Star Ratings, Color Labels or Flags: One Culling System

Three tools, three different questions

The three tools are not equally expressive, and that decides what each should do. A flag has three states, a color label has five (plus none), and a rating has six (0 to 5). The more states a tool has, the more judgment calls it demands per frame, and judgment calls are what slow a cull down.

Notice that the strengths barely overlap. Flags answer "is this in or out?" Stars answer "how good is it?" Labels answer "what happens to it next?" Trouble starts when you ask one tool to answer two of those questions.

It helps to think of the tools as columns in a spreadsheet rather than as competing options. A frame has a decision (flag), a grade (stars) and a destination (label). Three separate columns, three separate questions. Problems appear only when you try to read two columns out of one, for example inferring the destination from the grade.

That is also why the common advice to simply use stars for everything tends to work for a while and then stop working. It works on a small shoot, where you can hold the meaning of each number in your head. It breaks on the fifth or tenth shoot of the month, when a 3 on Tuesday and a 3 on Friday no longer mean the same thing.

Flags: the cheapest decision is yes or no

Flags are the right tool for the first pass because a binary decision is the cheapest decision a person can make. You are not comparing a frame to anything; you are deciding whether it survives. In Lightroom Classic the keys are P for pick, X for reject and U to clear, and arrow-key advance means one keystroke per frame.

The limit is that a flag cannot rank. After a pass over 2,000 frames you might have 500 picks, and all 500 look identical to the software. That is fine, because the first pass was never meant to rank. It was meant to shrink the pile.

A second practical point: rejected is not deleted. Lightroom lets you review rejects before removing them from disk, so a heavy hand on X on the first pass is recoverable until you choose to purge.

A practical way to keep the first pass honest is to set a rule for what a reject is. Reasonable candidates are closed eyes with no better frame in the burst, missed focus on the subject, or an accidental shot. Anything that is merely "fine, but not special" is not a reject; leave it unflagged and let the second pass decide. Rejecting on taste during a fast pass is how you lose frames you would have wanted.

Burst sequences deserve a note of their own. When ten near-identical frames exist, pick the best one in the burst on the first pass and leave the rest unflagged. This keeps the picks list free of duplicates, which means the star pass starts from distinct moments rather than from ten versions of the same one.

Star ratings need a written definition

Stars are useful when you cull in stages, because each pass raises the bar. The trap is that a star count has no definition until you write one. Is 3 stars "usable" or "good"? Two months later you will not remember, and neither will a second shooter or an assistant.

Stars also compose well with filters. Filtering the grid to "3 stars or higher" is one click, which makes a staged cull easy to resume after a break.

Staged rating also means you can stop early. If a client only needs 40 frames, you may never need to separate 4-star from 5-star frames across the whole shoot; you stop the moment the 3-star-and-up filter returns about the right number. Stars let you cut as deep as the job requires, and no deeper.

If you shoot with a second shooter, ratings are the field most likely to be applied differently. Agree the definitions before the shoot, or have each person rate in a separate pass and compare, rather than discovering at the edit that one person rates generously and the other severely.

Color labels track state, not quality

Color labels shine when you stop thinking about quality and start thinking about state. Red for "needs retouching," green for "ready to deliver," blue for "in the client's selects." Because a label has no built-in ranking, it does not collide with stars: a frame can be 4 stars and red at the same time without contradiction.

The risk is the opposite of stars. Five colors invite five meanings, and the meanings multiply with every new project type. Pick a small set, usually two or three, and reserve the rest. If you work with other people, the same discipline applies to names; the same reasoning is behind a file naming convention that survives handoff, and the approach to labels is similar: a convention only helps if everyone reads it the same way.

Labels also pair naturally with a minimal status model. If you track whether each frame is approved, edited or delivered, a label is a convenient place to show it in the grid. See the four-state tracking model for how to keep that small.

Because labels are the most flexible tool, they are also the one worth locking down per project type. A wedding might use red for retouch and green for album; a commercial job might use red for retouch and blue for a specific deliverable. What matters is that within a given job type the mapping never changes, so a glance at the grid always tells the same story.

A culling system rule set you can adopt

Here is a rule set that gives each tool one job. It is one reasonable arrangement, not the only one, and the point is that it is written down.

To see why staging helps, run illustrative arithmetic on a 2,000-frame wedding. These are assumptions, not measurements; substitute your own timings. Say the flag pass takes 2 seconds a frame: 2,000 × 2 = 4,000 seconds, about 67 minutes. If you keep 25% (500 frames) and the star pass takes 4 seconds a frame: 500 × 4 = 2,000 seconds, about 33 minutes. Total: about 100 minutes, with the slower, more careful look spent on a quarter of the frames.

Compare a single pass where you grade everything with stars at 4 seconds a frame: 2,000 × 4 = 8,000 seconds, about 133 minutes. The staged approach is cheaper only if your assumed per-frame times hold, so time yourself on one shoot before trusting the shape of this math.

Once you have your selects, the next question is how much of the cull the client should see; showing clients every frame or only your selects covers that decision.

Finally, decide when each tool is allowed to change. Flags change only during the first pass. Stars change only during the second. Labels change only after the cull, when work is being assigned. Once a tool is closed for the shoot, resist the urge to reopen it; reopening is how a finished cull quietly becomes an unfinished one.

Keep the client's decisions in a separate place

Your flags, stars and labels are your decisions. A client's decisions are a separate set, and mixing them into the same fields is how a photographer's "3 stars" ends up read as the client's rating, or the other way round.

Keep the layers apart. Your catalog marks describe your editorial judgment. The client's picks, comments and markup should live in the place where they make them, tied to the frame, and never overwrite your own marks. There is no Lightroom plugin for Appruva, so today that means the client's decisions are in the review link, and you apply whatever you need back to your catalog yourself.