A Craft Rubric That Does Not Kill Your Taste

One piece on an agency’s reel stops you, and you cannot say why.

Two ways to be wrong about the same post

Picture a hypothetical feed post from no real brand: crop deliberately off-centre, type uneven, caption a sideways joke that ignores tidy brand copy. It reads as one confident voice.

One person says it is great and cannot say why. That is pure gut, unrepeatable: you cannot explain it to a colleague, and it drifts with your mood. Another runs a checklist and marks it down for loose palette and no call to action. Checklists reward what is easy to tick, so competent but lifeless work outscores risky but alive work.

Thesis: a good rubric is a set of questions that direct attention, not a formula that outputs a verdict.

Why the tidy checklist misfires, and how to see it happen

A hypothetical illustration, not a case. Piece A is clean, on-palette, correctly captioned, with a clear ask, and forgettable. Piece B breaks the grid, runs long on a joke and uses one off-brand colour once, and you remember it a week later. A sheet favours A and has no box for being remembered. Checklist items measure the absence of error, while craft is the presence of a decision, which a sheet cannot see unless it anticipated it.

Once a checklist becomes the target, work is made to satisfy it. That is Goodhart’s law, often put as: when a measure becomes a target, it ceases to be a good measure. It is a widely cited adage, not a finding about agencies. See also: work can perform well and still be bad work.

The tell for a heavy checklist: if someone with no taste could use it and reach roughly your verdict on ordinary work, it will fail you on extraordinary work.

What gut is actually good for, and where it lets you down

A fast reaction is real information: compressed experience that catches something off or alive before you can say why. Alone, it fails three ways. It is swayed by who made the work, the halo effect, where a positive impression in one area colours others. It is swayed by the last thing you saw. Nobody else can check it. Nor is it steady: judgments can vary a great deal between people, and for one person from one occasion to the next.

So a rubric does not override the first reaction. It translates it into reasons afterwards.

Five questions that structure attention without dictating the verdict

  1. Intent. What is this trying to do, and can I say it in one sentence? Strong: you can, and it never wanders. Weak: you can only list contents.
  2. Choices. Which decisions could easily have gone another way, and were they made on purpose? Strong: one unusual cut the whole edit is built around. Weak: a stock transition, a default font pairing, a caption that could sit under any product.
  3. Fit. Does the execution serve the intent, or is it decoration over a different idea? Strong: humour, pacing and framing pull one way. Weak: a serious idea buried under trending effects.
  4. Risk. What could have gone wrong, and did the work take that risk or dodge it? Strong: it leans into something that might flop. Weak: every element is the safest option.
  5. Residue. What is left after I scroll away, and would I notice if a specific element were removed? Strong: a line or image stays and its loss would hurt. Weak: nothing remains.

These produce notes, not numbers. Two questions can pull opposite ways, and that is allowed. The verdict is a sentence you write, not a total you compute.

Why five? Few enough to hold in your head while scrolling, many enough to stop one impression dominating. It is a working choice, not a research finding.

Testing the questions on work that breaks the rules well

Pick three to five pieces you admire, at least two breaking a conventional rule: off-brand, overlong, ugly on purpose, no call to action. Answer the five questions for each. If a question marks down something you love, the question is wrong, not the work.

A hypothetical example: an anti-polished single-take video, badly lit, with a hard-to-hear opening line. Boxes for production quality and clear audio fail it at once. Intent survives: it wants to feel like a friend talking, not an ad. Choices survives, since the rough light and single take were kept when a cleaner version was possible. Fit holds. Risk is high. Residue is strong.

Now the reverse: a piece that ticks every conventional box and is forgettable. Residue catches it, and Risk shows every choice was the safe one.

Revise by dropping a question that only ever fires on polish, and rewording one that was a rule in disguise: “is the branding consistent?” becomes “did the piece commit to a voice?”

Using it without letting it turn into a form

React first and write the reaction in a line. Ask the five questions, then compare. When they disagree, investigate the gap instead of picking one. Record one sentence per question and a one-sentence verdict, so next month you can see why you thought it.

Warning signs of hardening: you score out of ten, add a sixth and seventh question every time you are surprised, or cannot like anything the questions did not predict.

For an agency’s portfolio and reel, ask the same five questions of every piece and watch closest where they disagree. Read across the whole body, since one excellent campaign tells you almost nothing. For groundwork, see what separates good social media work from average.

The honest limit: a rubric makes your reasoning visible and repeatable. It does not make it correct. It improves how you argue, and still depends on how much good work you have seen.

See the work for yourself

See this year’s selections, or nominate an agency.

FAQ

Why not just score work out of ten across a set of criteria?

Scored criteria reward what is easy to tick and cannot see a decision they did not anticipate, so they favour competent over risky.

How do I know if my rubric is any good?

Run it on admired work that breaks conventional rules. If it marks that work down, revise the question. Then run it on polished work you cannot remember and check it catches the emptiness.

Sources

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *