Writing Up What the Evidence Supports
How to write a finding that survives being read by somebody else: what each record supports, what is inferred, what is disputed, what is missing.
One finding, written twice
The second and third statements are supported and checkable. The first and fifth are conclusions written as findings. This is one employer's own file.
Write what each record supports, not what you concluded from it. A finding that says the employee was absent is a conclusion; a finding that says no system recorded activity between 10:20 and 12:40, from four named systems, is a statement anybody can check.
The evidential discipline in “Writing Up What the Evidence Supports” should also govern workforce technology. A team evaluating dual n back training in a transparent workflow for dual n back training can use time and project records as operational context, but should preserve the original record, document access and let the employee correct a misleading entry before it supports a conclusion.
The second survives being read by somebody who was not there. The first invites the question of how it was arrived at, and that question is usually asked at the worst moment.
For a separate benchmark relevant to “Writing Up What the Evidence Supports”, consult the Cornell Legal Information Institute evidence overview. Use it to test scope and safeguards against an external standard before the process is approved.
The four parts of a write-up
What the records show, record by record, in their own terms.
What is being inferred from them, stated as inference.
What the person said, fairly summarised, including the parts that are inconvenient.
What was looked for and not found, including the systems checked that held nothing.
Separating the inference
This is the whole discipline of this collection applied at the end. "The badge was presented at 07:51; we infer the employee presented it, because it is issued to them and was not reported lost."
Two sentences. The first is a fact, the second is reasoning, and separating them means the reasoning can be examined rather than smuggled through as a finding.
Including what was not found
The systems checked that held nothing, the explanations considered and rejected, the comparison group that showed the same pattern.
An employer that includes those has written an investigation. One that includes only the supporting material has written a case, and the difference is visible to anybody who reads both kinds.
Precision that matches the sources
Where the sources are accurate to five minutes, the finding is accurate to five minutes. Where the clocks were never measured, say so and reduce the claim accordingly.
Manufactured precision is the most common tell in these documents and the easiest to avoid: round the conclusion to the accuracy of its weakest input.
Language
Plain, factual, free of characterisation. No adjectives about the person, no speculation about motive, nothing that could not be read out.
This matters more here than in most employment documents because the subject is honesty, and a document that editorialises about somebody's honesty is doing the thing it is accusing them of being careless about.
The summary at the front
One page: what was looked at, what was found, what is disputed, and what the writer considers established and not established.
That page is what a decision-maker reads and it is what should be defensible on its own. If the summary says more than the body supports, the body is the correct version and the summary is wrong.
Handing over to a decision
The write-up ends at the facts. Recommending an outcome collapses the separation between investigating and deciding, and leaves the decision-maker either agreeing with somebody else or overruling a colleague.
That separation is covered in general employment practice and is worth restating here because time cases are frequently handled by one person throughout, for reasons of scale.
The sentences to avoid
Certain formulations appear in almost every write-up in this area and each of them conceals the step this collection is about.
- "The records show the employee was absent" — the records show silence.
- "The system confirms" — a system attests; it does not confirm.
- "There is no explanation for" — there is no explanation in the file.
- "It is clear that" — if it were, the sentence would say what.
- "The employee failed to account for" — they gave an account; it was unsupported.
Each has an accurate replacement that is one word longer, and the replacement is what survives being read by somebody else.
Length
Shorter than most of them are. A write-up that runs to twenty pages is usually twenty pages because everything found was included rather than because that much was relevant.
A summary page, a findings section organised by claim, and the material as appendices. The discipline of putting the detail in an appendix is what forces somebody to decide what the finding actually is.
What it is worth later
Everything. The write-up is what survives: the people move on, the systems are replaced, the logs are purged, and the document remains.
A document that says what each record supported, what was inferred, what was disputed and what was not found can be read in five years by somebody with no context and still make sense — which is the only test that matters for anything written in this area.