Before you let someone make materials alone, some weaknesses can be tolerated. But one cannot: can they trace every written word back to its source? This is the single floor you must never soften by averaging it into a total score.
Like the screening before you board a plane: no passing on average
Picture airport security. Ten bags, nine are fine, one hides a blade. No one says, "On average that's 90 percent, so let it through." One dangerous item and you stop. This is the idea of a floor. A floor is the minimum line that, once broken, means failure no matter how good everything else is.
When you judge whether someone can make materials alone, skill in presentation or word choice can be a little weak and made up for elsewhere. But "can they return to the source" is a floor. The source is the original material behind a claim: the trial paper or the data. When you point at one sentence and ask, "Which page, which number in which document does this come from?", they must trace it straight back. Someone who cannot do this cannot be trusted alone.
Persuasion can never fill the hole left in sourcing. So we make "can they return" the absolute floor.
Why returning to the source is the highest condition: grounding sets the ceiling
Think of a cook. However beautiful the plating, if the meat has gone bad, the dish cannot be served. The ceiling on presentation is set by how safe the ingredients are. Materials work the same way: however clever the delivery, if the underlying facts are not solid, the more skillful it is, the more dangerous. Fidelity to fact (grounding) sets the ceiling on design skill.
Grounding means a claim keeps its feet firmly on the original data. With feet on the ground, a strong wind (the wish to sell) will not knock it over. With feet in the air, skillful words carry the error further. So among the eight abilities, source-grounding alone gets special treatment: we do not dissolve it into an average; we demand it high on its own.
| Situation | One who can return | One who cannot |
|---|---|---|
| "What number is this difference?" | Shows the figure and value in the original paper on the spot | Answers from memory: "It was a good result, I think" |
| Checking a chart against source data | Axis and range match the paper | Redrew it for the slide; differs from the original |
| When asked for grounds | Returns to the source, confirms, then answers | Deflects to a person: "The professor said so" |
Breaking the floor, case 1: asserting that something "does not" happen
Picture a health checkup. For a test that was not run, a doctor does not write "no abnormality." If it was not examined, you write "unknown." That is honesty. Yet in materials, cases are reported where this line is crossed.
In a reported case, with no supporting data, a document flatly stated, "There is no increase in mortality risk." On inquiry, the maker said, "Because it is not clear at present, I wrote it this way." If it is not clear, the right answer is "we do not know," yet they asserted "there is none." In another case, for an important potential risk, they emphasized only that "a low risk is expected."
This is the moment where the sin of omission overlaps with motivated reasoning. The sin of omission is making something look safe by not speaking, by not saying "we do not know." Motivated reasoning is having the conclusion you want to sell come first, so the reading of data is pulled toward it. The maker does not feel they are lying. That is what is frightening. Someone who easily asserts the absence of something (proving a negative) is someone without the habit of returning to the source. The floor is broken.
Breaking the floor, case 2: claiming an effect from just nine patients
Picture a galley proof, the final check before printing. One wrong character against fact gets fixed. "Close enough" does not pass. The number of data points is the same: asserting "it works" from too few cases is a floor with no foundation under it.
In a reported case, a claim of effect rested on just nine patients (four versus five), shown in a chart with no statistical analysis at all. A difference among nine people may be mere chance. Flip a coin nine times and get more heads, and you still cannot say "this coin favors heads." Someone who can return to the source stops themselves here: "This number of cases cannot support that." Someone who cannot return looks only at the good-looking chart and proceeds with "there is a difference."
| Deviation case | Psychology behind it | Ability that stops it |
|---|---|---|
| Asserting "risk does not increase" without grounds | Sin of omission + motivated reasoning | Source-grounding (write unknown as unknown) |
| "It works" from 9 cases, no statistics | Local rationalization (just this one chart) | Source-grounding (see the limits of the data) |
| "The professor said it was fine" | Externalizing responsibility | Source-grounding (return to data, not a person) |
How to test whether they can return: read it from L1 to L4
Picture a driving license. Someone who can only drive a taught route differs from someone who reads the signs and drives safely in an unknown town. Source-grounding has stages too. L1: can attach a source to a single told sentence. L2: as a habit, attaches sources to every claim. L3: understands why that source is needed and can read its limits, such as case numbers and axes. L4: can design the checking system itself, a procedure where anyone can trace back to the original.
To be trusted alone, a person must reach at least L3. The danger of nine cases, or of asserting an absence, cannot be stopped just by returning to the source; you must also read what that source can and cannot say. The test is simple. Point at a few lines in the material and ask, "Which document, which number does this come from?" They go straight back, the axis, range, and case count match, and they can tell apart what can and cannot be said. Only then have they cleared the floor.
The judgment to avoid most is passing someone below the floor on the strength of their persuasive skill or likable character. High design times low fidelity equals persuasive misperception, the most dangerous combination. So while other abilities can be read as a total score, source-grounding is judged on its own as pass or fail.
Who Can Draft Unsupervised ── Map of all 10 episodes
- Vol. 1: How to Tell Who Can Build Materials Alone — Don't Judge by the Average of Skills ── Readiness to work alone is decided not by the average of eight skills, but by a floor: can the person return to the source.
- Vol. 2: Bending the Facts Is Far Heavier Than a Plain Mistake ── A persuasive piece that creates a false impression does far more harm than a dull but correct one.
- Vol. 3: The More Skilled the Communicator, the Less Their Misleading Slips Get Noticed ── Skilled presentation makes misleading framing look reasonable, so the most fluent creators are the ones whose errors slip through.
- Vol. 4: Fail One, Fail All ── Appeal Cannot Patch a Hole in the Facts ── Set the bar for independent work as a non-compensatory floor, not an average — strong appeal can never fill a gap in the source.
- Vol. 5 (this episode): Demand One Thing Above All: Can They Return to the Source ── Tracing every claim back to its source is the floor that persuasion can never fill in for.
- Vol. 6: Making It Look Good and Making It Right Are Two Different Things ── Visual polish and fidelity to the source are different skills. Judging correctness by beauty lets the most dangerous errors slip through.
- Vol. 7: People Who Are Sure Their Own Material Is Fine Cannot Be Trusted Alone ── Measure the gap between self-assessment and actual skill as a separate, independent gate from persuasive ability.
- Vol. 8: Four Gates — Draft, Self-Review, Source-Check, Balance-Check ── Tell whether someone can release materials alone by whether they pass four gates in order.
- Vol. 9: Judging Three People by One Standard: The Persuader, the Precise Craftsperson, and the Quiet but Trustworthy ── A story that lines up three character types against one floor (can they return to the source) and decides pass or fail by a non-compensatory gate, not by an average.
- Vol. 10 (final): The Judge's Responsibility — Line Up the Reference Cases, and Let a Person Make the Final Call ── The person who decides pass or fail must anchor the standard with real reference cases, not just words, and make the final call under their own name rather than leaving it to a checklist or a machine.
Returning to the source is a floor that persuasive skill or a likable character can never fill in for. Blanket assertions, claims from just nine cases, deflecting responsibility to a person: these are signs the floor is broken, and they must not be overlooked by averaging them away.
The test is always the same. Point at a few lines and ask, "Which document, which number is this?" They go straight back, the axis and case count match, and they distinguish what can and cannot be said. Only such a person can be trusted to make materials alone.
- Judge source-grounding on its own, pass or fail. Other abilities can be a total score, but "can they return" is not dissolved into an average; break it and you fail.
- Blanket assertions and tiny samples break the floor. Asserting an absence without grounds, or proceeding on nine cases, proves the habit of returning to the source is missing.
- The bar for working alone is L3 or above. Not just returning to the source, but reading its limits, such as case numbers and axes, before being trusted alone.
- Ministry of Health, Labour and Welfare, Compliance and Narcotics Division (commissioned project). Report on the Monitoring of Promotional Information Activities for Prescription Drugs (March 2024 and prior years). Flagged cases are published with company names anonymized; the deviation patterns cited here are generalized from this report.
- Ministry of Health, Labour and Welfare. Guidelines on Promotional Information Activities for Prescription Drugs. Principles of evidence-based information and explicit labeling of unestablished matters.
- Japan Pharmaceutical Manufacturers Association. JPMA Code of Practice. Principles of accurate, fair, objective information and explicit sourcing.
- Standards for Fair Advertising of Drugs and Quasi-Drugs (MHW Notification 1980, as amended). Standards on aligning efficacy claims with supporting source materials.