/Industry · E-commerce and retail
AWS for e-commerce: a checkout that holds up on Black Friday
Campaigns without outages, ERP and logistics integrations that never fail silently, and a bill that drops once the campaign is over.
Problems we solve in E-commerce and Retail
Checkout goes down mid-campaign
Load tests, autoscaling, CloudFront and a virtual waiting room.
Inventory and orders stuck between systems
EventBridge, SQS and Step Functions with retries.
Bots testing stolen cards
WAF with Bot Control, rate limits on login and checkout, and tokenization.
Peak capacity left running
Scale-down after the peak and cost per order in plain view.
Regulations:PCI DSSLocal data protection lawsGDPRISO 27001
Services for E-commerce and Retail
More about E-commerce and Retail on AWS
An online store on AWS ready for Black Friday and Cyber Monday
In e-commerce, traffic is never evenly spread: it piles up around campaigns, launches and dates like Black Friday or Cyber Monday. We design the store for that pattern. Static content and product pages are served from CloudFront, sessions and carts live in ElastiCache or DynamoDB, and catalog, inventory and checkout services run on containers or Lambda with autoscaling that is configured and tested.
Weeks before the campaign we run load tests with a realistic profile: homepage traffic spikes, concentrated searches, full carts and simultaneous payments. We review AWS service quotas, set up waiting rooms so shoppers never see errors when too many arrive at once, and leave an on-call plan with alarms and clear owners for the biggest sales day.
Catalog, search and integrations with ERP, logistics and payments
With thousands or millions of products, catalog and search have to respond fast even as prices and inventory change all day. We index the catalog in Amazon OpenSearch Service with filters, synonyms and tuned relevance, separate reads from writes and cache whatever repeats. That way a bulk price update never slows down browsing for your customers.
Integrations with the ERP, the WMS, logistics carriers and payment providers move to an event-driven model with Amazon EventBridge, SQS queues and Step Functions workflows. Every message is processed with retries, idempotency and dead-letter queues, and failures show up on CloudWatch dashboards. If an external system goes down, orders wait in the queue instead of getting lost.
Checkout security: PCI DSS, fraud and bots
The most effective way to meet PCI DSS in an online store is for your platform never to touch card numbers. We use the payment processor's tokenization, isolate the services involved in payment in dedicated accounts and networks, encrypt personal data with AWS KMS, and record access and changes in CloudTrail, with AWS Config and Security Hub generating continuous evidence.
Bots are part of every e-commerce site's traffic: they hoard deals, scrape prices and test stolen cards. We put AWS WAF with Bot Control and Fraud Control in front of the store, rate-limit requests on login, cart and checkout, and protect accounts with Amazon Cognito or your identity provider. Rules are tuned with real data so legitimate customers are never blocked.
Generative AI in e-commerce, with costs under control
Generative AI adds value when it solves concrete store problems: recommending products based on what the customer is looking for, answering questions about shipping, returns and sizing, and writing product descriptions from catalog data. We take it to production with Amazon Bedrock and models like Claude, with data kept in your account, guardrails, response evaluation and cost control per invocation.
After every campaign, capacity has to return to its normal size. We configure scale-down, Savings Plans for the stable baseline, Spot or Graviton instances where they fit and log retention based on value. With tags and cost-per-order or cost-per-channel dashboards, the team can see whether a campaign was profitable on the infrastructure side too, and blue/green or canary deployments let you keep shipping changes without interrupting sales.
/FAQ
Frequently asked questions
Another question? Write to hello@nubent.com and an engineer will answer.
How do you prepare an e-commerce site on AWS for Black Friday or Cyber Monday?
With realistic load tests several weeks ahead, a review of AWS service quotas, autoscaling that is configured and tested, CloudFront caching for everything that can be served from the edge, waiting rooms to absorb the peak and an on-call plan with alarms and owners. It is also wise to freeze major changes in the days leading up to the campaign.
Does an online store on AWS need to be PCI DSS compliant?
If it processes, stores or transmits card data, yes. The simplest route is to use a payment processor's tokenization so your platform never touches the card number, which shrinks scope dramatically. AWS covers its part of the infrastructure, but network configuration, access, encryption and logging remain your responsibility.
How do you stop bots and fraud on an e-commerce site with AWS?
With AWS WAF in front of the store, adding Bot Control to detect automated traffic and Fraud Control to protect login and account creation, rate limits on cart and checkout, and rules tuned with real data. On the payment side, combine it with your processor's anti-fraud tools so legitimate customers are not turned away.
What is generative AI useful for in an online store?
The most direct uses are assistants that answer customer questions, natural language search, product recommendations and descriptions generated from the catalog. With Amazon Bedrock your data stays in your account, guardrails apply and cost per invocation is measured. Start with a narrow use case and measure its effect on conversion.
How do you lower AWS costs after a campaign?
By scaling down the capacity added for the peak, covering the stable baseline with Savings Plans, using Spot or Graviton instances where they fit and setting retention for logs and snapshots. Tagging resources by campaign and service shows you what was left running and lets you measure cost per order.
Next step
What should we build next on AWS?
30 minutes with an engineer. Concrete next steps, whether or not you work with us.