Challenge: Department crisis and onset of bots
An AdTech company faced a crisis within its Data Science department when an unmanageable bot problem emerged. The issue was so severe that it skewed ad targeting and wasted budgets on non-human traffic, resulting in extremely low click-through rates and financial damage.
Launching the restructuring process
The severity of the bot issue made it clear the existing Data Science team could not cope. After the dismissal of the Head of Data Science, the company sought ASAO DS’s expertise for guidance to address current technical, operational, and structural challenges.
Our answer: the Mary Poppins Option—our senior partner moved on-site to take on the role of Interim Head of Data Science.
Diagnostic: Identifying core issues
Upon arrival, our Partner interviewed everyone from founders to engineers, revealing several critical issues:
Outdated models: The existing models relied on biased data, heavily influenced by past ad placements and overrun by bot interactions, making them ineffective for current needs.
Team misalignment: Initial hiring favored academic technical skills without evaluating their relevance to real-world business challenges. The result was a team that was unfortunately poorly matched to the company's current needs.
Legacy issues: Lack of ownership and monitoring let legacy code and persistent bugs accumulate, slowing down necessary updates and causing data discrepancies.
Data Warehouse chaos: No ground source of truth meant data was unreliable.
Solutions: Roadmap and Departmental rebuild
With this in mind, the Partner worked with the Client to devise a strategic roadmap for immediate and future transformations:
Data enrichment: Identified third-party data enrichment services and supervised their integration to imrpove model accuracy.
Framework reworking: Established a new, practical model development framework, making sure new models will be useful, adaptable, and fast.
Team optimisation: Reevaluated each team member's skills, retaining those aligned with strategic goals and parting ways with others.
Over the next months, our Partner led a series of transformative initiatives:
Data Warehouse redesign: Reconstructed to serve as the single source of truth.
New protocols and collaboration: Implemented new standards for model development and set up effective cross-team communication.
Ownership and processes: Defined clear roles and zones of responsibilities, complete with processes for ongoing analytics, monitoring, and alerting to ensure constant oversight.
Skill-based hiring: Recruited new Data Scientists and Analysts with both technical skills and a strong business perception.
New bot detection: Developed an original bot detection model using integrated third-party data, significantly improving precision.