01Why unit economics is needed at all
“Unit” means one client. Unit economics answers one question: does EACH INDIVIDUAL client bring profit to the company, not just the business overall. This is a fundamentally different perspective from “how much did we make this month” — you can grow revenue and still go bankrupt if each new client costs more than they bring in.
So a marketer thinks not “how much was spent on ads” but “how much did one client cost and how much will they bring in”. Everything else in this document is just a way to neatly calculate those two numbers and compare them.
02The funnel: why 4 steps, not a single percentage
The junior approach is to take one number — “conversion from ads to sale”. The problem: that mixes four different processes, each with its own cause and fix:
- Ad click — depends on creative and targeting.
- Click → direct message — depends on how clear the offer is.
- Direct message → booking — depends on HOW and WHEN the person responds (the most powerful lever; see §9).
- Booking → showed up and paid — depends on operations: reminders, ease of rescheduling.
If you measure it as one number you can’t tell WHAT to fix when conversion drops. The breakdown is a diagnostic tool, not complexity for its own sake.
03CPA — not simply “spent ÷ bought”
CPA/CAC is the result of multiplying conversion rates along the chain. In medical cosmetology the funnel is shorter: lead (left contact) → booking → showed up for first visit. Paid budget ₴60,000, CPL ₴620 → ≈97 leads. Booking rate 38% (18% in the “slow” scenario), show rate 84% (no-show ~16%). CAC ≈ ₴1,800. CAC is the budget divided by the number who actually showed up.
CPA is not “the truth about the world” but a function of assumptions: multiplying percentages gives a nonlinear effect, so a small change at the input (response speed) yields a large change at the output.
04LTV — why “lifetime” isn’t literally lifetime
LTV (Lifetime Value) sounds like “how much a client brings over their whole life”, but in practice it’s a statistical expectation based on two things: how often a client visits and how long they remain active before they churn.
A client definitely generates revenue on the first visit. After that, it’s not guaranteed they’ll return: about ~42% make it to a second visit within 90 days (first-visit retention 30–45%, best-in-class 60–75% — Meevo, Phorest; the 50–60% 30-day retention in cms.beauty is a different metric and the two cannot be averaged). Only this retained portion “accumulates” additional value — via frequency. And here is a correction I made after auditing myself: the “8–12 visits/year” from cms.beauty describes a salon (nails, brows, hair), not medical cosmetology, where botox lasts 4–6 months. So the regular client is split into two real types: injectable (45%, ~5 visits/year, ticket ₴7,800) and skincare (55%, ~8 visits/year, ticket ₴2,800) → weighted 6.65 visits/year at a ₴4,490 ticket — and active lifetime (repeat ~75% per visit → ~70% annual retention (business-broker: “70%+ is the norm”) → effective lifetime ~2.8 years).
share of “regulars” × (visits/year × the regular’s own ticket × gross contribution margin × years of activity) + share of “non-regulars” × (a couple of visits × ticket × margin). A weighted average acknowledging that clients differ. By default ≈ ₴17,742 — LTV by gross profit. Annual revenue of an active client ≈ ₴29.9k — in the real corridor “₴30–100k/year” (business-broker); a previous model version gave ₴26k, below the floor.
05Contribution margin vs revenue vs profit
Almost every newcomer gets confused here — let’s go slowly:
| Revenue | the whole client ticket — weighted average ₴4,228 (from the table on the model’s main page) |
| Contribution margin | after direct costs: products (30–50% of the ticket for injections), doctor bonus from (ticket − products). ~60% of the ticket — the metric for the “should we accept this visit” decision |
| Gross margin for LTV | contribution margin MINUS the allocated share of rent and room salaries for that visit. ~44% — this is what we use in LTV, because the client “consumes” fixed costs too |
| Studio net profit | after ALL costs (rent, self-employed / staff, taxes, leasing, amortisation) — ~7–8% of revenue (the lower bound of a noisy corridor 10–20%: some equipment on instalment, premium-area rent in USD). This is P&L, not unit economics |
For the decision “take this visit” — contribution margin (~62%). For LTV — gross margin (~45%), because a repeat client occupies the room and staff the same way any client does. Calculating LTV by revenue nearly doubles it.
06LTV:CAC — the 3x rule and why it’s not a law of physics
The “healthy ratio — from 3x” rule comes from venture/SaaS: investors want at least a 3× buffer to cover costs not included in CAC — rent, salaries, support, taxes. It’s an empirical rule, not a mathematical law.
For LUMÉ the ratio is ≈ 9.8x for the paid channel and ~21x for the mixed channel — well above the threshold. The junior conclusion is “increase the budget”. The correct conclusion is the opposite: the constraint is not acquisition (word-of-mouth in cosmetology is cheap, organic is strong), but in room utilisation (in the chain model — 76%, the network is near capacity) and in retention. So in the tornado analysis on the main page budget and CPL are low on the lever list. A high LTV:CAC doesn’t mean “all is well”; it means “invest in capacity and retention, not in targeting”.
07Payback period — the second dimension people forget
LTV:CAC is about “is it profitable in the end”. Payback is about “will we have enough cash to live until that end”. A campaign can be profitable by LTV:CAC (5x) but pay back in 20 months — if the business doesn’t have 20 months of free cash it will go bankrupt BEFORE realising the profit, even if on paper everything looks great.
08ROMI, ROI, ROAS — three terms everyone mixes up
| ROI | the general term “return on any investment”, not only marketing |
| ROMI | the same, but specifically about marketing investments |
| ROAS | calculated on REVENUE, not profit — the simplest and most misleading metric |
The model uses ROMI on contribution margin — the honest version that shows real profitability, not just turnover.
09The model’s most valuable insight: response speed
HOW FAST the administrator answers the first inquiry is one of the strongest levers in the model, stronger than budget size. In medical cosmetology the effect is doubled: an injection visit ticket ≈₴7,800, the decision is considered over weeks, the client is anxious (“what if I ruin my face”) — and goes to whoever calmly replied first. In the “slow” scenario the model simultaneously lowers both “lead → booking” (38%→18%) and the share of “regulars” (42%→30%).
The speed effect is supported by independent research on 100,000+ leads (Oldroyd, MIT/InsideSales 2007; Harvard Business Review 2011): responding in 5 minutes vs 30 gives ~100× higher chance to connect and ~21× higher chance to qualify interest; within an hour — ~7× higher than after 2 hours. This is B2B sales in the US, not cosmetology direct messages — the decay of interest over time (attention pattern) is universal, and the specific multipliers are order-of-magnitude estimates rather than Ukrainian coefficients. The specific figure “−15% conversion in the slow scenario” is my model assumption [M], not a sourced fact. There is a direct link to Project 3 (routing inquiries and SLA ≤1 min) — and it’s logical to include Telegram there: time spent in Telegram in Ukraine increased 8× since the start of the war (Kantar Ukraine), 81% use it for communication (USAID/Internews 2024).
10Analogy: the leaky bucket
Business is a bucket of water. Advertising is the tap above (new clients). The bucket has holes: clients who leave (churn). If the holes are large (low retention), you can pour as much as you like — the level hardly rises. A junior sees an empty bucket and says “we need a stronger tap” (more budget). A senior looks at the bottom and patches the holes (retention). For LUMÉ there’s a third consideration: the bucket is almost full already (rooms at 76%) — pouring more simply has nowhere to go until you add another bucket (a new studio).
11Typical junior mistakes
- Calculating LTV by revenue — revenue is prominent in reports, but margin must be calculated separately.
- Judging a campaign by the first month — reports are monthly, and there’s a temptation for a quick “worked / didn’t work”.
- Looking at only one scenario — it’s easier to build one than two.
- Believing that increasing budget linearly increases clients — advertising has diminishing returns.
12How to estimate “what if” yourself
If asked about a scenario not in the model — talk the logic through out loud:
- Conversion drops somewhere → CPA rises (fewer paid customers for the same budget).
- Retention and the share of “regulars” rise → LTV jumps significantly (a regular brings ~8× more than a non-regular) — the main lever.
- Average ticket increases → contribution per visit and LTV increase, but CAC stays the same.
- Budget increases → proportionally more clients, CAC in the model stays the same (the model doesn’t account for audience exhaustion) — but since LTV:CAC is already ≈8x, the bottleneck isn’t here.
- Share of “regulars” rises by 5 percentage points → LTV jumps significantly (a regular brings ~8× more than a non-regular) — this is the main lever in the tornado analysis.
13Model limitations — say this yourself before they ask
What separates a good specialist is not a “flawless model” but an honest understanding of its bounds:
- Doesn’t account for CPA increasing with higher budget (in reality it usually rises).
- Some figures (CPL ₴620, “lead→booking” 38%, lifetime of a regular 2.8 years, shares of service lines in visits) are reasoned estimates [M], not confirmed facts for the Ukrainian market. In the primary-source registry they are marked with an estimation grade C.
- Calculates averages — in reality a client who gets “botox twice a year” and a client who takes “a course of biorevitalisation + meso + peels” have very different economics. The two-segment model (regular / non-regular) partially covers this, but not fully.
- Seasonality is modelled with a “Season” toggle (autumn–winter peak, summer drop) — based on Google Trends UA and business-broker (the first 3 quarters often loss-making, Q4 peak), but specific multipliers are estimations.
- No independent financial anchor: net margin of a mature studio ~9% and LTV ₴17,742 are cross-checked with industry corridors (business-broker: margin 10–20%; LTV cosmetology ₴30–100k/year for an active client), but not with real reporting of operating Kyiv chains (YouControl registries needed). So this is [M], not “market-verified”.
- Cohort retention curve is simplified: geometric decay (the same annual percentage). The real curve is steeper at the start.
- Mixed CAC uses one estimated figure for “alternative channels” (~₴330: referral ~₴300 + organic/offline) — exact proportions of paid/free traffic for the Ukrainian market weren’t found. Market size in ₴ also isn’t closed (Pro-Consulting report needed).
Stating this yourself in an interview is a strong move: it shows you understand the model rather than just believing the numbers on the screen.
14What’s added beyond the base model — and why
After the first version the limitations were honestly listed (section 13) — but some of them could be addressed rather than only acknowledged. Four additions turn “we didn’t account for this” into “here’s how to account for it, with caveats”.
Season — now a toggle, not just a caveat. The second switch at the top of the page (low/average/peak) changes client visit frequency: peak — 1.4 visits/month instead of 1.1, low — 0.8. Exact Ukrainian seasonality figures for cosmetology are unavailable; this estimates the direction of the effect.
Mixed CAC — not all economics flows through ads. A new block calculates not only paid CAC from the funnel but also mixed CAC — accounting for the share of clients from word-of-mouth and Google Maps/local search (default 20%, the slider is adjustable). The alternative channel figure (100₴) is an estimate, not a sourced fact.
Cohort retention curve — a trajectory, not a single number. Previously the model said “retention 35–50%” as one figure. The new block shows a monthly curve: what percent of the original cohort remains active month by month, simplifying that the same retention rate repeats each transition.
Model sensitivity — what to move first. The block on the main page changes each variable by ±20% one at a time and shows the spread of LTV:CAC. The key is visible: ad budget and CPL are low on the lever list, while share of “regulars”, visit frequency of a regular and gross margin are at the top. So money and attention should go to retention and capacity, not targeting — that is the senior conclusion of the model.
15Market retrospective 2021–2026 — why LUMÉ looks the way it does
The model didn’t “appear in 2026” — studio №1 opened in 2019, and the chain has gone through a full cycle. The retrospective is built on a real macro series (exchange rate, inflation, real wages — State Statistics Service / NBU), while LUMÉ’s trajectory over it is my model assumption [M].
- 2019–2021: one studio, 2–3 rooms, reaching breakeven. Ticket ≈ ₴1,850 (in prices of those years).
- 2022: collapse. GDP −28.8%, real wages −11.9%, exchange rate from ₴27 to ₴33. Demand for cosmetology fell (model multiplier 0.62), studio revenue −21%. But premium cosmetology clients postpone rather than cancel — and salons returned to work within weeks.
- 2023: a unique window. Service supply −30% since 2020, demand > supply ~7×, injection cosmetology in the capital +40%+. Studio revenue ×2.5. №2 opened in this window (end of 2023).
- 2024–2026: normalization with accelerating prices. Cosmetology prices rose (Barb.ua/Forbes give +33.6%, official medical CPI +6%; the model takes ~+12%) — the sector has pricing power. №3–4 opened in 2025, №5 — Q3 2026, already in the boom, but under tightening regulation (MOH licensing updates, ad labelling).
Retrospective revenue is calculated as: number of studios × ramp-up curve × category price indexation (not CPI — cosmetology became more expensive faster) × seasonality × macro demand multipliers by year. The full series is in the “Market Retrospective” research file.