How to read analyst reports

An analyst report should be read starting from its assumptions, not from its conclusion. The rating and the target price are the output of a model, and the model rests on the author's assumptions about revenue, margins, capital expenditure and the discount rate; change any of them and the conclusion changes too. The working order of reading is therefore the reverse of the usual: start with the date and the horizon, then the assumptions, then the factual part (it can be checked against primary data), and only at the very end the recommendation. The "buy" label adds no information to the report — it compresses the report into a single word, and what gets lost in the compression is exactly what makes the report worth reading.

What in a report is a claim and what is data

Every report mixes layers of differing reliability. Reported figures — revenue, debt, operating cash flow for the period — are verifiable: they are taken from the financial statements, and a discrepancy with the original source means either an error or a different methodology (adjusted EBITDA, restatement under IFRS versus Russian accounting standards, exclusion of non-recurring items). Forecast figures are unverifiable by definition. Valuation judgements ("the stock is cheap relative to its sector") are already a conclusion drawn from the forecast, and they inherit all of its weaknesses.

The habit of sorting these layers into separate boxes makes reading fast. Check the factual part against the financial statements section, take the forecast part as the author's version of events, and treat the valuation part as a matter for debate. How the statements themselves are structured, and where non-recurring items hide in them, is covered separately: how to read an income statement.

The cut-off date matters more than the publication date

A report has two dates: the day it was published and the state of the data it was calculated on. A gap between them is normal, and it explains most of the puzzlement along the lines of "the analyst missed the obvious". The model may have been closed before the results came out, before a dividend was announced, before the key rate was changed. Check what has happened to the stock since the cut-off using the corporate events calendar and the news feed — and only then judge whether the conclusion is out of date or not.

The same place reveals a common sleight of hand: a target price without a horizon is meaningless. A "target" for a year and a "target" for a quarter are figures of a different nature, and if the horizon is not stated, the report is unfinished.

Who pays and what follows from it

Sell-side reports are written by those who at the same time sell market access, arrange issues and hold positions of their own. This does not make them dishonest — it sets a systematic bias: there are noticeably fewer "sell" recommendations in the public flow than "buy" ones, and a rating downgrade more often comes after the price has moved, not before. Look for the disclosures section at the end of the document: whether the author or the organisation holds a position, any role in placing the issuer's securities, whether the research was paid for by the issuer. The absence of such a section is in itself a signal about quality.

The legal status of a particular document is determined by the document itself. If it is not marked as an individual investment recommendation addressed to you, it is an opinion, not a recommendation, and nobody bears responsibility for the outcome.

How to check a report without building your own model

A full re-check requires a model, but a quick check for internal consistency does not.

Step 1 — verify the multiples. If the report calls the stock cheap, look at the current value yourself: 3,78 and the instrument card {{instrument:SBER}}. A discrepancy with the report usually means that the author used forecast earnings instead of reported earnings — and then the comparison is being made with a future that does not exist yet.

Step 2 — find the source of growth. A conclusion about a re-rating always rests on some driver: volume growth, the price cycle for the company's products, the end of a capital expenditure programme, a rate cut. If the driver is not named, the conclusion is not substantiated.

Step 3 — check the dividend part. When a significant share of the expected return comes from payouts, look not at the analyst's forecast but at what has been declared: {{dividend_calendar|limit=5}} and the dividends section. What in a dividend policy is an obligation and what is an intention is a subject for a separate analysis.

What a report cannot do

A report does not know your portfolio, your horizon or the drawdown you can tolerate. It assesses the stock in isolation, whereas the decision is made within a portfolio, where the size of the position and its correlation with everything else matter. So even a correct report does not translate directly into action: between the conclusion "undervalued" and the purchase stands the question of position size, which the report does not answer.

The second limitation is the sample. The exchange does not publish an aggregate consensus on Russian securities, and it can only be partly assembled from the reports that are available: not all of them are published, and those that are published are skewed towards large issuers with liquid securities. A small stock without coverage does not mean "bad"; it means "uncovered". Reference points for your own analysis are the list of stocks and the glossary of terms.

How to actually use a report

The value of a report rarely lies in its conclusion. It lies in the way the problem is mapped out: the analyst has done the work of isolating what the company's results depend on and has listed risks you may not have known about. Read the report as a list of questions for the issuer, not as an answer. Then a difference between your conclusion and the author's stops being a problem and becomes the substance: you know exactly which assumption you disagree on, and you can keep watch on precisely that.