GOeureka

Background

After years in the hospitality industry, our customer made sure of the inefficiency of the hotel business interaction with online travel agencies (OTAs). The point is that hotels don’t have enough resources to compete on equal terms with OTAs, so they are forced to cooperate with third-party platforms, often on unfavorable conditions. Such a situation makes hotels dependent as they cannot increase their profitability or attract clients for direct booking as well as reduces the opportunities and limits choices for commission rates.

Our customer came up with the idea to create an affordable and transparent Booking.com alternative applying Blockchain and tokenization to help hotels connect with consumers and conduct transactions directly. Such a solution allows excluding intermediaries and ensuring almost zero fee commissions.

Problem
to Be Solved

The four wholesalers provided data and services. The number of operations performed by their services was limited to 5. The customer required a load of 500 requests per second only for each B2B user as well as B2C portal. When working directly with the wholesaler's API, one would have to wait for a response to requests from all wholesalers, perform internal processing of the data obtained, and only after that send the answer to the user's request

The response time to the user could reach, in the worst case, 70 seconds.

Blockchain should be applied to pay for booking using GO tokens.

Workflow

To ensure the required number of requests, we decided to apply Big Data with forecasting and minimize requests to wholesalers. We divided the data from wholesalers into: “static” data that changes rarely and is updated below the peak system load times, and dynamic data requests that we cannot ignore.

We’ve created a database of 400.000 hotels, and added other information—photos, rooms, facilities.

The dynamic data was the room availability for the requested period and its price. We forecasted the room price based on Big Data, historical data, and the current prices with coefficients. The forecast accuracy was within 5%. This actual price was clarified during the booking process.

The search engine allowed giving customers search results with the required QPS.

Since the projected load was 2-3 million users, we decided to apply micro-service architecture which allowed us to easily scale the system to the required and larger loads.

The second phase was about applying hotel room management to integrate hotels directly into the client’s system. It means that hotels could conduct bookings on their websites through the platform and manage hotel occupancy. We also needed to provide HRM system for hotels so that they could manage their inventory and bookings.