Why AI pricing still fails hotels—and what needs to change

Source Why AI pricing still fails hotels—and what needs to change, PhocusWire

What It Says Tong Yin, CEO of InsightBridge Global, argues that current "AI" revenue management systems are often just automated suggestion tools built on obsolete assumptions. These legacy systems rely heavily on historical demand, fixed competitor sets, and OTA pricing signals—data points that are increasingly unreliable in a post-pandemic, AI-driven market.

Yin proposes a "three-layer architecture" for the next generation of pricing technology:

1. Demand Reconstruction: Using real-time data like flight capacity, event calendars, and search behavior on platforms like ChatGPT and Perplexity instead of just historical booking curves.
2. Net Revenue Optimization: Factoring in commissions and cancellation costs to prioritize net contribution over gross ADR.
3. Human-in-the-Loop Learning: Treating every manual price override by a manager as a training signal for the AI, rather than "noise" to be ignored.

Why It Matters For accommodation providers, relying on outdated algorithms leads to "incorrect price floors" and mispriced high-demand nights. Research suggests hotels using these legacy models may be losing between 8% and 14% of annual revenue. Furthermore, as travelers shift their research from Google to AI assistants like Gemini and ChatGPT, properties that lack sophisticated, independent pricing intelligence become more dependent on OTAs, surrendering their strategic advantage to third-party platforms.

Useful Signals

• Override Data: If staff are manually changing AI-suggested prices more than 50% of the time, the current system is failing to capture local market intelligence.
• Net vs. Gross: Pricing strategy must shift to "net revenue" to account for the true cost of OTA distribution.
• AI Search Impact: Emerging AI travel assistants prioritize "trusted sources," meaning accurate, real-time pricing data is critical for visibility in non-traditional search results.

STR Tech Report Take While Yin’s focus is on hotels, the short-term rental (STR) sector is even more vulnerable to these "broken assumptions." STRs are inherently fragmented and often lack the multi-year historical data found in the hotel industry. Modern STR revenue management tools must move beyond "comp set" scraping and start incorporating the "human-in-the-loop" feedback loop Yin describes. For STR tech vendors, the opportunity lies in building systems that learn why a property manager rejected a price—whether it’s a local festival, a hyper-local construction project, or a specific property nuance—and then encoding that expertise into the algorithm.

Original Source Why AI pricing still fails hotels—and what needs to change "The future of hotel pricing will not be fully automated systems replacing humans. It will be human-amplified systems that learn faster than competitors."