# CSI-IN Register Protocol **How a published league table of named coaches is compiled, checked, noticed and corrected.** Version 1.0 · 17 September 2026 Shortlist mode and register mode are the same arithmetic and completely different acts. Scoring four coaches for your own hiring decision is private research. Publishing a score about a named competitor is publishing. The second carries duties the first does not, and this document is those duties. Nothing here is optional: an entry compiled outside this protocol is not a CSI-IN entry. > **Not legal advice.** This is an editorial protocol written to make the register accurate > and defensible. Indian defamation law is both civil and criminal, and the register should > be reviewed by a media-law practitioner in India before the first edition publishes. The > protocol is designed to make that review short, not to replace it. --- ## 1. The governing rule **Absence of evidence is not evidence of absence.** In shortlist mode a buyer asked the coach a question and got nothing back. A zero is a fair record of that exchange. In register mode nobody asked. A zero would assert something the compiler cannot support — and would do it about a named person, in public, in a ranked table. So the register never publishes a zero as a finding. Everything below flows from this. ## 2. What may be used as a source An indicator may be coded from public evidence only where that evidence is: 1. **Published by or about the coach**, and reachable by the reader — a website, a credential register, a company filing, a book, a recorded talk, a bylined article, a court or regulatory record, a named press report. 2. **Dated**, or captured with a retrieval date and an archive link. 3. **Specific enough to map onto a single anchor** without inference. "Works with senior leaders" does not code CTX2. "Coached 14 CXOs at three listed companies since 2021" does. Never usable: private information, leaked material, anything behind a login the reader cannot obtain, hearsay, a competitor's opinion, a rumour, or the compiler's own recollection of a person. Reputation is not evidence. Neither is silence. ## 3. Every published point carries a citation Each coded indicator above zero records `{source, retrieved, quote}` in the register file. A code with no citation is not a low code — it is **not a code at all**, and reverts to T4 unrated. The published entry shows the citation for every cell. A reader who disagrees with a score can see precisely which document produced it and argue with that document. ## 4. The coverage floor, and the unrated list An entry publishes a score and takes a rank **only** if all three hold: | Test | Threshold | |---|---| | Evidence coverage κ | ≥ 0.50 | | Indicators carrying T1/T2 evidence | ≥ 14 of 21 | | Eligibility gates | passed on the public evidence | Otherwise the entry is **unrated**: no score, no band, no rank, and it is listed in a separate alphabetical section headed *"Coaches for whom public evidence was insufficient to score."* Three properties of this design matter: - **Unrated is not last place.** It is not in the ranking at all, and is ordered alphabetically so that no reader can infer an ordering that does not exist. - **Unrated is a statement about the record, not the coach.** The published wording says so explicitly, every time, in the same sentence as the name. - **Unrated is the expected outcome for most of the market**, and that is the most useful single finding the register produces. If the great majority of practising coaches in India cannot be scored from what they publish, the buyer has learned something real about the market's disclosure standards — without a word being said against any individual. ## 5. Notice before publication Every entry receives written notice **at least 14 days** before its first publication, containing: the full indicator codes, every evidence tier, every citation, the resulting score or unrated status, a link to the methodology, and an invitation to correct. Notice is not consent and is not sought as consent. It exists because a compiler who publishes without giving the subject a chance to correct the record has chosen not to be accurate. The register records `notice_sent` and `reply_received` for every entry, and both are published. **A coach who does not reply is recorded as "notice sent, no reply received."** Nothing more is inferred from silence — not in the score, not in the wording, not in the ordering. ## 6. Corrections and right of reply - Any coach may submit documented corrections at any time, in any edition. - Documented corrections are applied within **10 working days** and logged in the changelog with the date and the cell changed. - Where a coach supplies private evidence (an hours log, an indemnity certificate) it is used, tiered normally, and the citation records that it was supplied by the coach. - Where a coach disputes a code but supplies nothing, their statement of dispute is published **beside the entry, unedited**, and the code stands. The reader decides. - Any coach may request that their entry be removed and be listed only by name as *"asked not to be included."* That request is honoured without argument. That last provision costs the register a row and buys it the only thing that matters. ## 7. What is never published - A zero, a band or a rank derived from missing evidence (§1, §4). - Anything about a coach's conduct, competence, health, finances or character that is not a coded indicator with a citation. - Complaints, sanctions or disputes unless recorded by a credential body or a court and linked. - Client names a coach has not already made public. - Any comparison in words. The table ranks; the prose never says one coach is better than another. - An entry for a coach who has asked to be excluded (§6). ## 8. The compiler's conflict Register mode requires the compiler to code competitors. §12.3 of the methodology previously forbade this outright. It is amended, openly, rather than quietly dropped: > **The author may code a competitor only under this protocol, and never alone.** Three constraints replace the prohibition: 1. **Mechanical coding only.** Every code traces to a citation a third party can open and check against the anchor. There is no cell in the register where the compiler's opinion is the input. If an anchor cannot be settled by a document, it is coded T4 unrated. 2. **Independent adjudication of contested cells.** Any cell a coach disputes is re-coded by someone with no commercial interest in Indian executive coaching. Where that has happened, the entry says so. 3. **The author's own entry is coded by the adjudicator, not by him**, compiled the same way as every other, published first, and published wherever it lands (§12.1–12.2). He takes a rank like anyone else — the control is that he does not set his own codes and does not choose when the result goes out. A reader who believes the compiler has shaded a rival's score can test that belief in about two minutes, against the citation, without asking anyone's permission. That is the only conflict control that works at scale. ## 9. Editions The register publishes as a dated, frozen edition with a changelog. Codes, citations and the weight vector in force are published with it. A re-scored coach can always see which of a change in their evidence, a change in the method, or a correction moved them — because all three are versioned separately. ## 10. Before the first edition ships 1. Indian media-law review of this protocol and of a sample entry. 2. A written notice template, and a correspondence log. 3. The author's own entry compiled, published, and open for challenge (§12.2). 4. A named independent adjudicator for §8.2. 5. A dry run: compile 10 entries, send notice, publish nothing, and see what comes back. Item 5 is the one to insist on. It costs two weeks and tells you what the register will cost in argument before a single name is public.