A content system for an aesthetic cosmetology studio. It starts with an uncomfortable finding: Reels in beauty deliver the worst engagement of any industry, and the gap between "medicine" and "beauty" positioning is almost sevenfold on the same platform.
Five findings. high — several sources, medium — one, estimate — the author's calculation.
Healthcare gets 2.22% engagement at two posts a week; beauty and personal care gets 0.32%. That's almost a sevenfold difference on one platform, and it's created not by the format but by the category the audience perceives the account as being in.
0.2% engagement, the worst result of any industry. Carousels beat Reels by 109%, and beauty and personal care is explicitly named as the exception where photo formats lead.
Advertising prescription drugs on websites is prohibited. Claims of a guaranteed effect, images of bodily change, and statements that instil a fear of falling ill are prohibited. The list of drugs banned from advertising was expanded to 404 items by a Health Ministry order of 24 April 2026.
Rejecting "before and after," result promises and fear-mongering is at once a legal requirement and the content of the brand platform. Compliance costs nothing here, because we weren't going to do it anyway.
Calculated from industry production norms: about 24 hours a month for the whole flow of two posts and two story chains a week. That's half a working day once a week, not a separate job.
A section whose absence was a direct defect of the matrix: the copy was written about injectable procedures with not a single check of the legal restrictions.
Naming prescription drugs. Their advertising on websites is expressly banned, and botulinum toxin is prescription-only
Talking about the procedure and the consultation without the drug name: "injectable correction of expression lines"
Claiming that the therapeutic effect is guaranteed
Naming observed timeframes and probabilities: "lasts 6–8 months," "bruising in about one case in five"
Publishing images of change in the human body due to illness
Showing the room, the physician's hands at work, documents and quotes
Statements that cause a fear of falling ill or being harmed by not using a product
Naming the audience's existing fears and removing them — the opposite action
What had to be fixed in the matrix itself. The word "botulinum therapy" from the brand glossary is acceptable in conversation and on a price list, but in an advertising post the drug name or the procedure by its brand name is already a violation. In the copy below it is replaced with a description of the action.
On triggers specifically. The law prohibits statements that instil fear. The matrix is built on the audience's fears — and that is fundamentally different: we don't create fear to sell a solution; we name an existing one to remove it. The difference is checked with one question: will the person feel more anxious after reading it. If so, the text is rewritten.
And the main thing. All four prohibitions coincide with what the brand platform already forbids itself: no "before and after," no result promises, no fear-mongering. Complying with the law costs nothing here, because we weren't going to do it. For competitors whose marketing is built on dramatic shots and promises, it's a different story.
The first version of this matrix was built on the general advice that "Reels carry reach." The 2026 industry data doesn't confirm that advice for our category.
Healthcare shows 2.22% engagement at two posts a week. Beauty and personal care shows 0.32%, the worst result of any industry. Reels in beauty drop to 0.2%.
Almost a sevenfold difference on the same platform — and it's created not by format or budget but by the category the audience perceives the account as being in. The brand platform's bet on a medical rather than a decorative frame turns out to be a decision about reach, not about taste.
The second consequence — the format mix. Carousels deliver 109% more engagement than Reels, and beauty and personal care is explicitly named as the exception where photo formats still lead, with Reels behind. So building the whole system on Reels means going into the most crowded format with the lowest return.
What follows for the distribution. Reels stay, but as a first-contact channel, not the load-bearing structure. The main weight moves to carousels and text: that's where the mechanic is explained, the price is named and the fears are worked through — exactly the things that need several screens and fit badly into twenty seconds of video.
Third — frequency. The healthcare benchmark is obtained at two posts a week, not daily. This matches the industry observation that clinics with three or four dense pieces beat those who post filler every day.
A week of warm-up is the launch. After that you need a system that produces content indefinitely and doesn't require inventing a topic from scratch. Six series, each tied to a brand pillar.
The system's rule: if a piece doesn't fit any series, it isn't published. Not because it's forbidden, but because a piece outside the series is almost always filler — the very thing that, per the data, loses to a rarer but denser post.
Seven days of the first warm-up. The logic of the sequence: first recognition of the problem, then proof, then removing objections, and only at the end a call to action. Asking for an action before the fear is removed is pointless.
| Day | Trigger | What we close | Main format | Support |
|---|---|---|---|---|
| 01 | Self-identification | Fear of visibility — 85% of the audience | Telegram text | Reels hook |
| 02 | Social proof | Distrust of the consultation | Carousel | Story chain |
| 03 | Transparency | Fear of upselling and a hidden price | Carousel | Reels |
| 04 | Authority | Not knowing who to trust | Text + a photo of a document | Stories |
| 05 | Loss aversion | Fear of irreversibility — 27% | Telegram text | Reels |
| 06 | Honest scarcity | Doubt about the quality of the flow | Story chain | Text |
| 07 | Commitment | The absence of a next step | Carousel | Reels |
The main carrier's format is chosen by the benchmarks: carousels where the mechanic needs explaining over several screens, text where tone is needed, Reels only as an entry point for people not yet following.
The copy is written in the tone of voice from the brand platform: numbers and timeframes, risks out loud, not one exclamation mark or diminutive.
The first version had none at all, even though two platforms are declared. Benchmarks: chain completion 70–87%, sticker interaction 12–18% of viewers. The chain is built around those two numbers, not around "having some."
The poll comes fourth, not first: a sticker at the start collects a tap but cuts the chain off. Ahead of it are three frames that have already delivered value — then the interaction doesn't come at the cost of completion.
The last frame is a question with no answer options. It collects fewer taps than a poll but gives material for the "A question from the DMs" series and the audience's real phrasing.
One hook per day isn't a test, it's a bet. Three variants with different hypotheses and a clear risk for each.
How to test at small numbers. A studio with six thousand followers can't reach statistical significance in a week. So the comparison runs not on conversion but on completions and saves — metrics that accumulate volume in days, not months. The winning hook moves into paid promotion, where significance appears faster.
Two posts a week per the healthcare benchmark, not a daily flow. One piece lives in four forms.
| Day of the week | What goes out | Platform |
|---|---|---|
| Monday | A question from the DMs — a story with an answer | |
| Tuesday | What we talked people out of — a carousel, with a text repost in the evening | Instagram + Telegram |
| Thursday | Working through a fear, or What we don't do | Telegram, Reels after |
| Friday | Behind the scenes — a story chain | |
| Once a month | The price out loud — a breakdown of a programme's cost | Both platforms |
The reuse chain for one piece. "Working through a fear" is written as a Telegram text — that's the source. A carousel is assembled from it: one point per screen. A Reels hook is cut from the first screen of the carousel. The topic of the next "A question from the DMs" is born from the questions under the post. In total, one hour of work yields four posts on two platforms.
Why the source is text. Text is edited in a minute; video is reshot in an hour. It's cheaper to catch a wording error before the shoot — and on a subject where every promise is checkable, wording gets edited often.
The warm-up leads into a conversation, and there were no conversation scripts. Five real questions with answers in the brand's tone — with an explanation of why exactly this way.
In the first version the prompts stood with no outputs — that is, as hypotheses. Below each is shown together with what it actually produced on an input from this matrix.
The key here is the ban on marketing language and the mandatory number requirement. Without these constraints the model produces advertising copy by default.
You are a copywriter for an aesthetic medicine clinic whose entire
positioning is built on talking patients OUT of unnecessary procedures.
CONTEXT
- Audience: women 27-48 considering injectables, mostly first-timers.
- Research: 85% require results to be non-obvious; 27% fear a "slippery
slope"; qualifications rank #1 as a trust factor, before/after photos
rank only 5th of 7.
- Brand voice: calm, direct, professional but not clinical, warm.
TASK
Write one Telegram post (120-180 words) that closes this specific barrier:
{BARRIER}
HARD CONSTRAINTS
- Include at least one concrete number or timeframe.
- State something the clinic will NOT do, or who this is not for.
- No exclamation marks. No emoji. No diminutives.
- Never use: "transformation", "beauty shots", "limited spots",
"individual approach", "affordable prices".
- Do not ask for a booking unless the barrier is "no next step".
ACCEPTANCE TEST
If the clinic logo were removed, a reader must still be unable to
confuse this with any competitor's post. If they could, rewrite.
What it produced for BARRIER = "fear of judgement from those around you"
A separate prompt rather than "rewrite it shorter": video has a different retention mechanic, and the hook has to sound like an admission against one's own interest.
Convert the Telegram post below into a 25-35 second Reels script.
STRUCTURE
- Hook (0-3s): one sentence. It must sound like an admission against
the clinic's own commercial interest. No greeting, no intro.
- Body: 3 beats, each with a timecode and what appears on screen.
- Close: one line. Soft CTA only.
RULES
- Spoken register: how a doctor actually talks, not how ads talk.
- Any number stated aloud must also appear on screen.
- No before/after imagery anywhere in the script.
- Never open with "Hi guys" or the brand name.
POST: {POST}
What it produced on the post above
The controller prompt. Run last on any text, including one written by a person or a contractor.
Act as a brand editor. Audit the text against this checklist and
return a corrected version plus a short list of what you changed.
CHECKLIST
1. Does it contain a number or a timeframe? If not, it is empty - fix.
2. Does it say who this is NOT for? If not, add it.
3. Remove every exclamation mark and diminutive.
4. Is the price named, or hidden behind "message us"? Name it.
5. Could a real doctor say this out loud without wincing? If not, rewrite.
6. With the logo removed, is it distinguishable from a competitor?
Output format:
CORRECTED TEXT
---
CHANGES: bullet list, max 6 items, each one line.
TEXT: {TEXT}
What it produced on the text "We'll select personalised care at affordable prices! Book a free consultation"
A week of warm-up is the launch. After that you need a horizon on which the series build a cumulative effect rather than a stream of separate posts.
Why a quarter, not a month. The "Refusals" series builds strength through quantity: one post reads as a gesture, twelve in a row with dates read as a system. Cumulative formats can't be assessed before enough has accumulated.
What carries into the second quarter. Everything that showed growth in saves and completions. What didn't show it is replaced, not "refined": a format that hasn't worked over a quarter at two posts a week has enough statistics.
Calculated from industry content-production norms. Without it the plan stays a wish.
| What | When | Time |
|---|---|---|
| A story reply to a DM question | Monday | 20 minutes |
| The "Refusals" carousel plus a text repost | Tuesday | 1 hour 20 minutes |
| The "Working through a fear" text and a Reels from it | Thursday | 2 hours 30 minutes |
| The "Behind the scenes" story chain | Friday | 1 hour |
| Pulling the metrics and a short review | Friday | 30 minutes |
In total about 5.5 hours a week, i.e. 24 hours a month. At the Kyiv average wage that's roughly ₴7,000 a month in the cost of working time — half a working day once a week, not a separate job.
Where the norms come from. Industry measurements give an average of 4 hours 10 minutes to write a long article, while 40% of writers spend less than an hour on a short piece, and social-media graphics take about one and a half hours per unit. Our posts are short and the templates are reused — hence the figures below the averages.
On model acceleration. Industry data shows a 75–85% reduction in production time when working with a language model. The calculation above already accounts for this: without the prompts from section 11, the same volumes would take about 15 hours a week, and the system would be unworkable with a three-person studio.
What to outsource. The industry most often outsources video editing — 43% of cases — then copywriting at 37% and editing at 31%. For this studio the logic is reversed: write the copy yourself, because the positioning lives in the wording, and outsource the editing, because it carries no meaning.
A diagnostic tree instead of the advice to "keep trying." Every branch leads to a specific action, not a platitude.
The first version had one figure for the whole table. Below are benchmarks from 2026 industry data and from the logic of the system itself.
| Metric | Target | Where the benchmark is from and what a deviation means |
|---|---|---|
| Post engagement | from 2.0% | Healthcare gets 2.22% at two posts a week. A fall toward 0.3% means the account is being read as beauty, not as medicine — i.e. the positioning isn't landing |
| Story-chain completion | 70–87% | A cross-industry corridor. Below 70% — the chain is longer than attention holds, or the sticker is too early |
| Sticker interaction | 12–18% of viewers | An industry benchmark. Higher is good; below 10% — the question is phrased so that there's nothing to answer |
| Saves | growth month on month | There's no absolute benchmark; the trend matters. The "Price out loud" and "Working through a fear" series should collect more than the rest — if not, the copy isn't practical enough |
| Reels completions | from 30% | Below that — rewrite the first three seconds, the rest doesn't matter. Reels here is an entry point, not the load-bearing structure |
| Share of "how much" questions | falling month on month | If it doesn't fall after the "Price out loud" series — the price list isn't published prominently enough |
| Unfollows after day 06 | no higher than the usual baseline | A spike means the honest scarcity came across as manipulation and the wording needs softening |
How to read the launch-week result. If enquiries came mostly on day 07, the warm-up worked as intended. If they're spread evenly — individual posts work but the sequence doesn't, and the matrix needs reassembling. If there are no enquiries but saves are high — the problem isn't the content but the absence of a clear next step.
A caveat on sample size. At six thousand followers, weekly metric swings are noise. Decisions are made on the monthly trend, and on a weekly basis you can only look at completions and saves: they accumulate volume fastest.