K
Knowles Training Institute
Guest
- Thread Author
- #1
Knowles Knowledge Centre•Report Writing
A Report Has to Survive
A practical framework for writing reports whose findings outlive the advice attached to them.
A report, in the sense used here, is not a narrative of work done. It is a set of findings and a set of asks, written in one voice by one author, and read by somebody who may accept the findings and refuse the asks — or, far more often, refuse the asks and lose the findings with them. What follows is a question of what was actually observed, where the author moved from observation to conclusion, which of the recommendations matters most, whose work has to change for each one, and what evidence would overturn the whole thing.
Evidence-led guide•Singapore context•Updated September 2026
Explore the five-stage method
About this guide
Published byKnowles Training Institute
FrameworkKnowles Separable Finding Framework — an editorial synthesis, not an official standard
Evidence baseSkillsFuture Singapore • IAL • WEF • ILO • OECD
REPORT FINDING
MARKTHESEEN
OWNTHEJUMP
RANKTHEASK
NAMEWHOACTS
SHOWTHEUNDO
The conventional response to an ignored report is a better executive summary. A tighter opening, a page of headline findings, a cleaner chart. All of that makes the document easier to read, which was rarely the obstacle. What usually happened is that a reader disagreed with one recommendation, declined the paper, and in declining it discarded eleven observations that were sound and that nobody will gather again for two years.
This guide explains the five stages of the method, how it relates to Singapore’s Critical Core Skills, how a report differs from a briefing note, a status update and a recommendation paper, and the case in which using it is the wrong move.
The practical answer
For practical use, the method can be organised into five stages, applied to one report at a time:
1
Mark the seen
What was actually observed, as distinct from what it means?
2
Own the jump
Where did the author move from observation to conclusion?
3
Rank the ask
Which recommendation matters most, and which are optional?
4
Name who acts
Whose work has to change for each recommendation?
5
Show the undo
What evidence would overturn this conclusion?
About this framework
This five-stage structure is the Knowles Separable Finding Framework used in this article. It is an editorial synthesis for workplace application—not an official government framework, accreditation standard or universal definition.
The account of how a claim is licensed by its evidence comes from the study of practical argument; the account of how people climb from a selected observation to a conclusion in one unmarked step comes from organisational learning; the account of a decision as something that has to become somebody’s work comes from management writing; and the account of a claim as useful only when it says what would count against it comes from the philosophy of science.
Stephen
Stephen Toulmin argued that a practical argument has three visible parts and one that is usually invisible: the data, the claim, and the warrant that licenses the step from one to the other. Most disputes, on his account, are not about the data or the claim but about an unstated warrant that each party assumes differently. For a report this is the whole difficulty in one sentence. A reader who rejects a finding often accepts the observations and rejects the warrant, and a document that never states its warrant gives them no way to say so.
Chris
Chris Argyris described what is often called the ladder of inference: people select from what is available, interpret the selection, draw a conclusion, and then defend the conclusion rather than the selection, because the lower rungs have become invisible even to them. The practical consequence for a report is that the author is usually the last person able to see where their own jump happened, which is why marking it has to be a deliberate act rather than a matter of care.
Peter
Peter Drucker made the point that a decision which has not degenerated into work is not a decision but, at best, a good intention. Applied to a report, this is the test that most recommendations fail. “The handover process should be redesigned” names no person and changes nobody’s week. The same finding expressed as “the operations manager should publish a revised handover checklist by the end of the quarter” can be accepted, refused, or negotiated, which the first version cannot.
Karl
Karl Popper held that what distinguishes a serious claim is that it says what would count against it. A report conclusion written so that no observation could contradict it is comfortable to defend and useless to act on, because a reader cannot tell what would have to change for it to stop being true. Stating the undo is also the author’s own protection: a conclusion that names its own refutation is much harder to dismiss as opinion.
Warrant
Inference
Ownership
Refutability
Voice
The recurring pattern matters more than the labels. Across the study of practical argument, organisational learning, management writing and the philosophy of science, the same five ideas reappear: a claim is joined to its evidence by a warrant that is usually unstated; the step from selected observation to conclusion is invisible to the person who took it; a recommendation becomes a decision only when it has become somebody’s work; a conclusion that cannot say what would refute it cannot be acted on; and a document written in one undifferentiated voice forces a reader to accept or reject all of it at once.
The last of those disposes of the usual remedy. An organisation unhappy with its reports will reliably respond by issuing a template with a mandatory executive summary and a recommendations table. Neither asks the author to mark which sentences are observation, and the summary makes matters worse: it compresses findings and asks into a single paragraph, which is precisely the fusion that causes the loss. Two authors can complete the template identically, and the one who marked three sentences as observed is the only one whose work survives a refusal.
The first stage separates what was observed from what it means. It is the cheapest stage and the one most often skipped, because the author already knows which sentences are which and cannot imagine a reader who does not.
It includes:
An author has completed this stage when a reader can point to a sentence and say whether it is something that happened or something the author thinks. “The handover is chaotic” is not an observation. “Of the eleven handovers examined, seven had no written record of what was outstanding” is one, and a reader who dislikes the recommendation still has it.
For example, an operations review opens with “the current process is not fit for purpose”. Nothing in that sentence can be checked, agreed or disputed by anybody who was not in the review. Replaced with the count of handovers lacking a written record, it becomes a sentence the operations manager can confirm from their own files, and the argument moves to what should be done rather than to whether the problem is real.
02
The second stage names the inference and whose it is. Between the observation and the conclusion there is always a step, and the step is a judgement somebody made and could have made differently.
This is the stage that decides whether disagreement is possible. A reader who cannot see the inference can only agree or dismiss, and dismissing is cheaper, so a report with an invisible warrant tends to produce a short meeting and no decision.
They include:
The test is whether a reader can disagree precisely. “These seven cases indicate a process failure” hides the step. “We read these seven cases as a process failure rather than individual oversight, because the same omission appears across four different teams” exposes it, and invites the one objection worth having.
There is a second-order effect worth naming. Exposing the inference makes the report look less authoritative, which is exactly why authors avoid it, and exactly why a report that does it is treated as more credible by readers who have seen a hundred confident papers and acted on few of them.
03
The third stage separates the recommendations from the findings and puts them in order. A report with nine equally weighted recommendations has made no recommendation; it has handed the ranking back to the reader, who will do it badly or not at all.
They include:
A ranking is not produced by numbering a list. It is produced when the author can say which single recommendation they would keep if the reader accepted only one, and can defend that choice against the others. Nine unranked asks are usually a sign that the author did not want to make that choice in public.
For example, a review carries nine recommendations, of which two require a system change and seven are procedural. Ranked honestly, one procedural change accounts for most of the observed failures and costs almost nothing. Unranked, it sits ninth on a list headed by the system change, and the paper is declined on cost with the cheap fix inside it.
04
The fourth stage attaches each recommendation to the one person whose work changes. A recommendation addressed to the organisation is addressed to nobody, and will be accepted in principle, which is the form of acceptance that produces no activity.
It includes:
?
?
A named recommendation can answer:
Who does something differently?
What, exactly, and by when?
Do they have the authority?
What happens if they decline?
Naming the actor is the stage that costs the author something, because each name is a conversation they now have to have, often before the report is finished. That is the point. A recommendation nobody has been warned about arrives as an accusation, and the first response to an accusation is a dispute about the findings rather than a decision about the ask.
05
The fifth stage states what evidence would overturn the conclusion. It is the shortest section in any report and the one that most changes how the rest is read.
Showing the undo includes:
A conclusion that cannot be wrong cannot be acted on. The sentence to insist on is the one saying what would change the author’s mind.
This fifth stage connects the other four. The marked observations supply what is checkable; the exposed inference supplies what is arguable; the ranking supplies what is decidable; the named actor supplies what is doable. All of that produces a report a reader can engage with precisely — and precision is exactly what makes the conclusion vulnerable, which is why the author has to supply the refutation themselves rather than waiting for somebody less friendly to find it.
The stages are one sequence applied to a single report, not five sections of a template.
Consider an illustrative Singapore workplace scenario: an operations review into repeated delays at a site handover is declined by the steering group on cost, and an almost identical review is commissioned eighteen months later.
The sequence runs:
01
Mark the seen:
the original report is re-read for sentences that describe something that happened. There are eleven, buried in a narrative; the opening summary contains none of them, and the recommendation table contains none either.
02
Own the jump:
the inference is located. The review concluded a process failure from omissions in seven of eleven handovers across four teams, and never said so: the text moves from “delays were reported” to “the process is not fit for purpose” in two sentences, with the warrant left in the author’s head.
03
Rank the ask:
the nine recommendations are ranked. One procedural change — a written outstanding-items list signed at each handover — answers most of the eleven observations and costs nothing. It appeared sixth, under a system replacement whose cost is what the steering group declined.
04
Name who acts:
actors are attached. Seven of the nine recommendations name no role. The cheap one belongs to a single operations manager, who had not been shown the report and first heard of it as a decision that had already been refused.
05
Show the undo:
the limits are stated. Eleven of roughly ninety handovers in the period were examined, all from two sites; a second look should start with the other sites, and the conclusion would not survive finding written records in most of them.
The steering group’s first instinct was a shorter report. A shorter version of this report would have compressed eleven observations into the same unfalsifiable opening sentence and lost the cheap recommendation faster.
The stages depend on one another in a specific way: marked observations with no exposed inference produce a report that is checkable and unarguable, so a reader who doubts the conclusion has nothing to object to and declines the whole paper; and a ranked, named recommendation with no stated undo produces an action that proceeds past the point where the evidence for it has stopped holding, because nobody wrote down what would have told them.
Four situations should be named before anybody adopts this. Where the report is a statutory or regulatory submission in a prescribed format, the separation may simply not be available — the honest course is to supply the form as required and keep a separable internal version, rather than fight the format. Where the reader has already decided and the report is a formality, separating the finding will not change that decision; it preserves the observations for the next reader, which is often worth doing and should be said plainly rather than presented as influence. Where there is genuinely nothing to recommend, a report that observes and concludes and asks for nothing is a legitimate document, and a template that insists on recommendations will produce invented ones. And where the finding concerns a named person’s conduct, the separation of observation from inference is exactly the right discipline and is not sufficient: that document needs a process around it — confidentiality, a right of reply, somebody independent — and a report-writing method is not that process.
No. The four overlap, and they differ in what the reader is being asked to do.
Term
Primary emphasis
Examples
Briefing note
Equipping a reader for an imminent conversation
Position, history, sensitivities, lines to take
Status update
Telling an owner what moved since last time
Progress, slippage, blockers, next dates
Recommendation paper
Asking for one decision, with the options
Options, costs, risks, a preferred course
Report
Establishing findings that outlive the advice
Marked observations, exposed inference, ranked asks, named actors, stated undo
Researchers have a name for the confusion that a family of overlapping labels creates: the jangle fallacy, the assumption that two things must be different because they carry different names. “Review,” “assessment,” “findings paper” and “study” describe substantially the same territory with different emphasis, and nothing in this method depends on preferring one of them.
The practical consequence is twofold. Calling the document a report rather than a recommendation invites the question this article asks — what survives if the recommendation is refused? — which a recommendation paper does not invite, because a recommendation paper is the advice and has nothing else in it.
Conversely, the vocabulary should not be stretched over documents it cannot carry. A report is not a briefing for a meeting that starts in an hour, and it is not a running account of progress.
A status update, for example, remains a periodic account of movement rather than a report. Its author may well separate what they observed from what they infer, and should. They are not establishing a finding that has to outlive anything, and a status update written to this method would be four times as long and no more useful to the one person who reads it.
For a fuller decision guide, see reports vs briefings, status updates and recommendation papers.
8
SkillsFuture Singapore identifies 16 Critical Core Skills that support workers’ ability to develop and use technical competencies. They are arranged into three clusters:
Creative Thinking, Decision Making, Problem Solving, Sense Making and Transdisciplinary Thinking.
Building Inclusivity, Collaboration, Communication, Customer Orientation, Developing People and Influence.
Adaptability, Digital Fluency, Global Perspective, Learning Agility and Self Management.
Conceptual overlap — not an official equivalence
Mark the Seen + Own the Jump→ Thinking Critically
Name Who Acts→ Interacting with Others
Rank the Ask + Show the Undo→ Staying Relevant
Two things in that mapping are worth stating here rather than leaving to the crosswalk. The first is that Own the Jump is reached by no named skill. Several of the sixteen describe reasoning well and interpreting evidence; none describes making your own reasoning visible to somebody who may want to contradict it, which is a different and less comfortable act. The second is that Communication maps to Mark the Seen — the one stage that is not a communication problem at all. Separating what was observed from what it means is a matter of discipline in the author’s own notes, performed before anything is written for a reader. The frameworks should not be presented as interchangeable. SkillsFuture’s Critical Core Skills are an official Singapore framework; the five-stage structure is a simplified editorial model.
A fuller mapping is available in the practical crosswalk between Singapore Critical Core Skills and writing reports.
This crosswalk is not a claim that a Knowles Training Institute programme is SSG-accredited, WSQ-approved or SkillsFuture-funded.
Reports are increasingly read in fragments, by more people than they were written for, and quoted long after the author has left. A finding fused to its recommendation survives neither the refusal nor the quotation: the advice is declined and the evidence vanishes, or a sentence is lifted and nobody can tell whether it was something seen or something thought.
World Economic Forum 2025 employer survey
Analytical thinking was the most widely cited core skillIdentified as essential by around seven in ten surveyed employers
Employers also gave prominence to:
Resilience, flexibility and agility
Leadership and social influence
Technological literacy
Creative thinking
Curiosity and lifelong learning
International employer survey — not a Singapore-specific forecast.
Singapore’s Critical Core Skills similarly combine transferable cognitive, interpersonal, adaptive and digital capabilities with technical competence.
Issuing a tighter template will not by itself improve what gets acted on, because no additional section asks an author to mark which of their sentences a reader could check.
9
Yes, and the gains come from the separation rather than from the prose. A working cycle contains:
1
A list of what was actually observed
2
The inference written as its own sentence
3
Recommendations ranked, with the optional ones marked
4
One named role against each recommendation
5
A conversation with that person before circulation
6
A stated condition that would overturn the conclusion
Evidence from Singapore’s Institute for Adult Learning
762worker-learners
43educators and facilitators
8organisations
The Institute for Adult Learning’s Adult Learning Collaboratory examined future-oriented teaching methods across 762 worker-learners, 43 educators and facilitators, and 8 organisations, studying 86 learning interventions.
The work focused on emergent scenarios — which ask learners to respond to situations that evolve, change or remain uncertain — and generative dialogue, which asks learners to challenge assumptions and build understanding collectively rather than absorbing information individually.
The findings do not prove that every course using scenarios will produce the same result. They support contextual practice, dialogue and reflection over information delivery alone.
Start with the separation—not with the summary.
1
Which sentences describe something that happened?
2
Where does the text move from observation to conclusion?
3
Is the obstacle the writing, the ranking or the decision rights?
4
Who does something differently if this is accepted?
Question 3 matters most. A report can be impeccably separable and still change nothing: if the reader has no authority over the named actor, or if the paper arrives after the decision, the writing was never the constraint. A structured approach is set out in how to identify whether your reports change any decisions.
Example: a review declined, then commissioned again
The cause may be a fused summary, an unranked recommendation list, or an actor nobody told—not a shortage of analytical rigour.
The response may require marked observations, a ranked single ask, or a conversation with the named actor before circulation.
10
1. Opening with a conclusion nobody can check
A first sentence that characterises rather than counts gives the reader nothing to agree with. They must accept the author’s judgement or reject the paper, and rejecting is cheaper. The same finding expressed as a count can be confirmed from the reader’s own records, which moves the meeting on to what should be done.
2. Writing observation and inference in one voice
Where the two are indistinguishable, a reader who doubts the conclusion has to doubt the evidence as well. The author has made the data hostage to the argument, and when the argument is refused the data goes with it and is not gathered again.
3. Nine recommendations of equal weight
An unranked list hands the ranking to the reader, who ranks by cost. The cheap change that answers most of the findings sits below the expensive one, and the paper is declined on the expensive one with the cheap change inside it.
4. Addressing recommendations to the organisation
“The process should be redesigned” names nobody and will be accepted in principle. Acceptance in principle is the form of agreement that produces no work, and it is indistinguishable from a decision for about two quarters.
5. Letting the named actor read it first in a meeting
A recommendation nobody has been warned about arrives as an accusation, and the first response to an accusation is an argument about the findings. Ten minutes with the named person before circulation converts that argument into a negotiation about the ask.
6. Omitting what would overturn the conclusion
A conclusion with no stated limits is read as either certain or unserious, depending on the reader, and cannot be revisited sensibly when circumstances change. Naming the refutation costs two sentences and is the author’s own protection.
7. Compressing the summary until nothing is checkable
A one-page summary that fuses eleven observations into a characterisation and nine asks into a theme is the most-read and least-usable part of the document. It is also the part most likely to be quoted a year later by somebody who never read the rest.
A practical development cycle contains six steps:
1
Mark the seenList what happened, with its source, before interpreting it.
2
Own the jumpWrite the inference as its own sentence, with your name on it.
3
Rank the askSay which one you would keep if the reader accepted only one.
6
Name who actsOne role per recommendation, with what changes and by when.
5
Warn the actorTen minutes before circulation, not in the meeting.
4
Show the undoState what observation would overturn the conclusion.
This cycle prevents the organisation from treating a decision problem as a writing problem. It also creates a clearer basis for deciding whether a report is wanted at all, or whether what is wanted is a briefing or a recommendation paper.
Organisations that decide facilitated development is the right response can review the Writing Reports Training Course in Singapore, a commercial programme option for facilitated development across the five stages.
Course format, facilitator suitability, customisation and expected workplace application should be evaluated against the organisation’s identified needs.
What is a report, in this sense?
It is a document whose findings are built to outlive the advice attached to them. The method separates what was observed from what it means, writes the inference as its own sentence with an author attached, ranks the recommendations and marks the optional ones, names one role against each, and states what evidence would overturn the conclusion.
?
How is it different from a recommendation paper?
A recommendation paper exists to obtain one decision and is finished when the decision is made; if the answer is no, the paper has done its job and nothing is left. A report establishes findings that remain usable after a recommendation is refused, which is why the separation matters and why the document is longer.
?
Does this mean an executive summary is useless?
Not useless, but much more dangerous than its length suggests. A summary is the most-read and most-quoted part of the document, and compression is exactly the operation that fuses observation with inference. A summary that keeps three checkable counts is worth four that characterise.
?
What should be measured?
The proportion of reports whose opening paragraph contains a sentence a reader could check against their own records, the proportion in which the inference is written as a separate sentence, recommendations with a named role against recommendations made, how many named actors were spoken to before circulation, and how many declined reports had any finding reused afterwards.
?
When is this the wrong method?
Where the format is prescribed by a regulator and the separation is not available; where the decision is already taken and the report is a formality; where there is genuinely nothing to recommend and a template would invent something; or where the subject is a named person’s conduct, which needs a process around it rather than a writing method. In those cases the separation is not what is holding the document back.
A report is usually treated as the case for its recommendations and is better treated as a set of findings that have to survive those recommendations being refused.
An author needs to know which of their sentences a reader could check, where they moved from observation to conclusion, which recommendation they would keep if only one were accepted, whose work changes if it is, and what would prove them wrong. A tighter summary answers none of those, which is why an organisation can improve its templates and go on commissioning the same review twice.
For employers, the first step is not to revise the template. It is to take a report that was declined, count the sentences in it a reader could have checked, and find out whether any of its findings were ever used again.
Published by Knowles Training Institute. Evidence base: SkillsFuture Singapore, Institute for Adult Learning, International Labour Organization, World Economic Forum, OECD.
The post A Report Has to Survive Its Own Recommendation appeared first on Corporate Training Courses in Singapore.
A Report Has to Survive
Its Own Recommendation
A practical framework for writing reports whose findings outlive the advice attached to them.
A report, in the sense used here, is not a narrative of work done. It is a set of findings and a set of asks, written in one voice by one author, and read by somebody who may accept the findings and refuse the asks — or, far more often, refuse the asks and lose the findings with them. What follows is a question of what was actually observed, where the author moved from observation to conclusion, which of the recommendations matters most, whose work has to change for each one, and what evidence would overturn the whole thing.
Evidence-led guide•Singapore context•Updated September 2026
Explore the five-stage method
About this guide
Published byKnowles Training Institute
FrameworkKnowles Separable Finding Framework — an editorial synthesis, not an official standard
Evidence baseSkillsFuture Singapore • IAL • WEF • ILO • OECD
REPORT FINDING
MARKTHESEEN
OWNTHEJUMP
RANKTHEASK
NAMEWHOACTS
SHOWTHEUNDO
The conventional response to an ignored report is a better executive summary. A tighter opening, a page of headline findings, a cleaner chart. All of that makes the document easier to read, which was rarely the obstacle. What usually happened is that a reader disagreed with one recommendation, declined the paper, and in declining it discarded eleven observations that were sound and that nobody will gather again for two years.
This guide explains the five stages of the method, how it relates to Singapore’s Critical Core Skills, how a report differs from a briefing note, a status update and a recommendation paper, and the case in which using it is the wrong move.
The practical answer
The practical answer at a glance
For practical use, the method can be organised into five stages, applied to one report at a time:
1
Mark the seen
What was actually observed, as distinct from what it means?
2
Own the jump
Where did the author move from observation to conclusion?
3
Rank the ask
Which recommendation matters most, and which are optional?
4
Name who acts
Whose work has to change for each recommendation?
5
Show the undo
What evidence would overturn this conclusion?
About this framework
This five-stage structure is the Knowles Separable Finding Framework used in this article. It is an editorial synthesis for workplace application—not an official government framework, accreditation standard or universal definition.
Four traditions supply the method, and none of them is a template
The account of how a claim is licensed by its evidence comes from the study of practical argument; the account of how people climb from a selected observation to a conclusion in one unmarked step comes from organisational learning; the account of a decision as something that has to become somebody’s work comes from management writing; and the account of a claim as useful only when it says what would count against it comes from the philosophy of science.
Stephen
Toulmin
Stephen Toulmin argued that a practical argument has three visible parts and one that is usually invisible: the data, the claim, and the warrant that licenses the step from one to the other. Most disputes, on his account, are not about the data or the claim but about an unstated warrant that each party assumes differently. For a report this is the whole difficulty in one sentence. A reader who rejects a finding often accepts the observations and rejects the warrant, and a document that never states its warrant gives them no way to say so.
Chris
Argyris
Chris Argyris described what is often called the ladder of inference: people select from what is available, interpret the selection, draw a conclusion, and then defend the conclusion rather than the selection, because the lower rungs have become invisible even to them. The practical consequence for a report is that the author is usually the last person able to see where their own jump happened, which is why marking it has to be a deliberate act rather than a matter of care.
Peter
Drucker
Peter Drucker made the point that a decision which has not degenerated into work is not a decision but, at best, a good intention. Applied to a report, this is the test that most recommendations fail. “The handover process should be redesigned” names no person and changes nobody’s week. The same finding expressed as “the operations manager should publish a revised handover checklist by the end of the quarter” can be accepted, refused, or negotiated, which the first version cannot.
Karl
Popper
Karl Popper held that what distinguishes a serious claim is that it says what would count against it. A report conclusion written so that no observation could contradict it is comfortable to defend and useless to act on, because a reader cannot tell what would have to change for it to stop being true. Stating the undo is also the author’s own protection: a conclusion that names its own refutation is much harder to dismiss as opinion.
Warrant
Inference
Ownership
Refutability
Voice
The recurring pattern matters more than the labels. Across the study of practical argument, organisational learning, management writing and the philosophy of science, the same five ideas reappear: a claim is joined to its evidence by a warrant that is usually unstated; the step from selected observation to conclusion is invisible to the person who took it; a recommendation becomes a decision only when it has become somebody’s work; a conclusion that cannot say what would refute it cannot be acted on; and a document written in one undifferentiated voice forces a reader to accept or reject all of it at once.
The last of those disposes of the usual remedy. An organisation unhappy with its reports will reliably respond by issuing a template with a mandatory executive summary and a recommendations table. Neither asks the author to mark which sentences are observation, and the summary makes matters worse: it compresses findings and asks into a single paragraph, which is precisely the fusion that causes the loss. Two authors can complete the template identically, and the one who marked three sentences as observed is the only one whose work survives a refusal.
The five stages of the Knowles Separable Finding Framework
011. Mark the seen
The first stage separates what was observed from what it means. It is the cheapest stage and the one most often skipped, because the author already knows which sentences are which and cannot imagine a reader who does not.
It includes:
- Listing what was directly observed, counted or read.
- Recording where each observation came from.
- Keeping the count separate from the interpretation of it.
- Noting what was looked for and not found.
- Saying how much of the population was examined.
- Marking anything reported rather than seen.
- Resisting the urge to characterise while listing.
An author has completed this stage when a reader can point to a sentence and say whether it is something that happened or something the author thinks. “The handover is chaotic” is not an observation. “Of the eleven handovers examined, seven had no written record of what was outstanding” is one, and a reader who dislikes the recommendation still has it.
For example, an operations review opens with “the current process is not fit for purpose”. Nothing in that sentence can be checked, agreed or disputed by anybody who was not in the review. Replaced with the count of handovers lacking a written record, it becomes a sentence the operations manager can confirm from their own files, and the argument moves to what should be done rather than to whether the problem is real.
02
2. Own the jump
The second stage names the inference and whose it is. Between the observation and the conclusion there is always a step, and the step is a judgement somebody made and could have made differently.
This is the stage that decides whether disagreement is possible. A reader who cannot see the inference can only agree or dismiss, and dismissing is cheaper, so a report with an invisible warrant tends to produce a short meeting and no decision.
They include:
- Writing the inference as a separate sentence.
- Naming who drew it, including yourself.
- Stating the warrant that licenses it.
- Saying what else the same observations could mean.
- Flagging where the sample is thin.
- Distinguishing a cause from a correlation in the text.
- Letting a reader reject the inference and keep the data.
- Avoiding the passive voice that hides the author.
The test is whether a reader can disagree precisely. “These seven cases indicate a process failure” hides the step. “We read these seven cases as a process failure rather than individual oversight, because the same omission appears across four different teams” exposes it, and invites the one objection worth having.
There is a second-order effect worth naming. Exposing the inference makes the report look less authoritative, which is exactly why authors avoid it, and exactly why a report that does it is treated as more credible by readers who have seen a hundred confident papers and acted on few of them.
03
3. Rank the ask
The third stage separates the recommendations from the findings and puts them in order. A report with nine equally weighted recommendations has made no recommendation; it has handed the ranking back to the reader, who will do it badly or not at all.
They include:
- Listing the recommendations apart from the findings.
- Putting them in a stated order of importance.
- Marking which are optional and saying so plainly.
- Naming the one that matters most if only one happens.
- Removing the recommendations added for completeness.
- Attaching each one to the finding it answers.
- Saying which findings have no recommendation attached.
A ranking is not produced by numbering a list. It is produced when the author can say which single recommendation they would keep if the reader accepted only one, and can defend that choice against the others. Nine unranked asks are usually a sign that the author did not want to make that choice in public.
For example, a review carries nine recommendations, of which two require a system change and seven are procedural. Ranked honestly, one procedural change accounts for most of the observed failures and costs almost nothing. Unranked, it sits ninth on a list headed by the system change, and the paper is declined on cost with the cheap fix inside it.
04
4. Name who acts
The fourth stage attaches each recommendation to the one person whose work changes. A recommendation addressed to the organisation is addressed to nobody, and will be accepted in principle, which is the form of acceptance that produces no activity.
It includes:
- Naming one role per recommendation, not a committee.
- Saying what that person does differently, and by when.
- Checking that they have the authority to do it.
- Telling them before the report is circulated.
- Removing any recommendation nobody can be named for.
- Saying who decides, where the actor cannot decide alone.
- Recording a refusal as an outcome rather than a gap.
- Recording the date agreed, where one is agreed.
?
?
A named recommendation can answer:
Who does something differently?
What, exactly, and by when?
Do they have the authority?
What happens if they decline?
Naming the actor is the stage that costs the author something, because each name is a conversation they now have to have, often before the report is finished. That is the point. A recommendation nobody has been warned about arrives as an accusation, and the first response to an accusation is a dispute about the findings rather than a decision about the ask.
05
5. Show the undo
The fifth stage states what evidence would overturn the conclusion. It is the shortest section in any report and the one that most changes how the rest is read.
Showing the undo includes:
- Naming the observation that would contradict the conclusion.
- Saying what was not examined and might matter.
- Stating how long the finding can be relied on.
- Naming the conditions under which it stops holding.
- Saying what a second look should check first.
- Distinguishing a limitation from a caveat.
- Accepting that an unfalsifiable conclusion is an opinion.
A conclusion that cannot be wrong cannot be acted on. The sentence to insist on is the one saying what would change the author’s mind.
This fifth stage connects the other four. The marked observations supply what is checkable; the exposed inference supplies what is arguable; the ranking supplies what is decidable; the named actor supplies what is doable. All of that produces a report a reader can engage with precisely — and precision is exactly what makes the conclusion vulnerable, which is why the author has to supply the refutation themselves rather than waiting for somebody less friendly to find it.
How the five stages work together
The stages are one sequence applied to a single report, not five sections of a template.
Consider an illustrative Singapore workplace scenario: an operations review into repeated delays at a site handover is declined by the steering group on cost, and an almost identical review is commissioned eighteen months later.
The sequence runs:
01
Mark the seen:
the original report is re-read for sentences that describe something that happened. There are eleven, buried in a narrative; the opening summary contains none of them, and the recommendation table contains none either.
02
Own the jump:
the inference is located. The review concluded a process failure from omissions in seven of eleven handovers across four teams, and never said so: the text moves from “delays were reported” to “the process is not fit for purpose” in two sentences, with the warrant left in the author’s head.
03
Rank the ask:
the nine recommendations are ranked. One procedural change — a written outstanding-items list signed at each handover — answers most of the eleven observations and costs nothing. It appeared sixth, under a system replacement whose cost is what the steering group declined.
04
Name who acts:
actors are attached. Seven of the nine recommendations name no role. The cheap one belongs to a single operations manager, who had not been shown the report and first heard of it as a decision that had already been refused.
05
Show the undo:
the limits are stated. Eleven of roughly ninety handovers in the period were examined, all from two sites; a second look should start with the other sites, and the conclusion would not survive finding written records in most of them.
The steering group’s first instinct was a shorter report. A shorter version of this report would have compressed eleven observations into the same unfalsifiable opening sentence and lost the cheap recommendation faster.
The stages depend on one another in a specific way: marked observations with no exposed inference produce a report that is checkable and unarguable, so a reader who doubts the conclusion has nothing to object to and declines the whole paper; and a ranked, named recommendation with no stated undo produces an action that proceeds past the point where the evidence for it has stopped holding, because nobody wrote down what would have told them.
Four situations should be named before anybody adopts this. Where the report is a statutory or regulatory submission in a prescribed format, the separation may simply not be available — the honest course is to supply the form as required and keep a separable internal version, rather than fight the format. Where the reader has already decided and the report is a formality, separating the finding will not change that decision; it preserves the observations for the next reader, which is often worth doing and should be said plainly rather than presented as influence. Where there is genuinely nothing to recommend, a report that observes and concludes and asks for nothing is a legitimate document, and a template that insists on recommendations will produce invented ones. And where the finding concerns a named person’s conduct, the separation of observation from inference is exactly the right discipline and is not sufficient: that document needs a process around it — confidentiality, a right of reply, somebody independent — and a report-writing method is not that process.
Is a report the same as a briefing note?
No. The four overlap, and they differ in what the reader is being asked to do.
Term
Primary emphasis
Examples
Briefing note
Equipping a reader for an imminent conversation
Position, history, sensitivities, lines to take
Status update
Telling an owner what moved since last time
Progress, slippage, blockers, next dates
Recommendation paper
Asking for one decision, with the options
Options, costs, risks, a preferred course
Report
Establishing findings that outlive the advice
Marked observations, exposed inference, ranked asks, named actors, stated undo
Researchers have a name for the confusion that a family of overlapping labels creates: the jangle fallacy, the assumption that two things must be different because they carry different names. “Review,” “assessment,” “findings paper” and “study” describe substantially the same territory with different emphasis, and nothing in this method depends on preferring one of them.
The practical consequence is twofold. Calling the document a report rather than a recommendation invites the question this article asks — what survives if the recommendation is refused? — which a recommendation paper does not invite, because a recommendation paper is the advice and has nothing else in it.
Conversely, the vocabulary should not be stretched over documents it cannot carry. A report is not a briefing for a meeting that starts in an hour, and it is not a running account of progress.
A status update, for example, remains a periodic account of movement rather than a report. Its author may well separate what they observed from what they infer, and should. They are not establishing a finding that has to outlive anything, and a status update written to this method would be four times as long and no more useful to the one person who reads it.
For a fuller decision guide, see reports vs briefings, status updates and recommendation papers.
8
How does this relate to Singapore’s Critical Core Skills?
SkillsFuture Singapore identifies 16 Critical Core Skills that support workers’ ability to develop and use technical competencies. They are arranged into three clusters:
Thinking Critically
Creative Thinking, Decision Making, Problem Solving, Sense Making and Transdisciplinary Thinking.
Interacting with Others
Building Inclusivity, Collaboration, Communication, Customer Orientation, Developing People and Influence.
Staying Relevant
Adaptability, Digital Fluency, Global Perspective, Learning Agility and Self Management.
Conceptual overlap — not an official equivalence
Mark the Seen + Own the Jump
Name Who Acts
Rank the Ask + Show the Undo
Two things in that mapping are worth stating here rather than leaving to the crosswalk. The first is that Own the Jump is reached by no named skill. Several of the sixteen describe reasoning well and interpreting evidence; none describes making your own reasoning visible to somebody who may want to contradict it, which is a different and less comfortable act. The second is that Communication maps to Mark the Seen — the one stage that is not a communication problem at all. Separating what was observed from what it means is a matter of discipline in the author’s own notes, performed before anything is written for a reader. The frameworks should not be presented as interchangeable. SkillsFuture’s Critical Core Skills are an official Singapore framework; the five-stage structure is a simplified editorial model.
A fuller mapping is available in the practical crosswalk between Singapore Critical Core Skills and writing reports.
This crosswalk is not a claim that a Knowles Training Institute programme is SSG-accredited, WSQ-approved or SkillsFuture-funded.
Why does separability matter now?
Reports are increasingly read in fragments, by more people than they were written for, and quoted long after the author has left. A finding fused to its recommendation survives neither the refusal nor the quotation: the advice is declined and the evidence vanishes, or a sentence is lifted and nobody can tell whether it was something seen or something thought.
World Economic Forum 2025 employer survey
Analytical thinking was the most widely cited core skillIdentified as essential by around seven in ten surveyed employers
Employers also gave prominence to:
Resilience, flexibility and agility
Leadership and social influence
Technological literacy
Creative thinking
Curiosity and lifelong learning
International employer survey — not a Singapore-specific forecast.
Singapore’s Critical Core Skills similarly combine transferable cognitive, interpersonal, adaptive and digital capabilities with technical competence.
Issuing a tighter template will not by itself improve what gets acted on, because no additional section asks an author to mark which of their sentences a reader could check.
9
Can report writing be learned?
Yes, and the gains come from the separation rather than from the prose. A working cycle contains:
1
A list of what was actually observed
2
The inference written as its own sentence
3
Recommendations ranked, with the optional ones marked
4
One named role against each recommendation
5
A conversation with that person before circulation
6
A stated condition that would overturn the conclusion
Evidence from Singapore’s Institute for Adult Learning
762worker-learners
43educators and facilitators
8organisations
The Institute for Adult Learning’s Adult Learning Collaboratory examined future-oriented teaching methods across 762 worker-learners, 43 educators and facilitators, and 8 organisations, studying 86 learning interventions.
The work focused on emergent scenarios — which ask learners to respond to situations that evolve, change or remain uncertain — and generative dialogue, which asks learners to challenge assumptions and build understanding collectively rather than absorbing information individually.
The findings do not prove that every course using scenarios will produce the same result. They support contextual practice, dialogue and reflection over information delivery alone.
Which part of report writing should be fixed first?
Start with the separation—not with the summary.
1
Which sentences describe something that happened?
2
Where does the text move from observation to conclusion?
3
Is the obstacle the writing, the ranking or the decision rights?
4
Who does something differently if this is accepted?
Question 3 matters most. A report can be impeccably separable and still change nothing: if the reader has no authority over the named actor, or if the paper arrives after the decision, the writing was never the constraint. A structured approach is set out in how to identify whether your reports change any decisions.
Example: a review declined, then commissioned again
The cause may be a fused summary, an unranked recommendation list, or an actor nobody told—not a shortage of analytical rigour.
The response may require marked observations, a ranked single ask, or a conversation with the named actor before circulation.
10
Common mistakes when writing reports
1. Opening with a conclusion nobody can check
A first sentence that characterises rather than counts gives the reader nothing to agree with. They must accept the author’s judgement or reject the paper, and rejecting is cheaper. The same finding expressed as a count can be confirmed from the reader’s own records, which moves the meeting on to what should be done.
2. Writing observation and inference in one voice
Where the two are indistinguishable, a reader who doubts the conclusion has to doubt the evidence as well. The author has made the data hostage to the argument, and when the argument is refused the data goes with it and is not gathered again.
3. Nine recommendations of equal weight
An unranked list hands the ranking to the reader, who ranks by cost. The cheap change that answers most of the findings sits below the expensive one, and the paper is declined on the expensive one with the cheap change inside it.
4. Addressing recommendations to the organisation
“The process should be redesigned” names nobody and will be accepted in principle. Acceptance in principle is the form of agreement that produces no work, and it is indistinguishable from a decision for about two quarters.
5. Letting the named actor read it first in a meeting
A recommendation nobody has been warned about arrives as an accusation, and the first response to an accusation is an argument about the findings. Ten minutes with the named person before circulation converts that argument into a negotiation about the ask.
6. Omitting what would overturn the conclusion
A conclusion with no stated limits is read as either certain or unserious, depending on the reader, and cannot be revisited sensibly when circumstances change. Naming the refutation costs two sentences and is the author’s own protection.
7. Compressing the summary until nothing is checkable
A one-page summary that fuses eleven observations into a characterisation and nine asks into a theme is the most-read and least-usable part of the document. It is also the part most likely to be quoted a year later by somebody who never read the rest.
Turning a framework into a working practice
A practical development cycle contains six steps:
1
Mark the seenList what happened, with its source, before interpreting it.
2
Own the jumpWrite the inference as its own sentence, with your name on it.
3
Rank the askSay which one you would keep if the reader accepted only one.
6
Name who actsOne role per recommendation, with what changes and by when.
5
Warn the actorTen minutes before circulation, not in the meeting.
4
Show the undoState what observation would overturn the conclusion.
This cycle prevents the organisation from treating a decision problem as a writing problem. It also creates a clearer basis for deciding whether a report is wanted at all, or whether what is wanted is a briefing or a recommendation paper.
Organisations that decide facilitated development is the right response can review the Writing Reports Training Course in Singapore, a commercial programme option for facilitated development across the five stages.
Course format, facilitator suitability, customisation and expected workplace application should be evaluated against the organisation’s identified needs.
Frequently asked questions
?What is a report, in this sense?
It is a document whose findings are built to outlive the advice attached to them. The method separates what was observed from what it means, writes the inference as its own sentence with an author attached, ranks the recommendations and marks the optional ones, names one role against each, and states what evidence would overturn the conclusion.
?
How is it different from a recommendation paper?
A recommendation paper exists to obtain one decision and is finished when the decision is made; if the answer is no, the paper has done its job and nothing is left. A report establishes findings that remain usable after a recommendation is refused, which is why the separation matters and why the document is longer.
?
Does this mean an executive summary is useless?
Not useless, but much more dangerous than its length suggests. A summary is the most-read and most-quoted part of the document, and compression is exactly the operation that fuses observation with inference. A summary that keeps three checkable counts is worth four that characterise.
?
What should be measured?
The proportion of reports whose opening paragraph contains a sentence a reader could check against their own records, the proportion in which the inference is written as a separate sentence, recommendations with a named role against recommendations made, how many named actors were spoken to before circulation, and how many declined reports had any finding reused afterwards.
?
When is this the wrong method?
Where the format is prescribed by a regulator and the separation is not available; where the decision is already taken and the report is a formality; where there is genuinely nothing to recommend and a template would invent something; or where the subject is a named person’s conduct, which needs a process around it rather than a writing method. In those cases the separation is not what is holding the document back.
Conclusion
A report is usually treated as the case for its recommendations and is better treated as a set of findings that have to survive those recommendations being refused.
An author needs to know which of their sentences a reader could check, where they moved from observation to conclusion, which recommendation they would keep if only one were accepted, whose work changes if it is, and what would prove them wrong. A tighter summary answers none of those, which is why an organisation can improve its templates and go on commissioning the same review twice.
For employers, the first step is not to revise the template. It is to take a report that was declined, count the sentences in it a reader could have checked, and find out whether any of its findings were ever used again.
Published by Knowles Training Institute. Evidence base: SkillsFuture Singapore, Institute for Adult Learning, International Labour Organization, World Economic Forum, OECD.
The post A Report Has to Survive Its Own Recommendation appeared first on Corporate Training Courses in Singapore.
Last edited: