The one idea
Almost every sentence people call a fact is three things fused together: something observed, something concluded from it, and something preferred. Pull them apart and most confident statements get much smaller — and much more useful.
What did someone actually observe, and what did they add on top of it?
Tell the difference between what you saw and what you decided it meant.
Label each line of your analysis as evidence, inference, assumption or recommendation, so a reader can see where your reasoning could break.
Distinguish observation from interpretation from unstated premise. An argument is only as strong as its weakest unstated premise, and that is usually the part nobody wrote down.
The four things
| Type | What it is | Test |
|---|---|---|
| Evidence | Something observed, measured or recorded | Could someone else check it and get the same result? |
| Inference | A conclusion you drew from evidence | Could the same evidence support a different conclusion? |
| Assumption | Something you treated as true without checking | If this were false, does the conclusion collapse? |
| Opinion | A preference or value judgement | Would a reasonable person with different priorities disagree? |
None of these is bad. Inference is the whole point of thinking. Opinions are what decisions are made of. The failure is presenting an inference in the grammar of evidence, because then nobody knows which part to check.
The worked example
Here is a sentence of the kind you meet weekly, in a newsletter, a vendor deck, or a LinkedIn post:
“Company X increased sales 50% after adopting AI.”
Read it fast and you have learned that AI increases sales by 50%. Read it properly and you have learned almost nothing. Work through what is in the sentence.
What you actually know
Two things, and only if the source is honest:
- Someone reported that Company X's sales figure was about 50% higher in one period than another period.
- Somewhere in that stretch of time, Company X started using something they are calling AI.
That is the entire evidence content of the sentence.
What was added on top
The word "after" doing the work of "because". Two things happening in sequence is not one causing the other. This is the commonest move in business writing, and it is done with a preposition rather than a claim, which is why it slips past.
50% of what? Fifty percent growth from ₹2 lakh to ₹3 lakh is a slow month. Fifty percent on a base of ₹50 crore is a different story entirely. A percentage without its base is not a number, it is a mood.
Which period, compared to which? If the "before" period was their worst quarter, any comparison flatters. If the "after" period includes a festive season, Diwali did some of the work.
What else changed? Companies adopting a new tool are rarely doing only that. They may have hired salespeople, raised marketing spend, launched a product or cut prices — any of which produces the same number.
How many companies are there? One company is an anecdote. If a vendor tells you about one customer, ask what happened to the others. The ones who did not grow do not get case studies written about them.
Fact Sales at the reported company were higher in the later period.
Recommendation Before treating this as evidence, get the base figure, the comparison period,
and what else changed in the business.
Opinion A single company's result is not a reason to change our own tooling.
How to do this to any sentence
Taking a claim apart
1 of 5Underline the verbs that claim causation. Increased, drove, led to, resulted in, boosted, caused. Each one asserts a mechanism. Ask what would have had to be measured for that verb to be earned.
A colleague writes: “Support tickets dropped 40% after we published the new help articles, so the articles are working.”
Evidence: ticket count fell. Inference: the articles caused it. Assumption not stated: nothing else changed.
What else could it be? A bug was fixed. A big customer left. The contact form moved somewhere harder to find. It was December and half the users were on holiday. Any of these produces the same graph.
The fix is one line of work: check whether the drop is concentrated in the ticket categories the articles cover. If it is spread evenly across every category, the articles are not the cause.
In an interview you say: “I improved the process and reduced errors.”
The interviewer hears an unbacked inference and asks how you measured it. If your honest answer is “people said it felt better”, say that — it is a real observation. “We were rejecting around one in ten submissions; after the checklist it was closer to one in twenty-five over the following month, though volume was lower that month too” is a far stronger answer than a confident round number, because it shows you know which part is soft.
Try this
Take this sentence apart before reading on:
“Employees who use our productivity app work 25% faster, according to our internal study of 200 users.”
List: what is evidence, what is inference, what is assumed, and one specific reason the number could be true and meaningless at the same time.
Your challenge
Level 3 · IndependentFind a real claim this week — a vendor page, a news headline, a LinkedIn post, or a slide in your own company's deck. In under 200 words produce:
- The claim, quoted exactly.
- What is actually evidenced.
- Every inference the sentence makes without saying so.
- One assumption that, if false, sinks the whole thing.
- The honest rewrite.
Success criterion: your rewrite should be something the original author could not object to, while being noticeably less impressive than what they wrote.
What people usually get wrong
- Treating "after" as "because". The commonest error in business writing, and it is invisible because it lives in a small word.
- Accepting a percentage without a base. "Up 300%" from three to twelve is a rounding error in most companies.
- Assuming the absence of evidence is evidence of absence. "No studies show harm" often means nobody has looked.
- Rejecting a claim because you dislike its source. Interested parties are sometimes right. Their claims need checking, not dismissing — that is the subject of who-is-telling-you-this.
- Over-applying this and never concluding anything. Every conclusion rests on some assumption. The job is to know which ones you are standing on, not to refuse to stand.
How someone experienced does it
Experienced people label their own confidence inside the sentence, and it makes them more trusted rather than less. "Measured", "estimated", "I'm inferring", "my read is", "I could not verify this" are precise instruments, not hedges. Someone who uses them consistently gets believed when they say "this one is certain".
The second habit is stating the falsifier. Add the sentence that would break your claim: "If the drop is spread evenly across categories rather than the ones the articles cover, this explanation is wrong." Handing people the tool to disprove you is, counterintuitively, what makes them stop trying.
The third is noticing when a disagreement is really a definitional one wearing a factual coat. Two people arguing about whether the project "succeeded" are usually agreed on every fact and using two different definitions of success. Nothing is resolved until someone says that out loud.
When not to use this
Do not run this in conversation as a correction habit. Taking apart a colleague's sentence in a meeting reads as point-scoring even when you are right, and it stops people telling you things.
Use it silently on things you are about to repeat, forward, or write your name on. The test is not "is this person being sloppy" — it is "am I about to become the source of this claim for someone else?"
Why the same evidence supports opposite conclusions
Two people can read the same dataset and reach opposite conclusions without either lying, because a conclusion is evidence plus a bridge, and the bridge is made of assumptions.
Sales rose 50%. Bridge A: the market is expanding and we should hire. Bridge B: we cut prices and bought unprofitable revenue. Same evidence, opposite actions. Arguing about the 50% will never settle it. Naming the two bridges will.
This is why the useful question in a disagreement is rarely "what is your evidence?" — usually both sides have the same evidence. It is "what are you assuming that I am not?" when-sources-disagree takes this further, into conflicts between sources rather than between colleagues.
Prove it
Take the last document you wrote that made any claim — a report, a summary, a recommendation, an assignment. Go through it and mark every sentence E, I, A or O: evidence, inference, assumption, opinion.
Count how many inferences are written in the grammar of evidence. That number is your baseline. Rewrite the three worst offenders and keep both versions.
Keep learning this
Paste this into any AI assistant. It turns the assistant into a tutor that tests you instead of just answering you.
Act as an experienced practitioner who is good at teaching. I have just learned separating evidence from inference in a specific claim. Assume I am intelligent but relatively new to this — treat me as beginner level. Work through this in order, and wait for my reply at each step: 1. Ask me 5 questions that test whether I actually understood separating evidence from inference in a specific claim. Do not reveal the answers yet. 2. After I answer, tell me which parts I got right, which I got wrong, and which I only half-understand. Explain only what I misunderstood — do not re-teach what I already know. 3. Give me one practical challenge based on something I could genuinely encounter at work or in daily life. Do not solve it for me. 4. Evaluate my solution the way an experienced person would judge it, including what a professional would have done differently. 5. Tell me what to learn next, and why that comes next. 6. Give me trustworthy sources for deeper study — prefer official documentation, primary research or standards bodies over blogs and videos. Rules for you: no buzzwords. No motivational filler. Say "I'm not certain" when you are not certain, and tell me which parts of your answer I should verify myself. Clearly separate facts from your recommendations and your opinions.
Become independent at this
Use this when you want a path from where you are to actually good, with checkpoints you can test yourself against.
I want to become independently capable at taking a claim apart — not permanently dependent on AI, tutorials or step-by-step guides. Design a progression for me with five stages: Beginner, Guided practice, Independent practice, Real-world application, Professional level. For each stage tell me: - what I must know - what I must be able to do without help - the mistakes people make at this stage - one practical challenge - one real project that would prove I reached this stage - one way I can test myself honestly Then tell me the signals that I am ready to move to the next stage, and the signals that I have skipped ahead too early. Keep the theory to the minimum I actually need. Focus on ability I can transfer to situations you and I have not discussed.