What Makes Equity Research “Institutional-Grade”? A 12-Point Standard for Indian Markets

Every research shop in India claims its work is “institutional-grade”. Few say what that means. This guide sets a concrete standard, twelve checks grouped into four layers, that you can apply to any report on an Indian listed company: a broker note, a newsletter, an internal memo, or work from an outsourced desk.

The short version: institutional research is research someone can act on without redoing it. Everything below is a way of testing that.

Why “institutional-grade” needs a definition

The phrase gets used for anything with a DCF and a dark-blue cover page. But the people who actually consume institutional research (portfolio managers, family office CIOs, investment committees) judge it on a narrower question: does this change, or confirm, a decision I am about to make, and can I trust every number in it?

That framing kills a lot of common “research” habits:

• Restating the annual report. A business description is context, not research.
• A single target price. A point estimate hides the assumptions that matter. A PM wants the range and what drives it.
• Risks as boilerplate. “Competition may intensify” is not a risk analysis. A risk needs a mechanism, a magnitude and a trigger.
• Unsourced numbers. If a figure cannot be traced to a filing, a transcript or a named dataset, it cannot go into an investment committee deck.

The four layers of an institutional report

Think of a report as four layers stacked on top of each other. Each layer depends on the one beneath it. A brilliant thesis built on a broken model is worthless, and a perfect model with no thesis is a spreadsheet.

The 12-point standard

Score each check 0 (missing), 1 (partial) or 2 (fully met). A report scoring 18 or more out of 24 is usable by an investment committee. Below 12, treat it as a starting point, not research.

The radar below compares how a typical retail-style stock report and an institutional initiation score on these dimensions, grouped into six themes. It illustrates common patterns. It is not a statistical study.

Layer 1: evidence, and the India-specific traps

Indian disclosure is better than its reputation. Listed companies file quarterly results under SEBI’s LODR regulations, publish shareholding patterns every quarter, disclose promoter pledges, and upload investor presentations and call transcripts to the exchanges. The data is there. The trap is that the important data is often in places a casual reader skips.

• Related-party transactions (RPTs). Look for them in the annual report notes and in the separate RPT disclosures. Large, recurring payments to promoter-linked entities are a margin and governance question, not a footnote.
• Contingent liabilities. Tax disputes, guarantees given for group companies and litigation sit off the balance sheet. Compare them to net worth.
• Auditor’s report and CARO remarks. Qualifications, emphasis-of-matter paragraphs and auditor resignations are some of the strongest red flags available to an outside analyst.
• Segment reporting changes. When a company reorganises its segments, historical comparisons break. Rebuild history on the new basis before modelling.
• Promoter pledges. A high or rising pledge share can force selling in a drawdown. We cover this in detail in our guide to shareholding signals.

For a page-by-page approach, see how to read an Indian annual report like a buy-side analyst.

Layer 2: a model that earns trust

The model’s job is to make assumptions visible. Two tests separate a real model from a dressed-up spreadsheet:

1. It is driver-based. Revenue for a cement company is capacity × utilisation × realisation per tonne. For a private bank it is average advances × yield, minus cost of funds. If the model just grows revenue by a percentage, the analyst has not understood the business.
2. It reconciles. Three to five years of history should tie to reported numbers. If it does not, every forecast inherits the error.

Layer 3: judgement, the variant view

This is the layer that justifies paying for research at all. A variant view answers three questions:

What does the market believe? What do we believe instead? What evidence will settle it, and when?

If a report agrees with consensus on every driver, it may still be useful as a model or a data pack, but it is not a thesis. The variant view should be specific enough to be wrong. “We think margins expand” is weak. “We think gross margin reaches 38% by FY28 versus consensus at 34%, because the new plant’s mix shifts toward specialty grades; Q3 realisation per tonne is the first test” is strong.

Valuation then expresses that view as a range. On DCF mechanics for Indian companies, especially cost of equity and terminal value, see our DCF guide.

Layer 4: control, which is what separates a desk from a blog

Control is the least glamorous layer and the one most often missing. It is also why institutional investors trust some research providers and not others.

How to use this standard

• If you buy research: score the last three reports you received. Most providers do well on the model and badly on control. Ask for the review trail.
• If you write research: paste the scorecard into your template and refuse to publish below 18.
• If you are building a desk: the standard is mostly process, which is good news. Process can be taught to motivated analysts and enforced with automation. That is the premise behind student-led research desks and research automation.