Where to Find Data for Models: Terminals, Databases, Screeners & Filings

At 8:58 a.m., an analyst is two minutes away from sending a valuation model - and the revenue number from the screener does not match the annual report. One number is convenient; the other is official. In real finance and consulting work, knowing where data comes from is often the difference between a sharp model and a dangerous one.

The big picture is simple: data sources sit in a trust-speed trade-off. Filings are slower but authoritative; screeners are fast but derivative. Strong analysts do not ask, “Which website has the number?” They ask, “What level of proof does this decision need?”

Core Explanation: The Four Places Analysts Actually Find Data

Data sourcing is not about memorising websites. It is a decision rule: match the source to the job. The same company can require four different sources in the same model - filings for revenue, terminals for beta, databases for industry growth and screeners for peer discovery.

Definitions You Can Say in One Breath

The Source Selection Ladder

When time is short, follow this ladder. Start from the business question, not from the website you like. A comps task, DCF task, credit note and market-sizing task all need different evidence.

When to Use Which Source

Here is the practical mapping interviewers expect. Notice the pattern: official source for facts, professional platform for speed, database for history, screener for discovery.

The Trust-Speed Matrix

This is the mental shortcut professionals use. If you need a number in 30 seconds, a screener may be fine. If the number drives a buy/sell recommendation, move to filings or a terminal-backed official dataset.

How to Judge Whether a Data Source Is Good

Do not say “Bloomberg is reliable” or “Screener is easy.” Say how you would test the source. These six checks make your answer sound like an analyst, not a browser user.

A Simple Data-Sourcing Playbook for Common Placement Tasks

Case Study: Trent and the Zudio Growth Story

Situation: Trent became a frequent discussion point among Indian equity analysts because Zudio expanded rapidly in value fashion while Westside remained the more established lifestyle format. The story sounded attractive, but a serious model could not stop at “Zudio is growing fast.” The analyst had to verify revenue growth, store expansion, margins, capital intensity and peer positioning.

The move: A disciplined analyst would build the dataset in layers. The annual report and stock exchange filings provide the official financial statements and management discussion. Investor presentations help understand store count, format strategy and segment commentary where disclosed. A terminal or database helps compare valuation and historical ratios against retailers such as Avenue Supermarts, V-Mart Retail or Aditya Birla Fashion and Retail. A screener helps identify initial peers, but the peer set must be checked manually because grocery retail, value fashion and premium fashion have different economics.

Outcome and lesson: The primary driver of a strong Trent analysis is not just “retail growth”; it is the disciplined separation of official reported performance from market narrative. Supporting drivers include correct peer selection, format-level understanding, store expansion context, margin interpretation and valuation consistency. The case proves the core rule: use screeners to discover the story, but use filings and structured data to defend the model.

How AI Changes Where to Find Data

AI does not remove the need for source judgment. It makes source judgment more important because answers can look polished even when the underlying citation is weak.

1. Filing extraction becomes faster: LLMs can summarise annual reports, extract risk factors, compare management commentary across years and find relevant notes faster than manual PDF search. The analyst must still verify the extracted number against the original page.
2. Search shifts from keywords to cited answers: Tools like Perplexity can surface filings, exchange announcements and credible articles with citations. This is useful for discovery, but the final input should still come from the primary source.
3. Data cleaning becomes semi-automated: AI can map messy line items, standardise company names and flag outliers in a peer set. The risk is false matching - for example, confusing consolidated and standalone numbers or different fiscal year-ends.

Interview Relevance

Common Mistake

What to Revise Next

Now that you know where reliable data comes from, revise Case Study: A Timed Modelling Test Like the Ones Firms Set. That is where source judgment turns into execution - downloading the right inputs, making clean assumptions and building a defensible model under time pressure.