How PUMA Broke Down Data Siloes and Built Observability Into Their Systems

PUMA’s production environment includes elements of Salesforce Commerce Cloud (SFCC) and uses Fastly CDN, which produces enormous volumes of data. With most observability solutions, there is an unavoidable tradeoff between the centralization of growing data and the associated costs and performance deficits that come with it.

With that said, not having visibility and proactive monitoring in place put PUMA at risk of damaging their customer experience and sustaining revenue loss due to incidents that didn’t trigger alerts to the team or took too long to resolve.

During a big sale event on PUMA.com in the US, they experienced a problem with a 3rd-party inventory service that was meant to check if an item was in stock and return a response. Subsequently, none of the orders received the response they needed, and they all failed.

Without comprehensive monitoring and alerts, it took an hour for the team to notice an issue with all orders failing before investigating and resolving it. During Private Sales, an hour of downtime can mean upwards of $100,000 in lost revenue. The team needed a way to monitor their systems’ data comprehensively and trigger notifications for events and anomalies such as higher than normal ‘request fail’ or ‘error’ ratios.

With Coralogix, PUMA could leverage their log data in many ways. For example, they have logs that signal when an order is transitioned to ‘new’ or ‘failed.’ They can easily look at the ratio of ‘failed’ to ‘new’ orders and create a Ratio Alert to trigger when the ratio is higher than usual. This helps them respond to critical situations faster.

As part of an initiative to build a new headless frontend and native mobile app, PUMA’s DevOps team for Global E-Commerce built a new public cloud infrastructure. In doing this, they improved their ability to access system data and saw the opportunity to approach observability from the ground floor. This is where Coralogix comes into the story.

Michael Gaskin, Senior DevOps Manager for the group, met the Coralogix team at an industry event and was intrigued by the potential of the unique in-stream data analytics engine and the ability to aggregate logs as metrics without indexing the raw log data.

When he continued exploring the platform’s capabilities, he saw that he could easily install a serverless layer to collect logs from their serverless functions without needing a big professional services engagement or a ton of developer time.

“Once I installed that serverless function layer, I saw the logs from our serverless functions streaming into Coralogix in the LiveTail view. I just thought, wow, this is good time-to-value.”
– Michael Gaskin, Senior DevOps Manager at PUMA

With the streaming architecture, the Coralogix platform offers unparalleled cost savings that ultimately allow for more expansive data collection and better observability. During the evaluation process, the team worked with the customer success team at Coralogix to ensure full data visibility across all environments and quickly saw the platform’s value.

Almost immediately, the team decided to move forward with Coralogix and hit the ground running with the Coralogix team providing an incredibly smooth on-boarding.

“The willingness of Coralogix to do anything that needs to be done to get the data in and get customers up and running is admirable.”
– Michael Gaskin, Senior DevOps Manager at PUMA

To ensure the teams at PUMA had full visibility into their application data, Coralogix’s Customer DevOps team built custom serverless application repository apps that create serverless functions to pull logs in from Salesforce as well as get order data from Salesforce using the Open Commerce API.

Adoption in the development and infrastructure teams has been great. Developers and infrastructure teams immediately saw the potential in Coralogix. Moving forward, they plan to run an additional batch of training with another one of their development partners, as well as some of their regional business teams.

With the platform’s ease of use, the team hopes to onboard non-developer operations staff on the frontline in managing these websites and reporting issues. Coralogix gives them the ability to investigate issues on their own and approach the right person with the contextual information needed to quickly resolve the issue.

“[The onboarding] was about two months, and given the custom development that the Coralogix team put in during that time, I think it was a very good time frame.”
– Michael Gaskin, Senior DevOps Manager at PUMA

CDN Monitoring

Due to their volume, Fastly logs used to be sent to Puma’s cloud archive storage without an efficient way to query them. Now, Fastly logs are ingested to Coralogix for holistic monitoring with Logs2Metrics capabilities, alerting, and more.

With Coralogix, all Fastly logs are parsed, enriched, and sent to their archive storage where they can be queried directly from the Coralogix UI. Coralogix offers unparalleled access to archived data with the ability to query the data using multiple syntaxes and provide up to 72 hours of data in just a couple of minutes.

They are then directed down different data pipelines according to use case. Only Error and Critical-level logs are indexed to provide the team with lightning-fast queries. Info-level logs are kept in the Monitoring pipeline for monitoring and alerting – without needing to index them.

This allows for substantial cost savings while giving the team full flexibility to leverage all of their data at whatever granularity they need. With all data being stored in their own cloud archive storage and with powerful query abilities, the team essentially enjoys infinite retention.

Cloud-Native Monitoring

All the data from PUMA’s frontend, application and service layers running in GCP is now centralized and can be correlated with their Salesforce data. This gives a holistic view of their system behavior and health, and provides real-time insights.

Now, if a customer sees a blank screen when they visit the site or orders are failing, the team only has to look in Coralogix and can pinpoint the root cause using intuitive filters, ML-generated log templates, and more. It’s been perfect for troubleshooting issues raised by the developers, not only in production but in earlier development and staging environments as well.

For example, when the lead mobile app developer tested the credit card order in the UK in their staging environment, he kept getting an error response. Upon searching Coralogix over that time, they could immediately see that a fraud service was blocking the order from going through because of a required property missing in that environment.

In just a few minutes, they had an answer. The ease with which the team can pinpoint issues in their system and the power of Coralogix’s proactive alerting has sped up the development process by helping the team find the root cause of issues faster.

Coralogix’s pricing model makes it possible. CDN logs have always been considered too noisy to do anything with. It’s too expensive to ingest them into any traditional log analysis platform. With Coralogix’s unique architecture supporting a new approach to observability pricing, it’s possible to tap into these sources and control spending through the Total Cost of Ownership (TCO) Optimizer.

During the initial setup of the TCO policies and exploration of their data in the platform, there was an effort to optimize how they handled GraphQL logs coming out of their serverless functions. Like most teams during on-boarding to the platform, PUMA was able to optimize how their data is handled once it’s ingested to Coralogix and how they were logging data in the first place.

By reducing the volume of noisy logs, they can now monitor their data without any cost concerns or barriers. Plus, they can afford to expand their data coverage to bring on whatever other data sources we might come across. Next, the team hopes to begin sending tracing data as they consolidate observability across the organization.

The ease with which the team can search across many different systems in the Explore UI improved observability for PUMA tremendously. Whether they’re looking for an error or wondering why an order didn’t go through, the team can easily investigate the root cause of issues for a quick resolution.

Real-time insights powered by Streamaⓒ, Coralogix’s in-stream analytics engine, enable the team to monitor and get proactive alerts so they can stay on top of issues when they have extensive sales and product launches. Using Logs2Metrics, the team created custom dashboards with a lot of detail about site operations without needing to index the raw log data.

Additionally, the partnership with Coralogix and the customer success team to get metrics set up, turn those metrics into dashboards, do the custom development for data integrations, has all made the difference. The organization can now move a lot faster when compared with solutions they used in the past.

Video | Cost-effective Observability: How to scale without the overages

Speakers: Chris Cooney, Developer Relations, Coralogix & Michael Gaskin, Senior DevOps Manager, Puma

The big barrier to success with logs, metrics and traces, is scale. Scale impacts performance, cost, usability and more. At a talk delivered at an industry event in 2023, Puma and Coralogix explored what is driving the need to scale, and how to organize and architect your data to get the most out of it, while driving down costs and maintaining performance.