As part of the strategic acquisition of an American startup publishing a cutting-edge analytics platform, a global leader in advertising technologies and Retail Media called upon MARGO to master and integrate this critical data asset.
This platform, essential for retailers to steer the performance of their products, relied on a complex, poorly documented legacy architecture, initially maintained by a team of 40 engineers.
Intervening during a critical transition phase where headcount had already fallen to 20 people, MARGO deployed an elite engineering unit to ensure the reverse engineering of the system, guarantee absolute service continuity, and orchestrate the sunsetting strategy of the application. This intervention secured the intellectual property while drastically reducing the technical resources mobilized (from 40 to 3 engineers), thereby ensuring the technological success of this high-tension M&A integration.
Challenges: mastering a critical technological asset in a merger and acquisition context
During an acquisition, the target's technical debt often represents a major risk. Following the buyout, the group imperatively needed to regain control of the target platform, a complex analytical engine blending advertising campaign data, massive web scraping, and delivery via dashboards.
The stakes went beyond mere maintenance in operational conditions. It was a matter of defusing a technological « black box » to avoid any impact on our client's business:
Financial and image risk:
Prevent any loss of visibility for retailers on their performance KPIs.
Operational risk:
Secure the Sales teams' ability to support their clients with reliable real-time data.
Continuous development:
Maintain velocity to develop new strategic features requested by clients.
Risk of Intellectual Property (IP) loss:
Capture the business knowledge of a system maintained by a departing American team.
Objective: a rescue mission and strategic transmission
The challenge for MARGO was not to indefinitely maintain an aging infrastructure, but to steer a true technological safeguarding operation during a pivotal phase. In this uncertain transition context, the mission transformed into a resilience challenge built around a critical triptych.
First, the absolute imperative was business continuity: guaranteeing « Zero Downtime » so that the platform buyout remained seamless and painless for end retailers.
Next, the challenge was human and operational: it was necessary to industrialize the impossible by enabling a commando team of 3 MARGO experts to absorb the workload and complexity previously managed by the initial team of 40 engineers.
Finally, the mission targeted the future. It was not just about « running » the system, but preparing its intelligent sunsetting. The objective was to transform a complex and undocumented legacy code into a structured knowledge base. By extracting calculation rules and data schemas, MARGO enabled the group's internal teams to take over the torch and rebuild these strategic functionalities on their own technological standards, without loss of know-how.
Approach: industrialization, microservices and asynchronous agility
To achieve this maximum efficiency, the MARGO team replaced a manual approach with aggressive industrialization on the Google Cloud Platform (GCP) foundation.
Resilience through orchestration (Kubernetes): since large-scale scraping (a minimum of 2 million web pages scraped per day, covering products, categories, and keywords) is inherently unstable, MARGO leveraged Kubernetes to isolate each collection engine in ephemeral containers. In case a scraper failed, the infrastructure self-healed without human intervention, allowing a reduced team to manage massive flows.
Microservices architecture and asynchronous communication (GCP Pub/Sub): the architecture relied on decoupled microservices communicating via Google Cloud Pub/Sub. This model enabled high scalability, better resilience, and independent evolution of services despite headcount reductions.
FinOps and scalability (BigQuery): to absorb large variations in data volume inherent to migration phases, the architecture was oriented towards Serverless via BigQuery. This model guaranteed consistent performance for users on Looker Studio.
Asynchronous governance (Kanban): faced with the constraints of the time difference with the United States, MARGO established a strict Kanban framework, removing human bottlenecks and ensuring a smooth knowledge transfer.
Delivery: the 3 steps of a high-tension transition
MARGO's intervention was not limited to simple maintenance; it followed a methodical execution narrative to transform a major risk into a documented success.
Step 1: immersion and risk mapping
From day one, the challenge was to absorb a complex architecture (Python, SQL, Angular, Django, PostgreSQL) in record time. This Flash Audit and Knowledge Transfer phase allowed for mapping functional risks and securing critical access, thereby creating a knowledge bridge between the outgoing American team and the MARGO crisis unit.
Step 2: industrialization at the service of the run
Once the platform was mastered, the team moved into the industrialization phase. The objective: stabilizing data pipelines via proactive monitoring and automating flow restarts. By becoming the guarantor of the data, MARGO ensured high-responsiveness support for the Sales teams, allowing client requests to be handled without any delay despite the massive reduction of the original technical team. It was during this phase that the scale capacity was proven, with the scraping of several million pages per day.
Step 3: sunsetting and heritage transmission
The mission concluded with a secured shutdown phase orchestrated alongside SRE teams. The achievement of this step lies in reverse engineering: the MARGO team formalized all processing workflows and business rules. By documenting every KPI and every flow, we guaranteed that the startup's data heritage and intellectual property were fully preserved and ready to be reintegrated into our client's proprietary solutions.
Results: a secured integration and multiplied efficiency
The MARGO team managed to take control of a complex platform very quickly and ensure its stability within a highly constrained post-acquisition context. Responsiveness to incidents and the rigor of their methodology were decisive assets.
Group Engineering Manager
Are you facing similar challenges? Contact us!
This intervention illustrates MARGO's ability to position itself not merely as a player in operational maintenance, but as a strategic partner during critical phases of technological integration (Tech M&A).
By combining Cloud expertise (GCP/Kubernetes/Pub-Sub), reverse engineering methodology, and operational pragmatism, MARGO enabled this Retail Media leader to secure a major business asset while drastically rationalizing the associated operating costs.
Turn your challenges into opportunities with innovative and tailor-made solutions!
How does MARGO transform technical debt into an actionable documentary asset?
During a buyout (M&A), legacy code is often a black box. MARGO's approach goes beyond simple operational maintenance: our engineers perform functional reverse engineering. By decrypting and formalizing the KPI calculation rules hidden within the code before decommissioning the system, we transform obsolete software into a clear knowledge base, enabling the reconstruction of strategic features on the client's target technology stack.
Why is Kubernetes orchestration critical for industrial web scraping?
Massive scraping is subject to high volatility from target e-commerce sites. Using Kubernetes allows for compartmentalizing each collection engine. This « self-healing » architecture guarantees total resilience: failing processes are automatically restarted by the system, allowing a tiny team to manage colossal data flows while minimizing manual support work.
What is the benefit of a « Serverless » cloud architecture during a technological transition phase?
During a migration, ingested data volumes can experience unpredictable peaks. By centralizing workloads on a Serverless technology like BigQuery (GCP), rigid hardware infrastructure is replaced with elastic computing power billed on usage (a FinOps approach). This guarantees end users optimal response times on their dashboards, even during heavy processing, while controlling cloud costs.
