The three objects in every situation
Start with the thing that makes this difficult: facts arrive already interpreted, and the interpretation does not announce itself. "The finance director blocked the request" feels like an observation. It is not one. What happened is that the request was submitted on the ninth and no approval had been issued by the twenty-third. "Blocked" imports an actor, an intention, and a decision, none of which you observed.
Three objects sit inside almost every professional situation, and they arrive fused.
The fact. What happened, with a date, stated so that someone else present would report roughly the same thing. It contains no verbs of intention. It survives being read out to the person it concerns.
The inference. What you concluded from the fact. Inferences are not errors. Most professional work is inference and you could not function without it. The error is not making them, it is losing track of which of your sentences are them.
The unknown. What you would need to know to tell whether the inference holds. This is the object people skip, because writing it down converts a settled question into an open one, and open questions cost effort.
Everyone believes they already distinguish fact from inference. The belief is not the capability. Do this in your head and you will conclude, correctly, that it changes nothing, because inference only stops passing as fact when it is written in a different column with nothing else in it.
Here is one situation, run through.
Fact. The grant agreement required a narrative report by 30 June. On 4 July the programme officer searched the shared inbox and found no report. The partner's last email on file is dated 12 May.
Inference. The partner has not produced the report. The partner is struggling with capacity. This is part of a pattern with this partner.
Unknown. Whether the report was sent anywhere. Which address the partner holds for us. Whether anyone has asked.
The three inferences in that example are not equally supported, and they cost different amounts to check. The first would be settled by one phone call. The second is a general judgement that would need several instances. The third is a claim about a pattern, and a pattern requires the other members of the pattern to be written down somewhere, which they almost never are.
Notice what the split has produced. Before it, there was a single sentence available to management: the partner failed to submit. After it, there is a decision with a five-minute action in front of it.
The specific verbs to watch are the ones that place something inside another person: ignored, refused, dismissed, blocked, delayed, disagreed, resisted, prioritised, avoided. Each of them describes a mental state you did not observe and an event you did. Replace the verb with the event and the sentence gets narrower, more useful, and considerably harder to argue with.
"Procurement dismissed our specification and is dragging the process out to protect a preferred supplier."
"The specification was submitted on 2 May. On 30 May procurement returned it with four queries, two of which had been raised and answered in the March meeting. No revised timetable has been issued."
The second version is what you can say in a room without a fight, and it is also the version that lets someone else spot the answer. The two repeated queries are the interesting fact, and they were invisible in the first version because the first version had already decided what they meant.
There is a second discipline hiding inside this one. When the split is honest, the fact column is usually much shorter than people expect, and the missing rows are almost always the same kind: nobody wrote down when something was received, by whom, or what was said in a conversation. That is a finding about your organisation's records, not about the situation, and it is worth collecting separately. Most of the recurring judgement failures in an organisation are not failures of reasoning. They are failures of provenance that only look like reasoning failures at the point where someone has to decide.
One warning about doing this well. The split can be performed as a ritual: the fact is written vaguely enough to contain the conclusion, and the unknown column is filled with things nobody could ever find out. The test is whether the unknown column contains at least one item that could be resolved this week by one action, and whether you have named the action.
Which of these is written as a fact rather than an inference?
You write a situation split and your unknown column reads: "whether senior management really supports this project". What is wrong with it?
Notes are kept with your account, alongside your progress and your gate claims. The lesson itself is readable without one.
This lesson has a tool
Open it and get your draft reviewed. Drag-and-drop tools need a wider screen; the review works anywhere.