Travel planning involves vetting multiple choices for destinations, hotels, travel agencies, and more. Starting this process from scratch every time can be a tiring process. That is why our client, a travel technology company, provides user-catered smart ads to website visitors based on ML-backed insights. Customized ads also help the travel companies, suppliers, and publishers to increase revenue by getting more visitors to click through the ads shown to them.
During each session when a user loads a web page, ads from across 50+ different publisher sites are displayed, and a ‘conversion’ happens when a visitor clicks an ad and fills the form.
A data pipeline built on a leading public cloud provider’s platform facilitated the whole process of carrying out and tracking the conversion, that is, linking publisher website visits to the source ad. This helps the client determine the advertisements’ impact on the visitors and the travel companies’ outcomes and optimize the ad units.
The client wanted to increase the processing speed and decrease the costs incurred for each conversion. They collaborated with Persistent, a marketing analytics award winning consultancy, to evaluate the effectiveness of executing the conversion linker data pipeline on Google Cloud.
The engagement went through the following steps:
- In a design workshop, the Persistent team explored the client’s data and assessed the architecture to create a low-level architecture design
- Persistent synced the raw data between the previous cloud and Google Cloud
- The team set up BigQuery & BigTable and created DataFlow pipelines for data ingestion and conversion linking
In just 8 weeks, the Persistent team provided the end-to-end data pipeline architecture and developed both the stream and batch versions of the pipelines on Google Cloud.
50X improvement in the conversion process with high-speed, high-throughput data pipelines
The previous cloud provider could carry out 500 conversions in 40 seconds whereas Google Cloud could carry out about 25K conversions in the same time period for the same amount of data.
