What we learned from 90 days of Autopricer in production
Channel Autopricer for Lodgify went live for our first batch of paying hosts on 20 March 2026. Ninety days later, 340 hosts are running it on around 4 200 units, and the module makes roughly 4 200 pricing decisions per night — one per unit, one per night, on a rolling 90-day horizon. That's about 380 000 decisions a month. Here are the three things I did not expect.
1. Hosts don't want a "smart" pricer — they want a defensible one
Our first prototype was fancy: a small ML model taking booking pace, comp-set position and lead-time curve as inputs, spitting out a price. Hosts hated it. Not because it made bad decisions — it actually did fine — but because when a guest asked "why is the same room 240 € tonight and 190 € tomorrow?", the host couldn't answer. So we ripped it out and replaced it with a rules engine: floor, ceiling, weekday multiplier, occupancy stepping, comp-set adjustment. Every price is now explained by a two-line rule you can read in the audit digest. Complaints dropped to zero.
2. The ceiling matters more than the floor
Everyone worries about pricing too low — a race to the bottom that empties the calendar at 39 €/night. In practice, floors are boring: hosts set them once and never touch them. What matters is the ceiling. If you leave the default at 3× the base rate, Autopricer will happily push a Cannes villa to 1 100 €/night for a Formula 1 weekend, and 80% of those hosts then manually cap it back because "that just feels wrong." So we now surface the ceiling prominently on the setup screen and suggest a 2× default. Adoption of manual overrides has since fallen from 22% of hosts to 6%.
3. Sunday nights are the leading indicator
We looked at where Autopricer's changes correlated most tightly with a booking within the following 24 hours. It wasn't Fridays, which we expected. It was Sunday nights. A price drop on a Sunday night for a stay two to three weeks out was the single biggest predictor that the unit would sell. We now weight Sunday-night decisions slightly higher in the rules engine, and the module's contribution to weekly RevPAR jumped from +4.1% to +5.7% in the following month.
What's next
Two things on the roadmap for Q3: a longer 180-day horizon for hosts serving corporate long-stays, and a soft floor that decays automatically when a date approaches without bookings (instead of the current hard floor). Neither changes the "explainable" promise — you'll still be able to read every price back to the rule that produced it.
If you're on Lodgify and would like to try Autopricer, the module page has the full spec and 14-day trial with no card required.