From bare metal to cloud native: how we rebuilt our ads operations tool to empower business analysts

For years, our marketing operations team relied on a homegrown, on-premise tool to bridge the gap between business analysts and major ad platforms like Google Ads, Bing, and Meta. Our analysts use the tool to upload CSV files and to orchestrate campaign operations without ever touching a line of code.

For years, our marketing operations team relied on a homegrown, on-premise tool to bridge the gap between business analysts and major ad platforms like Google Ads, Bing, and Meta. Our analysts use the tool to upload CSV files and to orchestrate campaign operations without ever touching a line of code. Some examples are how these CSV files are used to trigger the creation of campaigns and their associated budgets, as well as to provide customization parameters that control how campaign information is exposed to end users via TripAdvisor’s partner APIs. Beyond streamlining campaign orchestration, the tool standardizes interactions with each partner’s API, significantly reducing the learning curve and operational burden on analysts who would otherwise need to understand the unique nuances of each platform’s technical requirements. While the tool remained functional, its aging architecture was becoming increasingly apparent.

The pain of staying on-prem

Running this tool on-prem came with mounting challenges:

• Infrastructure maintenance was tedious. Any server issue meant downtime.
• Costs weren’t transparent since the server was shared with other teams.
• Modernization felt out of reach. Integrating newer services meant reinventing the wheel.

Worse yet, as our ad strategies grew more complex, analysts needed faster development iterations.

The migration mandate

We had a clear goal: move to the cloud, reduce costs, and rebuild using cloud-native technology.

But we couldn’t just perform a “lift and shift.” This was our opportunity to rethink the architecture to:

• Empower analysts with a smoother, faster experience.
• Make the system modular, observable, and scalable.
• Leverage managed services wherever possible.

Re-imagining the tool as cloud native

Before we got to work redesigning the tool, we thought about the flow of the core user journey:

1. An analyst uploads a CSV file describing the operation.
2. The analyst selects the ad platform (Google, Bing, Meta) and the operation type (for example, create campaign).
3. The tool validates the data and orchestrates API calls to the relevant platform.
4. Status and results are shown on the frontend.

Designing the architecture

To execute the business logic, such as creating campaigns, we use Apache Airflow on MWAA (Amazon Managed Workflows for Apache Airflow) to orchestrate each step. Apache Airflow is an open-source workflow management platform that allows users to programmatically author, schedule, and monitor complex data pipelines. At Tripadvisor, we use Airflow not only to coordinate campaign workflows but also to manage our account structures across each partner’s advertising platform and to handle all types of business-related jobs.

Gallery