Turn Agency and Client Notes Into a Workable Treatment Revision

The note that ruins a revision is rarely the one you argued with. It's the one that had already been answered, and got obeyed anyway.

A notes round looks like a single instruction set. It isn't. A margin comment on the PDF you sent last week, two paragraphs in an email that never name a scene, a line you half remember from a call — each was written about a particular document at a particular moment, and each is asking for a different kind of thing.

So before you change a word, establish two facts for every note: which document and version it refers to, and what it's actually asking of you — a correction, a preference, a question, or a choice that is still open. Then work out who owns the decision when two notes can't both be true, and revise everything the decision touches rather than the sentence the comment happened to sit beside. A revision is a record of a commissioned decision, not a place to file everyone's words.

The note set I'll use runs through this article. The brand, dates, roles and treatment are invented; it's a construction, not a report of any real exchange.

Put each note back beside its source version

Treatments circulate. Drafts get renamed. A brief gets revised while you're shooting a location. By the time feedback lands, four documents may be in play, each with its own author and its own revision owner, and the newest message in your inbox may be about the oldest file.

Here is the invented file:

• Brief, rev 2 (dated 14 April). Governs the job. Names the hero product — the Carillon Duo — and puts story and performance decisions with the agency's creative lead, and product presentation and claims with the client's brand manager. Your brief will say something else, or nothing at all; if it says nothing, that silence is worth resolving before the next round rather than after.
• Draft A (8 April). The rider rings the bell and a passerby flinches and drops a paper bag.
• Draft B (15 April). The creative lead records a decision on it: the bell never plays at someone else's expense. The flinch goes. The bell lands on the rider.
• Draft C (6 May). The current treatment. No passerby. The rider treats the bell as a small ceremony; the note punctures his own self-importance; he twitches wryly at himself, we see the product, and he rides off.

And here is the notes round, with each note put back where it came from:

Two things are visible in that table and nearly invisible in an inbox. First, the note dated 12 May is about a beat that was cut on 15 April, and it asks for a bigger version of a reaction that no longer exists. Second, the most decisive instruction on the page — the one that governs the current cut — is the oldest entry, and it isn't styled as an instruction at all. It's a line in a treatment file.

Recency is a filing fact, not an authority. If the newest timestamp automatically won, draft B's decision would have quietly died in April and nobody would have noticed until the shoot. Read the passage a comment was written against, not the passage with the same number in the current version. If the comment is a shapes on a PDF, look at what it's near. If it's an email paragraph that names no scene, find the scene by matching its wording against the draft it quotes — usually a phrase like "the bell lands soft" exists in exactly one place in one version.

Turn vague feedback into a film question

"Warmer." "Bigger." "More human." "It feels flat." These aren't instructions; they're a reviewer pointing at a feeling and handing you the search.

Warmth in the invented spot could mean at least four different films:

1. The rider's own reaction opens up. He doesn't twitch; he grins at himself. Same scene, same cast, a different performance.
2. Someone else hears it. The bell stops being private. A stranger looks up and smiles, and the ring becomes something shared.
3. The light goes late. Same action, sun instead of overcast, a different grade and a different shoot window.
4. The relationship changes. He's ringing to someone specific, which rewrites why the bell matters.

Each of those is cheap, and each pulls the same beat in a different direction. They overlap at the edges — (b) changes his face too, because a bell someone else hears lands on him differently — but only one of them governs the beat. Applying all four is how a treatment ends up with a comment in the margin for every reviewer and no governing idea. Applying one silently is how you spend a week building the wrong film and get a second round you didn't need.

So name your reading and ask. What I'd actually send:

On beat 2: I read "warmer" as four things and I won't choose for you. (a) his own reaction opens up; (b) someone else hears the bell and smiles — the ring becomes shared; (c) the light goes late; (d) the relationship changes and he's ringing to someone. My reading is (b), with the joke still on him and nobody startled. If that's it, I'll revise the beat, its reference line and its timing. If you meant (a), it's a performance note and a smaller change — no second face, nothing downstream of beat 2 moves.

The answer back, in this construction, is (b): someone else gets to hear it, nobody jumps, the joke stays on the rider.

Notice what the question bought you. It confirmed a change and it also checked the change against draft B's decision, because adding a second face is exactly the kind of edit that looks like reopening a closed argument. It isn't. Draft B closed a reaction at someone's expense — a flinch, a dropped bag, a laugh earned off a bystander. It didn't close the existence of other people. Reversing that distinction is a small thing in a treatment and a large thing on set, where a performance pitched at a stranger's embarrassment and a performance the stranger enjoys are two different afternoons.

Keep questions and instructions apart while you're doing this. "Does the bell ring twice?" is a question. Answer it. "Cut the second ring" is an instruction. And a question can carry a request inside it: "Is the bell supposed to be this loud?" may mean turn it down. If the answer is that the note is deliberate, say so plainly, rather than quietly softening the sound to end the discussion — which gives you a mix nobody asked for and a note that will come back.

Separate decisions that can be implemented from those that conflict

Sort the round into buckets before you edit:

• Required corrections. The treatment contradicts the brief or a locked decision. These have one answer.
• Settled choices. Already decided, in the file, by someone entitled to decide.
• Preferences. Worth hearing, no obligation attached.
• Questions. They need an answer, not a revision.
• Live alternatives. Two outcomes are genuinely still in play on the same beat. You cannot shoot both.

Draft C's product-model error sits in the first bucket. The hero is the Duo; the treatment names the Solo and the end card carries the old name. No version of warmth touches the model, so the producer's correction can be applied today, before anything else in this round is resolved. That's the useful test for parallel work: does the change depend on an open question? Renaming a product doesn't. Rewriting beat 2 does, because it depends on what "warmer" turns out to mean, and if you guess wrong you'll have rewritten it twice.

The incompatible pair is the account director's note and the April decision. Read against draft C, "bigger reaction" has no target — there's no flinch to enlarge. Read against draft A, where it was written, it asks for the beat that was deliberately cut three weeks earlier.

Which means the question isn't "who is right," it's "who resolves this particular thing." Here's where people reach for the wrong signal. The account director's note is the most emphatic thing in the file: it's short, it's blunt, and its question mark won't be read as one by anybody. The client's note carries the weight of the client. Neither of those facts makes either sender the owner of this beat, because ownership is a property of the assignment, not of the sender's volume or title. In this invented production the brief puts story and performance with the agency's creative lead, so that's who confirms what warmth means. In yours it might be the client, or the director, or a person nobody has named yet — which is itself the finding, and worth one short email before you build.

So you ask the creative lead two things at once. What did "warmer" mean — and does the May note about a bigger reaction change the April decision? You get an answer to the first. You put the second in the return note as open, with a direct question, because a guess dressed up as a revision is worse than a question on the record.

Trace the chosen revision through the film

Once warmth means a shared smile, the change runs through beat 2 and stops where it stops. Following it properly:

• Setup. The woman at the kerb needs to be in the frame before the thumb presses the lever, or her smile reads as coincidence rather than answer. One added line to beat 1.
• Performance. The rider's reaction changes too, because it is now answering someone: "A wry twitch" becomes a grin he can't hold back, then her look up, her smile, his embarrassed nod. The rider is delighted with himself, not with having startled anyone.
• Reference. Draft C's reference line for beat 2 says Reaction: dry amusement only. No sentiment. Left alone, that caption keeps instructing the old performance to whoever boards the shot. It gets rewritten, and the reference frame should show both faces in one composition.
• Edit. The cut that came straight off the twitch now comes after her smile, so the beat is longer. The time comes back out of beat 1, which has two glove adjustments where one will do. That's a change to beat 1 — but it's one this decision caused, which is different from reopening beat 1 because you happened to be in it.
• Sound. The note stays one small, bright thing. If the mix swells it into an event, a second face looking up reads as a shock response, which is the reading April closed. Worth writing down, because "make it warmer" applied at the mix can undo a story decision nobody meant to touch.
• Ending. The push-off stays on the rider alone. She isn't in the final frame and the end card doesn't change; the spot is about him.

That last point is a kind of editing that gets skipped. A revision applies one confirmed intention consistently while leaving the surrounding work alone: the ceremony, the product arc, the tagline, the costume, the street. Reopening those costs a round and buys nothing.

And delete the obsolete material rather than leaving it in the margin. Draft C still carries an alternative under beat 2 — ALT (from draft A): passerby jumps, coffee, bell as a fright. Any reader picking up the next draft sees a live option that was closed in April, and you're back in the same conversation.

Return a readable account of what changed

The return note is not a defense of the draft and not a summary of the film. It's the record that lets the next reader see the decisions without re-reading the whole round.

Carillon treatment — return note, draft D, 22 May

Applied from brief rev 2: hero product corrected to the Carillon Duo (product section and end-card super). Independent of this round; nothing else moves with it.

Decision taken: beat 2 now shares the bell. A woman waiting at the kerb hears the note and smiles; nobody is startled; the joke stays on the rider. Confirmed by the creative lead on 21 May — recorded here because it was confirmed on a call.

Passages changed: beat 1 (she registers before the ring; one glove adjustment cut), beat 2 (rewritten), beat 2 reference line, beat 2 timing, one sound note.

Deleted: the draft-A alternate in the margin of beat 2, so it doesn't get reopened.

Open: the 12 May note on a bigger reaction was written against draft A, where the flinch existed. That beat was closed on 15 April. If it meant the rider's reaction rather than a passerby's, tell me — the grin that came in with the shared bell may already cover it, but I won't assume.

Short enough to read before opening the treatment, specific enough that someone who disagrees with you can disagree with the actual decision rather than the wording. That last entry is the point of the whole exercise. It stays visible, in plain language, instead of being smoothed into "we've softened the beat slightly to add warmth," which sounds agreeable and can be read four ways by four people.

A revision is finished when the decisions are traceable, the affected passages have moved together, and the things still undecided are named as undecided. Not when every comment has a sentence somewhere that appears to satisfy it.

Frequently asked questions

What two facts should be established for every note before revising a treatment?

Which document and version the note refers to, and what it is actually asking for: a correction, a preference, a question, or a choice that is still open. You also need to know who owns the decision when two notes cannot both be true.

Why does the newest note not automatically win?

Recency is a filing fact, not an authority. A note may have been written against an older draft, and the most decisive instruction in the round may be the oldest line in the file. Read the passage the comment was written against and match its wording to the draft it quotes.

How should vague feedback such as 'warmer' be handled?

Turn it into a film question by naming the possible readings rather than applying all of them or silently choosing one. In the invented Carillon example, 'warmer' could mean the rider's own reaction opens up, someone else hears the bell, the light goes late, or the relationship changes. State your reading, ask which is meant, and check that the answer does not reopen a decision that was actually about a different thing.

What categories help sort a notes round before editing?

Required corrections, settled choices, preferences, questions, and live alternatives. Apply independent corrections first, such as fixing the product model, and wait on changes that depend on an open question, such as rewriting beat 2 before 'warmer' is defined.

What should a return note contain, and when is a revision finished?

It should record corrections applied independently, the decision taken and who confirmed it, the passages changed, obsolete material deleted, and open items named in plain language. A revision is finished when the decisions are traceable, affected passages have moved together, and anything still undecided is named as undecided.

Further reading

• Bring PDF Review Comments Back Into the InDesign Treatment — Use when margin comments must land in the right document and version: it covers anchoring, authority and giving every imported note a disposition.
• Use Feedback From a Lost Commercial Bid Without Inventing Why You Lost — Useful when a note names no scene: it shows how to place a comment beside its passage and keep what was said apart from what you concluded.