We’ve really been able to simplify in terms of management, so our team of eight is capable of solving any network issue that comes up, 24/7.
Rai Umair CTO, Daraz

As the sixth-most-populated country in the world, Pakistan presents a large customer base for retail operations. Launched in 2012, Daraz offers a diverse assortment of products to online shoppers in Pakistan and four additional South Asian markets: Bangladesh, Myanmar, Nepal, and Sri Lanka. The e-commerce provider is focused on providing a superior customer experience, including a user-friendly interface and comprehensive customer care. With 2 million products, 30,000 sellers, and 5 million customers, Daraz has become a household name in the region.

In November 2016, during the busiest sales campaign of the year—Black Friday—the unthinkable occurred: Daraz’s backend systems crashed. The Daraz mobile app was extremely slow, even unresponsive, and customers could not view product images. Since launching, Daraz had maintained a hybrid IT network, running its applications virtually through a third-party data center. The company had leveraged two third-party providers, a Frankfurt company for customized network hosting and a Portuguese provider for development and infrastructure support. When the Frankfurt company’s routers went down in 2016, Daraz began evaluating cloud-based alternatives. The company considered several providers but found Amazon Web Services (AWS) the most mature in terms of breadth of services offered, and it also liked the fact that AWS can be configured for multiple availability zones. Two additional network glitches interrupted the Daraz website in early 2017, prompting its management to urgently migrate its primary workloads to AWS before the next annual Black Friday campaign.

Rai Umair, chief technology officer at Daraz, listed three goals for moving to the AWS Cloud. First and foremost was stability. “Especially during campaign periods, our site cannot be down for even a minute,” he says. Stability was also essential during the migration process; Daraz could migrate its customer-facing workloads for all its five markets onto AWS in just four months, with little or no downtime. The company measures performance based on customer orders, with no significant decrease reported during the four-month migration period.

Simplicity was the next goal. Daraz wanted more control over its network. Umair started with a team of four, which has since doubled in size. Most employees had little, if any, experience with AWS, but they were able to experiment and learn on the job, taking advantage of the AWS Business Support plan as needed. “We’ve really been able to simplify in terms of management, so our team of eight is capable of solving any network issue that comes up, 24/7,” Umair explains.

Flexibility was the final goal. During campaigns such as Black Friday, Daraz experiences rapid spikes in traffic, from 200,000 requests per second—the average on any given day—to more than a million during campaign periods. Umair expects site traffic to continue rising exponentially. Daraz also needs to quickly scale down its infrastructure after big campaigns to save costs, which it can now do with just one click, by turning off inactive AWS instances.

Daraz uses Amazon Elastic Compute Cloud (Amazon EC2) instances of various sizes for all workloads: the Daraz online mall, its order management system, and business intelligence (BI). It also employs Elastic Load Balancing to distribute incoming traffic across Amazon EC2 instances and multiple availability zones. Distributing workloads across separate availability zones also ensures redundancy and stability. Should servers in one zone go down, Daraz’s web traffic is rerouted to a secondary zone, with no interruption to site functionality. In addition, the company relies on Amazon CloudWatch to monitor servers and perform minute-by-minute reporting.

Umair and his team have simplified day-to-day operations with Amazon Relational Database Service (Amazon RDS), using it to run the open source database MariaDB. “Prior to this, we were running our databases directly on our own managed boxes, taking a lot of time to set up redundancy and take backups. We definitely saved a lot of time and have reduced complexity by moving to Amazon RDS,” says Umair. “We have many replicas, and by using Amazon RDS, it's easy to set them up. The complexity is hidden behind the servers.” Automated backups from Amazon RDS are then stored in Amazon Simple Storage Service (Amazon S3).

As a customer-centric company, Daraz always seeks to improve its customer experience. Prior to using AWS, nearly half of all customer calls received were simple inquiries about order status, which Daraz wanted to expedite using automation. Faster calls would enable each customer service agent to handle more calls each day, which would boost productivity. Experience had also shown that the less time a customer has to wait to hear their order status, the happier they are with the call. The company now uses Amazon Polly to deliver automated real-time information when customers check on their order status. Edouard Gheerbrant, head of customer experience, says, “We were able to shorten the duration of such calls by more than 40 percent.” The majority of these calls are now handled in 15 seconds, Gheerbrant explains, “which has had a positive impact on customer service without increasing the number of agents.”

After each call, Daraz customers are asked to complete a survey evaluating how satisfied they are with the experience. “Today, we’re standing on the verge of 4.8 out of a total score of 5, and before we introduced Amazon Polly we were at about 3.5,” Gheerbrant says. Looking ahead, he is considering using Amazon Lex to develop a chatbot that would automate up to 80 percent of all calls, beyond just providing order status. He is also now able to train his agents on more complex issues that would be the subject of the remaining 20 percent of calls.

Since moving to AWS, Umair and his team are able to design more monitoring and BI applications. He can manage all IT issues with just eight engineers because they no longer have to worry about setting up data centers or brokering power and bandwidth agreements. The team has used AWS Support to work through any issues and will continue to do so in the future.

“AWS Support has been fantastic,” Umair says. “The responses are so detailed. When we faced difficulties because we had no formal AWS training, they responded with exactly what must be done and what the real issue was. There were many instances where we were addressing the symptoms of a problem without understanding why it was happening. Once we had an issue with managing container images and spent a week trying to identify the cause, but AWS was able to help pinpoint root cause and resolve it quickly.”

Learn more about AWS Cloud Solutions.