begins after the answer
Imagine that someone has investigated a movement in revenue. They have selected a period, established the measure, checked the calculation and decided that the result is worth discussing. They now need to explain it to a colleague who did not take part in the original investigation.
A screenshot can communicate the number. It usually cannot communicate everything that was required to establish it. The recipient may not know which dataset was used, how the measure was defined, whether there were exclusions or why this particular movement was considered noteworthy. If the screenshot is copied into a board pack, the relationship to the source analysis can weaken further.
The question I set myself was not how to export a nicer chart. It was how to carry a governed analytical result into a different context without letting the presentation become a new source of analytical truth.
Analysis and reporting have different responsibilities
The analytical engine establishes supported calculations. The evidence and historical record preserve information about how those calculations were reached. A report has a different job: to arrange selected material for an audience, provide appropriate explanation and retain the basis of the result being communicated.
That means a report should be able to change its structure, order, introduction or explanatory language without silently changing the underlying figures. Different readers may need different levels of detail. What they should not receive is a different calculation merely because the report has been written for them.
Research figure 01
Evidence must survive communication
What changes when a result becomes a report
- GOVERNED ANALYSISEstablish the resultAn authorised operation produces supported values
- EVIDENCE RECORDKeep the basis inspectableRetain relevant calculation and source relationships
- REPORT COMPILATIONArrange and explainA presentation layer works with established results
- PUBLICATION CONTROLReview and shareExplicit publication and access decisions are made
- READER JUDGEMENTUnderstand and questionA recipient can trace the conclusion to its basis
An editorial model of responsibility, not a claim that every export route is certified.
In determinIQ this separation has influenced the Reports workspace, the historical Ledger and the report-publishing model. It is also the reason I regard a Management Report as a structured evidence-bearing document, not a collection of application screenshots.
A report should be traceable without becoming unreadable
Provenance is useful only if people can reach it. But putting every technical detail at the top of a management report can make the actual business question harder to see. Good communication requires a deliberate hierarchy.
A reader should first understand what was established and why it matters to the purpose of the report. Supporting detail should be available in a recognisable form: sources, relevant questions, supporting calculations, qualification and important limits. The arrangement may differ between an everyday operational report and a board or assurance-oriented document, but the source relationship should remain intelligible.
This is another application of progressive disclosure. The reader should not be forced through an audit trail to understand a simple result, and the person who needs to scrutinise that result should not be left with an unsupported statement.
Frozen evidence and changing presentation
One particular engineering problem emerged in work on comparison reports. An earlier saved comparison could retain a historical presentation format, while a newer view needed a clearer multi-series chart and chronological table. It would have been tempting to regenerate the underlying analytical values for the new layout. That would cross the wrong boundary.
The November 2026 implementation documentation instead describes a presentation-only adapter that reads frozen report values, arranges compatible comparison series for the reader and leaves the stored analytical evidence unchanged. The same authority is used by individual report inspection and Management Report compilation. This is a concrete example of the architectural distinction between a calculation and the way it is shown.
It is important not to overstate what that proves. The documented tests cover specific historical comparison and report-presentation cases. They do not amount to universal certification of every publication, export or print route. Native document generation and live print behaviour have also required separate development and verification.
Controlled publication is part of the analytical lifecycle
There is a further distinction between creating a report and making it available to someone else. An unfinished analysis may be useful to its author but unsuitable for wider circulation. A working copy may contain provisional wording. A report shared externally may need access controls and a clear publication state.
The determinIQ reporting architecture includes controlled sharing and persistent report records. These are not cosmetic administrative features. They answer basic questions about which material was deliberately made available, how it relates to its source and whether a later version should be treated as a separate artefact.
The presence of sharing controls does not remove the need to assess recipients, authorisation, sensitive content or retention policy. Publishing a document makes an information-governance decision as well as a design decision.
Why layout matters to evidence
During the later product work, I found that document layout was not independent of the trust model. Repeating a report title as though it were an analytical question can misrepresent the source. An unexplained label can make an audit record harder to interpret. Reordering dates visually without moving their values consistently can create a persuasive but false comparison.
Some of these were ordinary presentation defects. Their consequences were not ordinary. They reinforced the need to treat reporting transformations as explicit, testable operations rather than assume that anything downstream of analysis is harmless.
The documentation records work on report provenance labels, source-report identification, chronological proof rows, front matter and publication-safe presentation. Those details have the same purpose as the larger architecture: a reader should be able to tell what has been established and where it came from.
What remains to be tested
The October 2026 white paper established reporting as part of the end-to-end research proposition. Subsequent source documentation records implemented report and comparison-presentation behaviour, including tests of particular frozen-value and historical-report cases.
The remaining assurance questions are practical. Does each share and export route preserve the required evidence? Can a recipient understand the distinction between a source report and a new analytical claim? Are protected identities handled appropriately in every output? Do a print view and a native document retain the intended meaning? These need checks against the actual route and its current version, not confidence borrowed from a different test.
For me, the useful endpoint of analytics is not a number that looks convincing. It is an explanation that can travel to the people who need it without leaving its evidential basis behind.
Connected work
This study follows the work on evidence and verification. It also connects to the more recent comparison investigations, in which a change of presentation must not be mistaken for a fresh analytical calculation. The broader research programme examines how those separate responsibilities fit within one product.
Research context
Research programme: Making data analysis understandable, reproducible and accountable
Topic: Evidence and verification
Explore further
Supporting material is listed below.
Further reading is listed below.