End-to-end commerce architecture processing 10,000 orders/day with 99.9% uptime

Mars Rejects had cracked the hardest part of the fashion business: building cultural relevance. Their limited drops consistently sold out, their Instagram following had grown to 600,000, and their collaborations with artists and musicians were generating national press coverage. The brand had momentum that most fashion founders spend years trying to create. What they hadn't built was the infrastructure to monetise that momentum without it collapsing under its own weight.

During their most anticipated drop — a collaboration with a well-known Mumbai artist — their Shopify store received 15,000 concurrent sessions within 60 seconds of launch. The storefront went down completely within four minutes. The checkout was inaccessible for 2.3 hours. An estimated ₹45 lakh in sales was lost in that single afternoon, and the brand's social media was flooded with angry customers. The post-mortem revealed three critical failure points: inadequate server capacity, a slow database query pattern under concurrent writes, and a payment gateway timeout configuration that was killing sessions rather than queuing them.

Beyond the drop infrastructure problem, Mars Rejects had no retention marketing whatsoever. They were acquiring customers through organic hype and paid social, but making no systematic effort to bring buyers back for subsequent purchases. Fulfilment was being managed from a shared spreadsheet that broke down above 200 orders per day. The business was generating genuine demand but leaving significant revenue on the table at every stage of the customer lifecycle.

We conducted a full architecture review over two weeks and presented the founding team with a three-part engagement: infrastructure rebuild for drop-day resilience, fulfilment automation, and a retention marketing programme. All three were sequenced carefully — the infrastructure work came first, as another failed drop would have been brand-damaging at a critical growth stage.

The infrastructure approach centred on a headless commerce architecture: Shopify remained the commerce engine and checkout, but the storefront was rebuilt as a Next.js application hosted on Vercel's edge network. This decoupling allowed the storefront to serve from Vercel's globally distributed CDN under any load while Shopify handled only authenticated checkout traffic — dramatically reducing the peak load on Shopify's servers during drop events. Drop-day queue management was built using a virtual waiting room system that metered customers into the checkout flow at a controlled rate.

The retention programme was built on Klaviyo and integrated with Shopify's customer data. We segmented the customer base by purchase frequency, average order value, and product category preference, and built automated flows for post-purchase engagement, cross-sell sequences, VIP recognition, and win-back campaigns for lapsed customers. Fulfilment automation via ShipBob API eliminated the spreadsheet dependency and provided real-time inventory visibility.

The headless commerce stack launched six months into the engagement. The first major drop post-relaunch handled 52,000 concurrent sessions with no performance degradation — a 3.5× increase over the previous record, at which the old stack had failed. All 4,200 items sold out in 11 minutes. The waiting room system queued 48,000 customers and metered them through checkout over 35 minutes, generating thousands of positive social posts about the experience rather than the frustration of a previous crash.

Fulfilment automation via ShipBob reduced order processing time from an average of 8 hours (manual) to 45 minutes from order placement to warehouse pick instruction. Returns processing was automated with a self-service portal, cutting customer service email volume by 40%. Real-time inventory syncing between Shopify and ShipBob eliminated the overselling incidents that had occurred twice in the previous year.

The retention programme, running for six months, achieved a repeat purchase rate of 38% — up from a baseline of 11% before any structured retention marketing existed. A VIP cohort of 2,200 customers — defined as those with three or more purchases and an average order value above ₹4,000 — were enrolled in an exclusive early access programme, giving them 30-minute pre-drop access. This cohort converted at 71% on drop day and drove 22% of total drop revenue despite being less than 5% of the active customer base.