Prom Saturday Turns Clients Away While Tuesday 2pm Sits Empty. 2026 Demand Models Reprice the Gap.

Seventy-two percent. That is the share of retail and e-commerce businesses now using AI to set prices, according to the 2026 adoption figures — alongside 66% using it for inventory and 48% of manufacturers using it to forecast demand. Those three jobs are, almost exactly, the three jobs a spa owner does badly on a Sunday night with a legal pad: how busy will we be, how much product do I need, and what should Tuesday cost.

The reason those uses took off in 2026 is not that someone invented forecasting. It is that the newer models finally cope with the kind of history a 6-chair salon or a 2-room med spa actually has: three years of bookings, gappy, seasonal, interrupted by a remodel, with a stylist who left in March and took 200 clients with her. Classical forecasting maths choked on exactly that. It wanted clean, long, boring history. You do not have any.

October to February is laser season. June is not.

Every operator in this trade knows the calendar in their bones and almost none of them have it written down as numbers. Laser hair removal and IPL load up from October through February, because clients have to stay out of the sun between sessions and nobody wants to do that in July. Body contouring gets sold in January and February for a body someone wants in June. Injectables run hot from the last week of October to about the 20th of December, then fall off a cliff. Bridal trials cluster nine months before the wedding, which means your May bookings were decided in August. Prom is two or three specific Saturdays in April and May where you turn away money. Mother's Day weekend is the single biggest gift-certificate event of the year. And the last ten days of December sell colour, blowouts and gift cards that will be redeemed in a January you have not staffed for.

Demand forecasting in a spa means one plain thing: a number for how many of each service you will sell in each week ahead, built from your own booking history instead of from last year's feeling. Not a trend. A number, per service, per week, that you can staff and order against.

The four things you are currently guessing

First, injector days. Your RN or NP injector is the most expensive hour in the building, and you either book her for a Wednesday that produces four patients or you do not book her for a Thursday that could have produced eleven.

Second, product. Toxin comes in vials, filler comes in syringes, both have expiry dates and lot numbers, and your purchase price per vial depends on hitting volume tiers. Over-order in November and you are sitting on cash and a date stamp. Under-order and your injector is turning away a $700 ticket in the second week of December, which is the worst week of the year to be out.

Third, retail. The SkinMedica, ZO and Alastin shelf turns four times a year if you are good at it and once if you are not, and December is when clients actually buy skincare as gifts.

Fourth, price. Not your headline price — your off-peak price, your membership price, and whether the Tuesday 2pm facial should cost what the Saturday 11am facial costs. Most spas price by service. Demand does not arrive by service. It arrives by hour.

Why the report inside your booking system never answered this

Boulevard, Zenoti, Vagaro, Mindbody, Meevo and Phorest all have reporting. What they give you is history: last month versus the same month last year, revenue per stylist, retail attachment rate. Useful, and backward-facing. None of them tell you how many 100-unit vials to have in the fridge on 8 December, or what a Tuesday 2pm blowout should cost in June to fill rather than in December to ration.

flowchart TD
  A["Export 3 years of tickets from Boulevard"] --> B["Model reads demand week by week, service by service"]
  B --> C["Injectables: late Oct to mid Dec spike"]
  B --> D["Laser and IPL: Oct to Feb season"]
  B --> E["Colour and blowouts: Saturday crush, June lull"]
  C --> F["Vial order and injector days for Nov and Dec"]
  D --> F
  E --> G["Off-peak price for Tue and Wed 11am to 3pm"]
  F --> H["Owner approves the order sheet and the price sheet"]
  G --> H

What this looks like the week you actually run it

You export three years of completed tickets — date, time, service, provider, price paid, product used — and hand the whole file over at once. What comes back is not a chart to admire. It is two lists.

The first is an order and staffing sheet: for the week of 8 December, expect this many toxin patients and this many syringes, hold these injector days, and here is how confident the number is. The second is a price sheet: these named hours in these named weeks run below 60% booked every year, so these are the hours that carry an off-peak price, a member rate or a new-client offer — and these hours never get discounted because you already turn people away in them.

The owner approves both. That approval step is not ceremony. A model reading your booking history does not know your landlord raised rent, or that you are opening a second treatment room in March, and you do.

Eight empty chair-hours, priced

Illustrative numbers for a 6-station salon with two treatment rooms. Use your own.

The leakage line is the honest one and the one most vendors leave out: some clients who would have paid full price on Saturday will move to the cheap Tuesday. Your model should be asked to estimate that, and you should track it. If more than about a third of your off-peak fills are existing regulars shifting, you have cut your own price rather than found new demand.

What the history cannot see

A forecast reads what you sold, not what you turned away. If your phone went unanswered for 40 calls last December, those 40 clients are invisible to the model and your December looks smaller than it was. Same with the online booking page that showed no availability — the demand was real and left no trace. Before you trust a seasonal number, ask whether the trough is real demand or just an hour nobody could book into.

It also cannot see a new competitor opening two blocks away, a stylist's maternity leave, or the fact that you are dropping a service line. Those go in by hand.

And keep a human on the price sheet permanently. Aesthetics is a relationship business with a memory. A client who paid $700 for a syringe in November and sees a $499 offer in January does not think about demand curves, she thinks she got taken. Discount the hour, discount the new-client visit, discount the membership — but do not reprice the same service, for the same client, week to week, the way an airline does.

Start with one service and one quarter

Take a single service line where you already suspect the seasonality — laser hair removal is the usual candidate because the pattern is so strong — and forecast just that, just for next quarter, against three years of your own tickets. Write the number down before the quarter starts. At the end of the quarter, compare it to what happened and to what you would have guessed. If the model beats your gut, widen it to injectables and product ordering. If it does not, you have spent an afternoon and learned your own calendar.

Frequently asked questions

We have only 18 months of clean data because we switched from Mindbody to Boulevard. Is that enough?

For weekly patterns and day-of-week patterns, yes. For year-over-year seasonality, 18 months gives you one pass at each season, so treat the first year's seasonal call as a hypothesis rather than a fact. Old exports from your previous system are usually worth the hour it takes to pull them, even if the service names do not match perfectly.

Does off-peak pricing make my brand look cheap?

It does if you advertise a discount on the service. It does not if you price the hour or the membership. "Weekday member rate before 3pm" reads differently from "20% off facials" on the same page as your $1,200 laser package. Same money, different signal.

How do I use a forecast for toxin ordering when the tier pricing pushes me to buy more?

Run both numbers side by side: what the forecast says you will use, and what the next tier would save you per vial. Buy to the tier only when the forecast says you will use the product before it dates, and when the cash is not needed elsewhere in a January that you already know is thin. The forecast makes that a calculation instead of a rep conversation.

Who on my team owns this?

Whoever already owns the order sheet — usually the practice manager or lead stylist-manager, not the owner. It becomes a 20-minute Monday review of two lists, not a new job.

Forecasts also tell you when the phone is about to get loud: the last week of October, the Thursday before prom, the whole of Mother's Day weekend. CallSphere builds AI voice and chat agents that answer the salon or med spa line and web chat 24/7, book and reschedule into your booking system, and capture the enquiries that otherwise hit voicemail during exactly those peaks — which also means next year's forecast finally sees the demand you used to miss.