Case Study

Invoice to Booking reconciliation in Travel

A leading cruise and package holiday agency partnered with Data Cactus to automate their manual invoice-to-booking reconciliation process. By leveraging AI-powered document processing, custom-built matching algorithms, and a centralised Azure data platform, we reduced staffing needs by 75%, improved accuracy, and enabled real-time booking updates—laying the groundwork for further innovations like a customer-facing AI chatbot.

Background

Client: A travel agency specializing in cruise retail and package holidays
Industry: Travel
Scale: Processing 500 bookings and 1,000-2,000 documents daily
Challenge: Manual reconciliation of supplier invoices with booking data

Challenge

Our client, a prominent travel agency focusing on cruise and package holidays, was facing significant operational challenges due to their document-heavy business processes. With approximately 500 bookings processed daily, resulting in 1,000-2,000 supplier documents (invoices, tickets, and ATOL certificates), the agency was struggling to efficiently reconcile this information with their booking system.
Key pain points included:
• A dedicated team of 8 full-time employees manually verifying tickets and invoices against bookings
• High risk of human error in the verification process
• Delayed processing times affecting customer service
• Revenue leakage due to missed discrepancies between supplier invoices and actual bookings
• Particular challenges during booking amendments, when customer changes frequently led to incomplete information being passed to suppliers
• Wide variation in document formats across numerous travel suppliers
• Legacy systems in the travel industry creating additional complexity
• Inefficient utilisation of staff resources
This manual approach was not only costly but also created bottlenecks in their operation, preventing the agency from delivering the level of service they aspired to provide. Furthermore, the organisation was unable to scale effectively during peak booking seasons without significant temporary staffing.

Solution

DataCactus designed and implemented a comprehensive automation solution to streamline the invoice-to-booking reconciliation process. The solution included:

1.Centralized Data Infrastructure
• Implementation of an Azure data warehouse utilizing Azure Data Lake, Fabric, and Data Factory
• Development of data pipelines to extract booking information from the Traveltek iSell system and CRM
• Creation of a unified data repository for all booking and invoice information

2.Intelligent Document Processing
• Implementation of advanced AI document extraction using OpenAI with a bespoke training model specifically designed for travel invoices and tickets
• Automated extraction of key information from PDF attachments including passenger details, travel itineraries, flight information, and pricing
• Specialised handling for various document formats from multiple suppliers

3.Automated Reconciliation Process
• Development of matching algorithms to compare extracted document data with booking system records
• Implementation of alert systems to flag discrepancies requiring human intervention
• Creation of dashboards for monitoring reconciliation status and addressing exceptions

4.Booking Amendment Tracking
• Specialised workflows to monitor changes to existing bookings
• Verification that all booking amendments were properly communicated to suppliers
• Automated validation of updated passenger information, extended itineraries, or added excursions

Implementation

The project was executed over an 8-month period and included these key phases:

1.Discovery and Requirements Gathering (Month 1-2)
◦ Analysis of existing manual processes
◦ Documentation of various supplier invoice formats
◦ Definition of reconciliation rules and exception criteria

2.Data Infrastructure Setup (Month 2-3)
◦ Implementation of Azure data warehouse
◦ Configuration of data pipelines for booking system integration
◦ Design of data models and schemas

3.AI Model Development (Month 3-5)
◦ Training of the OpenAI model with travel industry documents
◦ Development of extraction algorithms for various document types
◦ Testing and refinement of extraction accuracy

4.Reconciliation Engine Development (Month 5-6)
◦ Creation of matching algorithms
◦ Implementation of discrepancy detection
◦ Development of alert mechanisms

5.Integration and Testing (Month 6-7)
◦ End-to-end system integration
◦ Performance testing with historical data
◦ User acceptance testing

6.Deployment and Training (Month 8)
◦ Phased rollout of the solution
◦ Training of the remaining reconciliation team
◦ Establishment of monitoring and support processes

"The automated reconciliation system developed by DataCactus has transformed our operations. What previously required a team of eight people can now be managed by just two staff members, with greater accuracy and speed. The system has paid for itself within the first year through direct cost savings, and the improved data quality has allowed us to implement additional customer-facing enhancements like our new AI chatbot. Most importantly, our customers are receiving better service, especially when making changes to their bookings."

Head of Operations, Travel Agency