The best unit on the property had one price all year

Forty-six beds had an engine. The best unit had one number.

Our hostel in Oaxaca has 46 dorm beds and, inside the same property, one whole-house unit: six bedrooms, a pool, sleeps twelve, three-night minimum. Since August those 46 beds have been repriced every morning by software. Five lead-time bands, three competitors checked on Hostelworld before anything is written, 32 logged runs with nobody watching, zero rates touched by hand.

The whole house had one price. MXN 5,520 a night, every night of the year, since whenever it was first typed in. On 18 August someone added a hand-made override called "Low season 4000", which expired on 1 October. That is the complete pricing history of the highest-earning single unit on the property.

I did not notice for eight months. Why I did not notice is the useful part.

Eight months at one price

February through September 2026, out of the booking system:

• 141 nights sold out of 242 available, 58% occupancy
• MXN 766,000 in revenue, ADR MXN 5,436
• About MXN 96,000 a month, from one unit

Fifty-eight per cent on a whole house is not a crisis, and that is exactly what hid it. Nothing went red, nothing failed, nothing demanded a decision. It produced its 96,000 a month in silence and never asked me a question.

The number that ends the argument: 17,378

Day of the Dead, 30 October to 5 November, is already sold in that house at MXN 17,378 a night. Same six bedrooms, same pool, same linen, same cleaning cost. Three point one times the flat price we charge the rest of the year, and it sold.

So it is not a 5,520 house. It is a house whose market clears over a range of at least three to one, and we answered that range with a single number. A flat rate is not a neutral choice. It is a bet that your demand is flat, and one sold week at 17,378 proves ours is not.

What the other channels were already paying us

Airbnb pays 6,200 to 7,200 a night for that house, and our own Airbnb listing shows 10,199 all-in to the guest on a comparable date. Booking.com takes our direct price and keeps about 15%, so at 5,520 direct we net somewhere around 4,400 to 5,800.

Read that as an operator. The channel we control, the direct rate we tell guests is the cheapest place to book, was set about 45% below what we charged for the same nights on Airbnb. Nobody decided that. It is what happens when one number sits still for a year while another gets maintained.

The comparable set agrees. Eight guests with a pool in Oaxaca centro, 14 to 17 November: four-bedroom houses ask 3,900 to 8,000 a night, six-to-eight bedroom houses ask 14,000 to 21,000. We are in the second group.

The discount nobody decided, and the calculation that judges it

The "Low season 4000" override is the uncomfortable part. Under it, in August and September, a Booking.com sale on that house netted between 2,600 and 3,900. September closed at 73% occupancy with an average rate of MXN 3,009 a night.

73% looks like a win. Here is the arithmetic that tells you whether it was one. Roughly 22 of 30 nights at 3,009 is about MXN 66,000 for the month. The low-season base we have now set is 4,400. At 4,400, that same 66,000 needs 15 nights sold, which is 50% occupancy. So the question was never whether the discount filled the house. It is whether that house would have sold 15 of 30 nights at 4,400.

I do not know yet. But break-even occupancy is two divisions, and I did not do them before setting the discount. Do them before, not after. If the break-even number is one you have already hit in that month in a previous year, the discount is buying you nothing but a fuller house at a worse rate.

Where we had put the advance premium

In the first version of the rules we put the advance-booking premium at 61 days out, because 61 days sounds like a long way off to someone who sells dorm beds. Then we looked at who actually books it: the median booking for that unit arrives 101 days before check-in, a quarter land inside 21 days, they stay four nights, they arrive seven at a time, and weekdays sell exactly like weekends.

A premium starting at 61 days charged more to almost everybody and signalled nothing. We moved it to 90, where the median booker actually sits. If you run a whole unit or a group room, its lead time is nothing like your dorm lead time, and the rules in your head came from the dorms.

Two checks worth doing this week

• List every sellable unit, not every room type, and mark the ones whose price actually changed in the last 90 days. For each unmarked one, divide six months of revenue by nights sold, then compare what it earns against the attention it gets.
• Open your direct rate and your Airbnb or OTA price for the same future date, side by side, on the unit you manage least. If the direct number is lower by a wide margin, that gap is not a discount you chose.

What we changed, and when I will know

From 1 October the house is priced every morning by its own engine, separate from the dorm engine. A seasonal base (4,400 low, 5,500 shoulder, 6,800 high, 7,200 at Christmas) times a lead-time ladder: an open night inside three days at -25%, four to seven days at -20%, both with a two-night minimum so a gap can still be filled, then -12%, -5%, base, and +8% from 90 days out, all clipped to a floor of 4,000 and a ceiling of 8,000. A pace controller deepens the near-term steps three points a run while the next 30 nights sit under 60% booked. Day of the Dead gets no automated rate at all, by code and by test.

First read on the night we turned it on: 43% booked across the next 30 nights, so the controller took three extra points, leaving an open night inside three days at 4,000 and one inside two weeks at 4,680.

That is a decision, not a result, and it is one day old. What I expect: November, at 27% booked today, reaches 55% or better by 15 November without the 4,000 floor breaking, and the Booking.com net rises from about 3,000 to about 4,700 in shoulder season. I judge it on 15 November against today's 27% and MXN 5,360, and again on 15 January against last Christmas and January, which ran 52% at MXN 6,190 and 61% at MXN 5,658.

If I am wrong, that gets published here too. The point is not the ladder. It is that the most expensive pricing mistake on a property is rarely a bad price. It is the unit nobody is pricing at all, sitting on a number someone typed once, performing well enough that it never comes up.

If this sounds like your property, see the three Nightfill tiers and start a free 14-day pilot on your own data: view pricing.