Cencosud creates an observability center to accelerate business transformation across 5 countries

In today’s digital world, brick and mortar businesses need online storefronts and digital experiences that offer customers virtual as well as physical engagements. Over recent years, Censosud has modernized operations to enhance their digital infrastructure while building a store-brand ecosystem of applications that focus on high performance with compelling customer experience opportunities.

Working across countries and with multiple applications and brands to manage, Cencosud needed a common observability culture and tooling to ensure high performance across all infrastructure and a continuously evolving focus on customer experience. 

Building a common culture and standardized processes became a top concern at Censosud. The company wanted to move beyond implementing improvements as one-off approaches, and create a culture of observability, to accelerate digital transformation securely, with data driven decisions that make optimal use of their modern Kubernetes microservices and Amazon Web Services (AWS) infrastructure. 

Starting change, enterprise-wide

Creating a common focus at an enterprise group can be a huge task. To do this, Censosud created a new team, Cencommerce. Cencommerce worked like an internal startup within the enterprise: testing new more agile and product-oriented ways of working that were more suited to digital delivery.

Starting with around 150 team members, Cencommerce focused on creating common observability tooling that could measure the impact of their online applications. Quickly demonstrating the business value of having high-performance, fast, and seamless digital customer interactions, the team grew to 400 developers by 2021 and doubled the year after in 2022.

Implementing observability

One of the biggest challenges Cencommerce faced during its initial establishment was the disparate cultural approaches across the enterprise. With multiple countries providing services, at different stages of maturity and each with a unique culture, observability was viewed in different ways by each team.

Engineering Director Pablo José Duran Rojas needed to find a way to unite team members across the continent and enterprise operations. Introducing a common observability framework. Tooling and automation was key to supporting better communication and a sense of shared identity for Cencommerce team members.

Drawing on New Relic as a common observability platform for all staff, Pablo created a framework for monitoring infrastructure, legacy applications, microservices, and more newly built solutions. The next challenge was getting teams to adopt and understand the framework. To help with this, the new framework established three levels of activity that helped team members take a path to maturity in their observability activities.

The framework defined three levels:

1. Gather information using application performance management

   While teams can understand the importance of monitoring, it can be deprioritized compared to business needs, especially when things get hectic, says Pablo. To reduce the burden of having to establish new systems for each application ahead of a deployment, Cencommerce built an automated solution so that as an application is created or upgraded, monitoring code is automatically injected which in turn adds the application to the New Relic observability platform.

2. Monitor KPIs using team requirements

   Now with all applications being monitored regularly, teams were encouraged to look at how to leverage observability to ensure their applications were performant, engaging, and available when needed. Teams were encouraged to use metrics to identify innovation and new feature opportunities to continually enhance customer experience.  

3. Standardize metrics using golden signals
   
   Drawing on the metrics identified by teams at Level 2, teams were now encouraged to create golden signals for their applications. To provide a common observability language, the enterprise used four key metrics:
   
   Latency: How long an application takes to respond.
   Error rate: What proportion of queries cause an error.  
   Saturation: Amount of resources that are being used by the application.
   Traffic: Volume of requests to the server. 

   
   By ensuring these four signals remain within standard thresholds that reflect high performance and robust delivery of applications, teams can quickly address any anomalies as they arise, reducing any potential negative impacts on the business. 

Proactive and responsive to business needs

With this observability framework in place, Cencommerce created a robust observability structure, where predictive and threshold alerts are automatically triggered whenever applications or infrastructure under-perform at expected levels. This allows teams to be both proactive (for predictive alerts) and responsive (for threshold alerts), continuously improving performance or tweaking infrastructure supports at peak moments of demand, whether occurring in modern or legacy applications.