How we score and write notes.

Tipsheet exists to make Indian-equity disclosures readable without losing the source. This is the standard we hold every note to.

The scoring rubric

Every disclosure is scored 5–10 by our upstream classifier. The score reflects how materially the filing changes what an investor would think about the company — not how dramatic the language is.

• 9–10 — material change, immediately consequential. New guidance, regulatory action, large M&A, governance break.
• 7–8 — notable but expected, or significant with caveats. Order wins, credit rating changes, concall takeaways that move the thesis.
• 5–6 — incremental, routine, or already-known. Earnings releases, transcripts that confirm prior guidance, board approvals of routine items.

Anything below 5 we do not publish. The upstream classifier filters it out before we see it.

How Tipsheet is produced

Every Tipsheet article is grounded in a primary source — a BSE or NSE filing, a regulator notice, an earnings transcript, or company-disclosed material. The upstream classifier reads the primary source, identifies what is materially new, and writes a structured rationale: what changed, why it matters, what to watch.

That rationale is then rendered into a published article by a large language model, under editorial supervision. The model does not select what to cover, does not interpret the filing, and does not introduce facts that are not in the source. Every article is verified against the source by our automated validator before it appears on the site.

For earnings calls, the source material is the Tijori Concall Monitor analytical block (which includes the management-consistency check that flags where management contradicts a prior call). We summarise their analysis in 120–180 words, surface the most consequential angle, and link readers back to Tijori for the full analysis. We do not republish their work.

Tipsheet does not write from secondary sources. We do not aggregate other publications' work. We do not run advertorials. We do not predict prices.

The note format

Every note has the same six parts so the reader knows where to look:

• Headline — sentence case, the company is the subject, journalist verbs only.
• The number — the single most consequential figure with one line of context.
• What's new — three bullets of material facts, boilerplate stripped.
• Why it matters — two to three sentences of editorial commentary.
• What we're watching — forward-looking points that would confirm or break the thesis.
• The full read — a tight prose paragraph synthesising what happened, what it means, and what to watch.

Grounding rules

Every concrete claim — every number, date, name, percentage — must appear in the source rationale. Editorial framing is permitted only when it is a fair characterisation of facts in the source. We do not invent sector context. We do not put words in management's mouth. We do not speak for "investors" or "the market" unless our source explicitly does.

Outputs that fail this grounding test are automatically rejected by our pipeline and either re-generated with feedback to the model or held back from publication.

Banned phrases

Certain phrases produce no information and erode trust. We do not use them:

• "It is worth noting", "Importantly", "Notably", "Of course"
• "Going forward", "In recent months", "In today's fast-paced environment"
• "Leverage", "navigate", "unlock", "transformative", "game-changing", "robust"
• "Investors will…", "The market expects…", "Some analysts believe…"
• "Pursuant to", "As per the filing", "In a regulatory filing"

Source links

Every note links to the company's BSE and NSE announcements pages so you can verify our reading against the primary source. We do this because our rationale is downstream of the actual filing, and you should be able to check our work.

What we do not do

• We do not write clickbait. Headlines are sentence case and specific.
• We do not run paid placements or sponsored content.
• We do not offer investment advice. Our articles are editorial reads, not recommendations.
• We do not republish the source rationale verbatim — we rewrite it as news.

How to read a note

If you have thirty seconds, read the headline and "The number". If you have two minutes, also read "What's new" and "Why it matters". If you want the full picture, read "The full read" — it stands on its own.

This methodology is a living document. We update it when we change how we work.