The beds filled on night four, and the week still took less money

The week at half price is over

On 27 September we overrode our own pricing engine at Maka, our 46-bed hostel in Oaxaca, and set the same-day and one-to-three-night rate bands to minus 50 per cent through 4 October. The engine had learned its way to minus 8 per cent on those bands over a month of daily runs. We ignored it and halved the price by hand on every bed a guest could book for tonight or the next three nights.

On 30 September we published a note here saying half price had not filled the beds. Two nights had settled at that point: 10.87 and 21.74 per cent occupancy. That piece said four nights is not a result, and it was right about that, because the beds then filled. Here is the whole window, now that it has run out.

Six settled nights, one price

• 27 September: 10.87 per cent, 5 beds of 46, MXN 737 for the night
• 28 September: 21.74 per cent, 10 beds, MXN 1,432
• 29 September: 26.09 per cent, 12 beds, MXN 1,665
• 30 September: 69.57 per cent, 32 beds, MXN 2,862
• 1 October: 50.00 per cent, 23 beds, MXN 2,544
• 2 October: 58.70 per cent, 27 beds, MXN 2,471

The first three nights of the override averaged 9.0 occupied beds. The last three averaged 27.3. Three times the beds, and the only thing that changed between them was the calendar. No price move, no new channel, no campaign.

One correction while we are here. That 28 September night was reported in this blog as MXN 1,554 when it was fresh. The ledger now reads MXN 1,432. A cancellation, a refund, a corrected channel payout, it does not matter which. If you are judging a pricing change on last night's number, you are judging a number that is still moving.

Where the three days came from

This is not mysterious, and it is the part worth keeping. A discount on the same-day and one-to-three-night bands can only be bought by someone deciding inside the next three days. It is invisible to everybody who booked us in August. So on night one, the house you are looking at was sold at the old price. On night two, mostly the same. By night four, the beds arriving are beds that were actually bought at the new one.

Which means a three-night read on a near-band discount measures the price you just stopped charging. We published one anyway. If you run this test, the window has to be at least twice the band it touches, and ours was not that until day four.

Then the money

The week before the override, 20 to 26 September, the engine ran the near bands at minus 8 per cent. Those seven nights averaged 31.06 per cent occupancy, 14.29 occupied beds, and MXN 2,755.60 a night at an ADR of MXN 193.

The three filled nights under half price, 30 September to 2 October, averaged 59.42 per cent, 27.33 beds, and MXN 2,625.70 a night at an ADR of MXN 96.

Nearly double the beds. MXN 130 a night less money. Across all six override nights, which includes the three the discount had not reached yet, the house took MXN 1,951.80 a night, 29 per cent below the week it replaced.

Put it in revenue per available bed, which is the only version of the number that survives a change in occupancy. The engine week: MXN 59.90 per available bed per night. The best three nights of half price: MXN 57.08. The override window as a whole: MXN 42.43. Occupancy up by 28 points, revenue per available bed down in both readings. An operator watching the occupancy report had a great week.

The calculation that actually decides it

On the filled nights we put 13 extra people in the building every night and took MXN 130 less room revenue for them. That trade is not obviously bad. Those 13 people drink, eat and book tours, and the engine tracks exactly that: it divides margin-weighted bar and activity income by occupied beds, at 27 per cent bar margin and 36 per cent activity margin, both taken from our 2025 profit and loss. The rolling figure right now is MXN 22 per occupied bed, against a 2025 baseline of MXN 32.

So 13.05 extra beds at MXN 22 of ancillary margin is MXN 287 a night. Against MXN 130 of lost room revenue, the week is MXN 157 a night ahead, before one peso of what those extra guests cost us.

Divide that 157 by the 13.05 extra beds and you get the number the whole week turns on: MXN 12. That is the break-even variable cost of one extra guest-night. Sheets through the laundry, water, gas for the shower, coffee, soap, the minutes of reception time. Under a dollar a head at any exchange rate you like.

I am not going to give you Maka's true variable cost per bed-night, because it is not cleanly separated in our accounts and inventing it would be worse than saying so. What I will say is that the verdict on a week of half-price beds sits inside a 12-peso window, and that nothing on an occupancy report comes close to telling you which side of it you landed on.

What else moved that week

The engine probes three competitors on Hostelworld every morning, for a dorm bed three days out. That floor read MXN 177 on 20 September and MXN 189 on 2 October. The market lifted about 7 per cent across the same window in which we halved. Some part of those 27 beds is a market coming out of a dead week, not a price, and one week at one property cannot separate the two. I would not trust anyone who told you it could.

The rest of the caveats, plainly: 46 beds, low season, one property, one owner override rather than a controlled test, no day-of-week control, six settled nights. This is a reason to change how we measure. It is not a finding about discounts in general, and it is certainly not a rate to copy.

What changes here

The override expires today, 4 October. The engine goes back to deciding, and it had settled on minus 8 per cent before we interrupted it. Two things change in how we test the near bands from now on. No near-band discount gets a verdict before night six. And the verdict is revenue per available bed plus ancillary margin per occupied bed, never occupancy. The next test runs ten nights minimum and we will publish the revenue per available bed, win or lose.

If you cut your near-term rates last month and called it a success because the house filled, the question was never whether the beds filled. It is what each of those beds left behind once the laundry was paid for.

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