Analyst coverage

On a real desk the analyst does not start from a blank page every morning: they cover a list of names, carry a view on each, and say out loud when that view changes and why. The Researcher does the same for your book. For every name you hold and the names on your focus list it keeps a rating (Buy, Hold or Avoid), a conviction from 1 to 5, a one-paragraph thesis, the drivers that actually move the name, the next catalysts with dates, and the levels or conditions that would change its mind. Every brief it writes starts from that list.

Read it on Research › Coverage, and per name in the symbol sheet's Analyst view (open any symbol chip). A name the next brief will be asked to look at again carries a due for review tag with the reason (moved 8 % since the last review, through one of its own levels, past a dated catalyst, or simply not reviewed for six briefs) and sorts to the top of the table.

What the Researcher does each brief

The morning and midday brief jobs are handed the current coverage list together with what changed since each name was last looked at: the price now against the price at the last review, whether it has gone through the view's own bull or bear level, whether a dated catalyst or a results date has passed, and how many briefs have gone by since the review. From that the brief must:

1. 1Initiate the held and focus names it has not rated yet: held names first, a handful per brief, and only names it has a price and news for. A name the tools return nothing for is not rated; the brief says so instead of guessing.
2. 2Review every name marked as due (not reviewed for six briefs, moved 8 % or more since the last review, through one of its own levels, or past a catalyst or results date) and either reiterate the rating with the reason it still holds, or upgrade, downgrade or update it with the reason it changed.
3. 3Leave the rest alone unless the news changes the view.

Every change is a new version. Nothing is overwritten: the symbol sheet lists each version newest first with what changed in plain words ("rating buy → hold · bull above 128 → 131") and the Researcher's one-line reason. A name it stops covering (sold and off the list, no edge until a new base forms) is suspended, kept for the record under No longer covered, and can be picked up again later.

Catalyst dates come from a tool, never from memory: a results date is the one the calendar shows, any other dated event names the feed item that stated it. A date the Researcher cannot back is dropped and the event kept undated.

How the PM uses it

The PM's session card carries a compact Analyst coverage section for the names it holds or might act on: rating, conviction, the thesis in one line, the next catalyst, what would change the view. It is labelled as the Researcher's, context and not an instruction. When the view bears on a decision the PM cites it among the inputs behind the proposal, next to the signal, the brief and your guidance. The PM is free to disagree: a proposal that goes against the rating (buying a name rated Avoid, closing out a high-conviction Buy) must say so in one line, and that disagreement is recorded so "the PM overrode the analyst" can be counted later. It is never a veto; the risk policy and your approval line are.

How it is measured

An analyst is judged on calls, not adjectives. The checkable calls a brief makes on a covered name ("holds 220 into results", "closes above 500 within ten sessions") are the same calls every brief makes: scored hit, miss or undetermined by code from the daily close at the horizon, no model grading. Each one is tagged with the rating and conviction it was made under, so Coverage can show two things:

• the record: hits over scored calls, for the book and per name;
• by conviction: whether a conviction-5 call hits more often than a conviction-2 call, which is the honest test of whether the confidence means anything.

Both are shown only once there are enough scored calls to mean something: twenty for the book's record, ten per name or per conviction bucket. Until then the slot reads not enough calls yet with the count, never a percentage from three calls. Expect that for the first few weeks.

What it is not

Coverage is the Researcher's opinion, written down and versioned. It does not place, size or block orders, it does not arm price levels, and no number on this page feeds back into the model as a score to chase. If a rating keeps being wrong, the record shows it; what to do about it is yours and the PM's call.