A direction is one of two to four plain-language options Rankfield proposes before it runs the expensive part of a request — a coarse choice of where to point the research, made cheaply, before the deeper pass commits budget to it.
Choosing a direction is not choosing a scenario: it narrows what gets researched, and the ranked list of scenarios is what that research still has to produce.
Committing a run's budget to the wrong direction is the expensive mistake this two-step shape exists to avoid — the cheap pass is what a person actually gets to choose from.
Choosing between directions is meant to be a fast decision, not a research project in itself — the deeper pass is where the actual evidence gets gathered, and the direction only decides which part of the account and its market that pass looks at closely.
A common misreading is treating a direction as a finished recommendation rather than a coarse choice of where to look — it is closer to picking a chapter to read than to reading it. Two directions can both look reasonable from the outside, and the cheap pass is not meant to settle which is actually right, only to avoid spending the expensive pass on the wrong one.
Choosing badly at this step is cheap to fix, since almost nothing has been spent yet; choosing badly one step later, after the deeper pass has already researched the wrong angle, is the expensive version of the same mistake.
Where it comes from
This is Rankfield's own vocabulary and it has no exact match in the public copy yet — the closest live description is the loop itself, the same research-then-recommend shape the landing page walks through, step by step.
Read the source: Rankfield — the loop, in its own words.
What Rankfield reads
The landing page answers this directly, in its own words:
Your account end to end, your competitors, and the wider market around you: posts, videos, captions, comments, rhythm, what you offer. It watches the videos, not just the numbers under them.