The November Rate Letter Goes to 85 Assisted Living Families. A Flat 6% Is the Costliest Habit You Have.

Eighty-five letters go in the mail the first week of November, and every one of them says the same number. Six percent. Maybe five and a half. The Executive Director picked it in a budget meeting in September by taking last year's number, adding what the insurance broker said about renewal, and rounding to something that sounds defensible when an angry son calls.

That single percentage is, in my experience, the most expensive habit in assisted living. It is too high for the studio on the second floor that has been vacant twice in two years, and far too low for the memory care suite where the resident has drifted from care level two to care level four without the rate ever following her. Both mistakes are invisible until they show up as a move-out or a margin that quietly disappeared.

How the rate letter actually gets written today

The mechanics are the same in nearly every community. The Business Office Manager pulls a rent roll out of Yardi or Eldermark or ALIS. The Resident Care Coordinator has an assessment score for each resident — care points, levels, whatever your system calls them — but those scores were last refreshed when someone had time, which for most buildings means at move-in and after a hospital stay. Then somebody applies a flat increase to base rent, and sometimes a separate flat increase to the care level fees, and the letters print.

What nobody does is look at each unit's history: how long that apartment sat empty last time, what concession closed it, whether the resident's care needs have climbed two levels since the last time you touched her rate, and what the two competitors within four miles are quoting this month. There is no time. The Executive Director is running a building, and this is a two-week job done in an afternoon.

Demand forecasting and pricing are now among the highest-adoption uses of AI in American business — roughly 48% adoption for demand forecasting in manufacturing and about 72% for pricing in retail and e-commerce — because the 2026 models handle exactly the kind of messy, sparse, seasonal history that senior living has always had and classical spreadsheet methods always choked on.

Why the January bump was never as predictable as it felt

Every operator knows the pattern in their bones. Families gather at Thanksgiving and Christmas, see that Dad has lost fifteen pounds and three weeks of mail is piled on the dining table, and the inquiry line lights up the second week of January. Tours peak late January through March. Move-ins follow three to seven weeks behind the tour. Then June and July go quiet, and August is the worst month of the year for net census in most markets.

Skilled nursing has its own version, keyed to respiratory season: hospital discharges climb from December through March, Part A census and PDPM revenue climb with them, and so does your agency nursing spend, because everyone in your market needs the same nurses in the same eight weeks. Snowbird markets in Florida and Arizona layer a second, opposite pattern on top.

Knowing the shape is not the same as knowing the size. “January is busy” does not tell you whether to hold the two-bedroom at full rate through February or discount it now, and it does not tell you how many agency shifts to pre-book in December before the rate goes up.

flowchart TD
  A["Three years of move-in and move-out dates"] --> D["Weekly forecast run"]
  B["Care-level points per resident, month by month"] --> D
  C["Local competitor rates and hospital discharge volume"] --> D
  D --> E["Inquiry surge and summer dip, by unit type"]
  E --> F["Draft rate letter by care level, not one flat number"]
  F --> G["Executive Director and owner approve unit by unit"]
  G --> H["Letters mailed 60 days before anniversary date"]

What the 2026 models do with a thin, messy history

An 85-unit community has maybe 300 move-ins across five years. That is nothing by statistical standards, and half the fields are inconsistent because three different sales directors entered them. The older forecasting tools needed clean, dense, regular data and fell over on this. The current generation does not — you hand it the rent roll history, the assessment records, the inquiry log from your customer relationship system, and it works with what is there, tells you where it is guessing, and gives you a range rather than a false single number.

The output that matters to an owner is not a chart. It is three things: an expected move-in count by month by unit type for the next two quarters, a list of residents whose care points have drifted above what they are being billed for, and a per-unit price recommendation with the reason attached — “this floor plan filled in eleven days at full rate twice; stop discounting it.”

The tiering example, worked

Assumptions, all illustrative: 85 units, average current rate $5,400, 92% occupancy, and a historical pattern where roughly 4% of residents move out within 90 days of a rate letter they consider unfair. Compare the flat increase to a tiered one built off the forecast and the assessment drift.

The gap is roughly $109,000, and almost all of it comes from one thing: the 22 residents whose care has increased without their rate following. That is not a price increase. That is billing for work you are already doing. Most buildings I visit have between 15% and 30% of their residents in that position, and the reason is always the same — the reassessment happened in the chart and never made it to the rent roll.

The agency line the forecast actually pays for

The same forecast that prices the building should be sitting in front of your Staffing Coordinator in October. If the model says December through February runs eight to eleven residents above your fall census, at higher acuity, you can post open shifts early, offer your own staff the incentive bonus in November instead of the agency premium in January, and lock what agency you do need before everyone in the county is bidding for the same LPN.

Run the number for yourself: the difference between a staff LPN at roughly $38 an hour loaded and an agency LPN at $82 is $44. Twelve extra shifts a week for ten weeks is 1,440 hours — $63,000 of spread. You will not eliminate it. Cutting a third of it is realistic, and that is the same order of magnitude as the pricing win, from the same forecast.

What the forecast cannot see

It cannot see a bad survey. A CMS-2567 with an immediate jeopardy citation will dry up your hospital referrals faster than any seasonal pattern, and no model reads that coming. It cannot see the new 60-unit community breaking ground two miles away, unless you tell it. It cannot see your hospital partner signing an exclusive with a competitor, or a regional Medicare Advantage plan tightening length of stay by four days, which shows up as a revenue drop that looks like a census drop and is not.

It also will not tell you when a price increase is wrong for a human reason. The widow in 214 whose husband died in March and whose daughter is already asking about cost — the model sees a unit with an above-market rate and a high move-out risk. You see a person you are not going to push. Keep the override, use it, and write down why, because that note is the thing that makes the next forecast better.

Frequently asked questions

We only have three years of clean data. Is that enough?

For seasonality by month and by unit type, yes — three years gives you three Januaries, which is thin but usable, and the current models are honest about their uncertainty rather than pretending to precision. What matters more than years is whether your move-out reasons are coded consistently. If “moved to higher level of care,” “deceased” and “price” are all recorded as “other,” fix that first; it takes an afternoon and doubles the value of everything else.

Won't tiered increases cause a fairness complaint?

They can, if you cannot explain them. The defensible structure is: one base increase that applies to everyone and is stated plainly in the letter, plus separate care-level charges that follow the published assessment schedule in your residency agreement. Tie every care-level change to a documented reassessment the family has seen. What gets you in trouble is a rate that moves with no assessment behind it.

Does this replace the market study we pay for every year?

No. The market study tells you about supply coming into your submarket and what your competitors are actually charging, which your own history cannot know. The forecast tells you what your own building does month to month. Use both — and feed the market study's competitor rates into the forecast rather than filing it.

Forecasting the January surge only helps if you catch the calls it produces. CallSphere builds AI voice and chat agents that answer the inquiry line and web chat 24/7, ask the qualifying questions a family expects, and book the tour on the calendar — so the second week of January does not depend on whether your sales director was at her desk when the phone rang.