Duetto's New Forecasting Engine Hits 95% Accuracy and Removes the Manual Tuning Revenue Managers Used to Be Judged On

95.22% average forecast accuracy across Duetto’s customer portfolio, with some properties hitting 99% — that’s the headline number behind Duetto’s replacement for its legacy historical-average forecasting model. The new engine blends booking pace, seasonality, competitive dynamics, and local events, then trains a model on each hotel’s own booking data instead of relying on manual property-level tuning. Duetto reports a 12% improvement in sMAPE and an 8% improvement in WAPE against the prior system, with monthly aggregate forecast errors running 1–6% inside 30-day windows. A new “driver waterfall chart” shows revenue managers which factors are actually moving a given forecast, rather than handing them a number with no explanation attached.

The property-level tuning Duetto is removing was, until recently, treated as a core revenue-management skill — the thing that separated a strong RM from an average one. That’s the same shift Duetto itself was naming back in August, arguing revenue managers need to move from RevPAR-only thinking toward “performance engineering” — a profit-centered discipline built around GOPPAR, TRevPAR, and CPOR — because automation was coming for the mechanical parts of the job regardless. This forecast release is that argument’s technical follow-through: it doesn’t just speed up rate-setting, it takes the manual-tuning variable off the table entirely.

It also lands next to independent evidence that automated pricing decisions can outperform manual ones at scale. A causal-inference study covering 6,000 hotels found Mews RMS’s Autopilot feature lifted revenue per square meter 13% over 18 months, driven by price changes running up to 14x more frequently than a human RM could sustain — and doing so without trading rate for occupancy, the tradeoff manual pricing typically can’t avoid.

The pattern across both vendors: the metric that used to signal RM skill — frequent, accurate manual repricing — is becoming the metric that signals the system is working without one.