How to Set Up an Equity Research Desk in India: A 90-Day Blueprint

A new investment firm should build its research system before research fragments into private spreadsheets, inboxes and individual memory. The first 90 days are the easiest time to decide what evidence the firm trusts, how decisions are recorded, who owns each assumption and what happens after a company enters the portfolio.

The technology comes second. A new PMS, AIF, family office or independent research firm does not begin with an “AI strategy.” It begins with an investment mandate and a repeatable path from information to decision.

This blueprint is about that path. It is operational, not legal advice; registrations, disclosures, custody, client reporting and compliance obligations should be designed with the firm’s qualified advisers against the current Indian rules.

Before day one: write the operating mandate

The investment mandate should be specific enough to constrain the research process.

Write down:

• the securities and markets in scope;
• market-cap, liquidity and ownership constraints;
• sectors or structures the firm excludes;
• the expected holding period and portfolio concentration;
• the role of quantitative versus qualitative evidence;
• who can propose, challenge and approve an investment;
• what requires a documented exception;
• what must be monitored after purchase.

Without these choices, software becomes a collection of features looking for a process.

Days 1–15: define the research universe and source policy

Start with the universe the desk can actually research and hold. More coverage is not automatically better. A small team with a disciplined 300-company universe can be more effective than a broad database nobody reviews consistently.

For each data family, define the source and the rule:

Data family	Policy question
Financial statements	Consolidated or standalone? What is the fallback?
Prices and corporate actions	How are splits, bonuses and dividends handled?
Filings and annual reports	Which exchange or company document is authoritative?
Concall transcripts	How are speaker errors and missing calls treated?
Management guidance	Is exact wording preserved and later graded?
Ownership	Are percentages, pledges and quarter-on-quarter changes consistently defined?
Alternative and macro data	What is the economic link to the company or sector?

Also decide the historical rule. If a screen or backtest asks what would have passed in June 2022, it should use only information that was knowable then. Point-in-time discipline belongs in the foundation, not as a repair after the first strategy looks suspiciously good.

Days 15–30: standardise the company research packet

Every company should enter the desk through the same minimum packet, even if the final analysis differs by sector.

A useful packet contains:

1. business and segment map;
2. five to ten years of sourced financial history where available;
3. sector-specific KPIs;
4. management guidance and delivery history;
5. cash-conversion and forensic checks;
6. ownership, pledging and governance changes;
7. valuation history and peer context;
8. key risks and disconfirming evidence;
9. the initial model or scenario range;
10. proposed thesis guideposts.

Standardisation does not mean every company gets the same score. A bank and a manufacturer should not be forced through identical economics. It means the firm agrees which questions cannot be skipped.

Days 30–45: build models with a clean boundary

Separate reported facts, calculations and assumptions.

• Reported facts should link to primary documents.
• Calculated values should use explicit, consistent formulas.
• Forecast assumptions should have a named owner, date and rationale.

Excel is often the right modelling surface for a new desk because the team can inspect formulas and run scenarios quickly. What Excel should not become is the sole database, filing archive, task tracker and monitoring engine. Those jobs become fragile as soon as more than one analyst touches the work.

AI can accelerate document reading, spreading history and checking model consistency. It should not silently choose the growth rate, terminal multiple or position size.

Days 45–60: design the decision record

An investment committee memo should make the decision inspectable, not merely persuasive.

Require:

• the thesis in one paragraph;
• the variant view;
• key operating drivers;
• base, upside and downside assumptions;
• the valuation method and sensitivity;
• disconfirming evidence;
• portfolio impact and concentration;
• the conditions that would trigger a review;
• the named decision and date.

The memo is not the end product. It is the initial state of a living record. When a later result changes an assumption, the desk should be able to reconstruct what it believed before the result arrived.

Days 60–75: turn the philosophy into rules

Most new firms can describe what they like in prose: clean balance sheets, durable growth, trustworthy management, reasonable valuations. The next step is to identify which parts can be made observable.

Examples include:

• an eligibility gate on liquidity or leverage;
• a quality score combining returns on capital and cash conversion;
• a valuation rule relative to the company’s own history;
• sector-specific KPI thresholds;
• an explicit governance veto;
• a review rule after a material guidance miss.

Rules do not eliminate judgement. They expose where judgement entered. If the portfolio manager overrides a score, the reason can be recorded and later evaluated rather than disappearing into memory.

Days 75–90: make monitoring part of the original design

Do not wait for the first portfolio company to surprise the desk.

For every holding, name the three to six conditions most likely to change the thesis. They may include volume, pricing, margin, order inflow, funding mix, asset quality, working capital, capex timing, guidance or a governance event.

Then define the monitoring loop:

new filing or datapoint
        ↓
attach it to the correct company and period
        ↓
compare it with the thesis, model and guidance
        ↓
route material exceptions to a named reviewer
        ↓
record the conclusion and update the watch condition

This is more useful than a generic feed of price moves and news headlines. The alert should explain why the event may matter to this portfolio.

The minimum viable research stack

A new desk needs seven connected capabilities:

1. Source archive: filings, results, calls and primary documents.
2. Structured data: financials, KPIs, ownership, factors and relevant market data.
3. Analysis: screens, peer comparisons, scorecards and models.
4. Decision workflow: memo, challenge, approval and exceptions.
5. Portfolio context: holdings, weights, exposures and constraints.
6. Monitoring: company-specific guideposts and material alerts.
7. Memory: forecasts, decisions, outcomes and lessons.

The firm can buy, build or combine these layers. The test is whether the evidence survives the hand-offs between them.

What to automate first

Automate work that is repetitive, high-volume and verifiable:

• collecting filings;
• extracting and standardising reported data;
• finding changes between documents;
• checking arithmetic and data freshness;
• running screens and scorecards;
• watching a universe for defined events;
• preparing first-pass evidence packs.

Keep humans explicitly responsible for:

• choosing the question;
• interpreting the industry;
• setting forecast assumptions;
• judging management and governance;
• deciding exceptions;
• sizing the position;
• communicating the decision.

The goal is not to remove the analyst. It is to stop paying analyst salaries for transcription and inbox monitoring.

Where Altys fits

Altys for new investment firms provides an India-first research foundation that can be configured around a team’s own universe, screens, scorecards, models and monitoring rules.

It brings point-in-time financials, filings, concalls, guidance, ownership, factors, mutual funds, macro and relevant alternative data into one research layer. Calculations are explicit, material claims remain linked to sources, and analytical work can be exported to Excel for independent verification.

A new firm does not need to become an AI engineering company to use AI well. It needs reliable data, explicit rules, human decision rights and a system that remembers what happened. That is the research infrastructure Altys is built to support.

Next, use the investment research software checklist to evaluate vendors, or read build versus buy before deciding what your team should own.