Data‑Driven User Experience: Machine Learning Applications in Hotel Booking and Marketing at Ctrip

At the 2015 China Hotel Marketing Summit organized by Global Travel News, Ctrip Senior Vice President and Chief Technology Officer Ye Yamin delivered a keynote describing how Ctrip uses machine‑learning models based on purchase behavior and order transactions to predict non‑reserved room status, shorten confirmation times, recommend hotels, and measure advertising effectiveness.

Host Wang Jing recalled a previous open‑day event in Shanghai and highlighted the impact of Ye's presentation, noting its open, relaxed, and egalitarian spirit.

Ye introduced himself, noting he joined Ctrip in 2011, worked overseas, and was involved in hotel product R&D and technical architecture; he emphasized Ctrip's shift from traditional internet to mobile internet, with mobile bookings now accounting for over 60% of total reservations.

He outlined three 2015 breakthroughs: using data to enhance user experience, applying technology to improve operational efficiency, and providing technical support for Ctrip's goal of ten‑fold growth.

He stressed that more than 60% of Ctrip's customers interact via the app, which offers a one‑stop travel solution and has earned numerous awards, including recommendations from the Apple App Store.

The current version 6.4 and the upcoming 6.5 will enrich the product lineup with hotels, apartments, inns, scenic spots, and surrounding services.

Ye presented a customer‑centric data framework covering pre‑travel, in‑travel, and post‑travel phases, citing KPIs such as arrival rate, secondary consumption, churn rate, and others that help evaluate service quality.

He introduced the concept of a retention‑rate KPI, noting that many of these metrics are adopted from successful overseas companies.

One scenario described how an instant order confirmation system—"smart room reservation"—reduces customer waiting time by immediately confirming bookings when possible, thereby increasing conversion and delivering more guests to hotels.

Another scenario explained a model that alerts hotels to respond after a predefined waiting period, improving hotel efficiency while enhancing the guest experience.

During travel, Ctrip can push personalized recommendations (e.g., nearby attractions or partner shops) based on a guest’s prior interests and purchases, such as suggesting a shop within one kilometre to a Chinese tourist who has shown shopping intent.

He gave an example of a churn‑detection model that identifies hotels losing about 200 guests per day, emphasizing the win‑win relationship between OTA platforms and hotels.

Ye discussed measuring the impact of a long‑running TV ad (“Say Go”) using big‑data techniques: comparing "Before Visits" and "After Visits" to assess changes in traffic, demonstrating a clear shift in distribution that indicates ad effectiveness.

He compared Ctrip’s customer retention rates (e.g., 62% after one year, over 50% after two years) with competitors, showing Ctrip’s superior performance.

The talk highlighted that customers may switch platforms, but many remain loyal to Ctrip due to its service value.

Beyond mobile, Ctrip is exploring device‑level experiences, such as an Apple Watch app, which required multiple design reviews with Apple to meet strict guidelines before launch.

Ye concluded his presentation.

He added that Apple chose Ctrip’s app because of its sleek design, relevant use cases, and the strong partnership between the two companies.

Host Wang Jing asked a follow‑up question about the depth of the data shared.

Ye responded that, as a technologist, he prefers to focus on technical details rather than business storytelling.

Li Chao mentioned an additional article about Ctrip’s Apple Watch strategy that was shared earlier in the morning.

Ye noted that the advertising content was minimal and straightforward.

Li Chao thanked Ye and invited further discussion during lunch.

Wang Jing asked about a Siri test regarding Shanghai‑dialect flight booking and its feasibility.

Ye explained that voice‑recognition technology has become highly accurate (over 90%) for various dialects, and Ctrip’s system can convert speech to text and provide instant, relevant suggestions, positioning Ctrip as a leader in voice‑driven travel search.

He demonstrated that the search box includes a microphone icon where users can speak their intent and receive accurate results.