Most reference libraries have a few hundred images that actually earn a look and thousands that don't

Rating is the fastest way to tell those two groups apart, but only if the stars mean something consistent — which they rarely do on a first pass, because rating while you're still excited about a folder produces four stars across the board. The fix isn't a stricter scale. It's a rating pass done separately from collecting, against a specific question, followed by a query line that actually uses the result.

The reason this is worth a whole workflow rather than just "rate the good ones" is that a rating only functions as a filter if it means the same thing on the day you set it and the day you search by it. A rating set in a burst of enthusiasm the moment you save an image measures your mood that day, not the image's actual usefulness — and a library where half the ratings were set that way and half were set carefully during a calm afternoon two years later is a library where rating:>=4 returns a mix of two different standards. Getting the rating pass right the first time, or at least consistently, matters more than almost anything else you'll do to this library.

1. Rate in a second session, against one question

A rating pass works better as a dedicated fifteen minutes than as something squeezed between other tasks — treat it closer to a small editorial review than a chore you're rushing to clear. Close whatever tab or project you were in when you saved the images, and come at the folder fresh, the way you would if someone else had handed it to you and asked which ones were worth keeping.

Don't rate as you import. Come back once the folder has cooled off and ask one narrow question per pass — "would this survive being shown to a client," or "is this the shot I'd reach for if I needed this exact pose again." A single clear question produces a rating that means the same thing across a thousand images; five vague ones produce noise.

1 star   — kept out of habit, would delete on a cleanup pass
2 stars  — usable, nothing special
3 stars  — solid, would use if nothing better turns up
4 stars  — reach for this one specifically
5 stars  — the reference I'd point someone else to

2. Leave unrated files unrated

Resist the urge to give everything at least one star. An unrated file and a one-star file answer different questions — "I haven't judged this yet" versus "I judged this and it's weak" — and a library that collapses the two loses the ability to ask either. no:rating becomes a genuinely useful "what haven't I gotten to yet" filter, but only if it's honest.

3. Build the shortlist with a query, not a scroll

Once even a fraction of a folder is rated, the shortlist stops being something you assemble by eye and becomes something you ask for:

-is:favourite in that line matters more than it looks: it keeps the query from re-showing you images you've already promoted to a board, so a rating pass and a favouriting pass don't fight over the same result.

4. Use favourites for "in the current project," not "good in general"

Rating and favouriting answer different questions and conflating them is the most common reason a filter stops being useful. Rating is a judgement about the image itself, and it should survive the project it was made for. Favouriting is "this is in the working set right now" — save it for what you're actively pulling toward a board, and clear it when the project ends. is:favourite then means something specific instead of accumulating every four-star image you've ever tagged.

5. Save the shortlist query instead of rebuilding it

A saved view stores the query text, not a fixed list of files, and re-parses it every time you open it. Save rating:>=4 tag:any:portrait,studio once and it stays current as you rate more images into it — no manual list to update, and renaming a tag doesn't orphan the view the way it would a fixed selection.

6. Filter by rating range, not just a floor

rating:>=4 is the everyday query, but the range operators are worth knowing for the opposite question too. rating:<=2 -no:rating finds the weak end of what you've actually judged — useful once a year, for a cleanup pass where you're deciding what to stop keeping around, without accidentally sweeping up everything you haven't gotten to yet.

6.5 Use a tag for the nuance a five-star scale can't hold

Five stars is coarse on purpose — a finer scale is harder to apply consistently, not easier, because the more gradations there are the more each one starts to blur into its neighbours. When you genuinely need a distinction the stars can't carry — "great pose, weak lighting" versus "great everything but the wrong mood for this project" — a tag is the better tool for it than trying to invent a half-star. rating:>=4 tag:wrong-mood narrows a five-star shortlist by a reason that a number alone could never express, and it keeps the rating scale itself simple enough to apply the same way six months from now as you did today.

6.75 Rate in context, not against the whole library at once

A rating is a judgement relative to something, and it's worth being deliberate about what that something is. Rating a folder of studio portraits against the best studio portraits you've ever seen produces a much harsher scale than rating it against "is this usable for the brief in front of me." Neither is wrong, but mixing the two standards within one folder — five stars for "genuinely exceptional" in one session and five stars for "good enough for this job" in another — is exactly the kind of inconsistency that makes rating:>=4 stop meaning anything specific. Picking one standard and writing it down somewhere, even just in your own head as a fixed rule, is worth more than getting the scale itself right on the first try.

7. Rate video the same way you rate stills

A clip takes a star rating exactly the way an image does, on the same query line, and there's no separate workflow to learn. If part of your reference is footage — motion studies, reference takes, film stills alongside the scenes they came from — rate the clips during the same cold second-look session rather than treating video as a category you'll get to eventually. type:video rating:>=4 then works exactly like the image version, and a shortlist that mixes stills and clips by rating doesn't need two separate passes to build.

type:video rating:>=4 -is:favourite     the clips worth a second look
type:image rating:>=4 -is:favourite     the stills, same standard
rating:>=4 -is:favourite                both, mixed in one result

8. Check your own drift, not just once but periodically

Rating standards drift over time even when you're being careful — a four-star image from two years ago and a four-star image from last week were judged by slightly different versions of your own taste, and that's normal, not a mistake. What's worth doing occasionally is a spot check: pull up rating:5 and scroll it end to end once every few months. If a chunk of it no longer feels like it belongs at the top, that's not a reason to panic and re-rate everything — it's useful information about how your standards have moved, and a nudge to be a little more critical on the next pass rather than a project to fix retroactively.

9. Use notes to explain a rating that isn't obvious

Occasionally a rating will look wrong to your future self — a three-star image sitting next to five-star company for no visible reason, or a four-star rating on something that looks unremarkable out of context. A short note explaining the judgement — "weak on its own, but exactly the pose I need for the brief" — saves you from either trusting a rating you no longer understand or second-guessing and re-rating something that was actually right the first time. rating:>=4 has:notes is a small but useful filter for exactly this: your best-rated images with an explanation attached.

10. Let the shortlist inform the next round of collecting, not just the current board

Once you've built a rating habit, the shortlist itself becomes useful information about what you're actually drawn to — pull up rating:5 across your whole library every so often and look at it as a set rather than individually. Patterns that weren't obvious image by image often are obvious once thirty of your five-star references are sitting in one grid together — a recurring colour palette, a type of composition you keep reaching for. That's not something a rating system was explicitly designed to surface, but it's a genuine side benefit of having one that's honest.

What to do once rating a whole library by hand stops being realistic

At a few thousand images, rating everything in one sitting is not a reasonable goal, and it doesn't need to be. Rate as you go — a fresh folder gets its cold-second-look pass the week after it's added, not years later — and let the query line do the work of surfacing what you've rated so far. rating:>=4 added:>2026-01-01 finds this year's best without needing last year's folders to be rated too. A library where every image has an opinion attached ages better than one where rating was a one-time project, because the habit keeps up with the folder rather than falling permanently behind it.

The failure mode worth actually planning against is not "I haven't rated everything" — that's fine, and permanent, and true of every large library that has ever existed. It's "I rated in a burst two years ago and haven't touched it since," because that's the version where the standard has drifted the furthest from whatever your taste is today without you noticing. A folder added last month, rated last month, against this month's standard, stays trustworthy. A folder added three years ago and rated in the same sitting it was imported carries whatever your judgement was worth on that one afternoon, frozen. Neither is wrong to have in the same library — old folders don't need to be re-rated on a schedule — but it's worth knowing which kind of rating you're looking at when rating:>=4 returns a mix of both, and treating an old five-star result with slightly more scrutiny than a fresh one if the two disagree with what your eye tells you today.

If ratings are the part of your workflow that's currently scattered across Finder tags or color labels, Finder tags versus Coii Ref covers what a dedicated rating and search step actually adds. For getting more out of the query line generally, using search operators for images goes past rating into the rest of the keys. And setting up saved views is the natural next step once you have a shortlist query worth keeping around permanently.

Photographers culling a shoot down to the keepers are usually doing exactly this rating workflow already, just without a query line to make it searchable afterward — see organizing for photographers for how that fits together. If you're coming from Adobe Bridge, where ratings already live in the files themselves, the comparison with Bridge covers what that means for a switch.

For the note habit mentioned above, writing notes on reference images goes into what makes a note worth the ten seconds it takes versus one you'll skim past without reading.

Rating a real folder against a real project is the only way to know if a five-tier scale actually fits how you work — the trial gives you thirty days and every feature to find out before paying for it.

Thirty days is enough to run the whole cycle once: collect for a week, rate cold the following week, build a shortlist query against the result, and see whether the query actually returns what you'd have picked by eye. If it does, the habit is worth keeping past the trial. If the shortlist feels wrong, that's useful information too — it usually means the rating question you picked in step one wasn't specific enough, and it's worth trying a narrower one before concluding the whole approach doesn't work for how you collect.