ResearchMedium

3 min read · 612 words

What the seat actually does

Research produces a written view — on a company, a sector, a currency, a curve — and publishes it to clients under a named author. The output is an argument with a number attached: a forecast, a target, a recommendation, and the reasoning that is supposed to make the number checkable.

The useful part is rarely the recommendation. A price target is one number produced by a model with a dozen assumptions; the assumptions are the content, and a reader who takes the target and discards the model has taken the least durable part.

• Coverage — a list of names this seat is expected to have a view on, permanently.
• Maintenance — earnings, guidance, a data release: the model is updated and the change explained.
• The occasional real piece — the work that reframes something, which is what the seat is remembered for.
• Client conversations — the half of the job that is not published, and on many desks the half that is actually paid for.

A day, and where it goes

• Before the open — anything published overnight by the companies covered, and whether it changes the view or merely the numbers.
• Publishing — through a process: compliance reads it, supervisory analysts read it, and a rating change goes out to everybody at once rather than to whoever called first.
• The phone — clients asking the question the note did not answer, which is usually the one that mattered.
• Model work — the unglamorous maintenance that makes the next piece possible.

What it is measured on

• Client votes. Institutions rank the analysts they found useful, and the ranking allocates the commission or the fee. It measures being read, which correlates with being right only loosely.
• Whether the calls worked — tracked, argued about, and harder to attribute than it looks: a target hit for the wrong reason counts the same in a table.
• Since research was unbundled, an explicit price. Buy-side firms in Europe must pay for research separately from execution, which turned an unpriced service into a line item and shrank it.
• Not on banking revenue. That link is what the independence rules exist to sever, and severing it is why the seat looks the way it does.

What it touches on this site

• What a view is built on — reading financial statements and an annual report.
• What the numbers mean — valuation and reading a market number.
• Where the conflicts are — investor protection.
• Who reads it — the buy-side seats on this map, especially long/short equity, which is paid to disagree with it.

How it goes wrong

• The rating that never falls below hold. A distribution of recommendations with almost no sells is a distribution that has stopped carrying information.
• Access as the product. When what is really being sold is time with management, the written view becomes the ticket rather than the argument.
• Consensus as a hiding place. Being wrong with everybody costs less than being wrong alone, and that asymmetry shows up in the forecasts.
• A model with one assumption doing all the work, unstated. See valuation on terminal value.

Concepts to master

• A target price is a model output, not a forecast of a price. It says what the analyst's assumptions imply, and the assumptions are listed.
• Consensus is a number about analysts, not about a company. It is useful mainly as the thing a surprise is measured against.
• Unbundling changed the economics of the seat — what was paid for inside a commission is now invoiced, and priced accordingly.
• Independence is structural. Reporting lines, wall crossings and publication timing, not good intentions.