Custom AWS commerce platforms for luxury brands that have outgrown Shopify Plus.

No. If the platform fits the workflow and the all-in fee math is acceptable, do not migrate for ideology.

The math changes when transaction size, operational exceptions, checkout constraints, payment economics, plugin marketplace dependencies, and data ownership make a SaaS commerce stack more expensive than a custom platform. The threshold is somewhere around $20K average order value, depending on order velocity. Below that, Shopify usually still wins on time-to-market. Above that, the fee math and workflow fit start breaking down fast.

More than the headline fee. The honest number is card processing (2.9% plus $0.30) plus platform-side transaction fees, required app subscriptions, tax-engine fees, fraud-protection fees, refund and chargeback fees, and a fixed monthly platform subscription divided across a small number of high-value orders.

On a $50K high-ticket order, the headline card fee alone is $1,450. With platform fees, required apps, and amortized subscription costs included, the all-in cost per revenue dollar can climb well past 10% on a low-velocity high-ticket seller. Across a year of high-ticket commerce, that is a six-figure tax on top-line revenue. The fix is a custom payment workflow engineered to your order-value distribution and operational realities. For the Fine's Gallery anchor client, that workflow settles balances via Stripe ACH with card accepted only as a $500-max non-refundable deposit. The right model for your business may look different; it is engineered, not prescribed.

Yes, and it is usually the safest path. Parallel operation, staged data migration with audit reports, reconciliation gates, and a controlled cutover after the new platform's core workflows are proven against real orders. Production cutover happens against a production-clone rehearsal first, with rollback criteria defined in advance.

The Fine's Gallery cutover ran with zero operational downtime and zero loss across 20+ years of legacy operational data.

A Shopify agency optimizes inside Shopify's model: themes, apps, checkout extensibility, third-party plugin selection. The agency's economic incentive is to keep you on Shopify because the agency's expertise is Shopify-specific.

This work starts when the model itself is the constraint. Custom checkout, custom payment flows (ACH / wire / check), custom order state machine, sovereign AWS infrastructure, event-driven async backbone, owned integration boundaries. The economic incentive is to retire SaaS layers that no longer pay for themselves at this order velocity, not to keep adding plugin subscriptions.

Staff workflow is treated as a first-class requirement, not cleanup after launch. The safest migrations preserve familiar operating patterns where they matter (record-oriented found-set navigation, sequential invoice numbers, the way staff actually edit a quote) while removing the platform limits underneath them.

On the Fine's engagement, the legacy invoicing system staff had used for 20+ years was replaced with a platform-native invoicing surface inside Payload that staff could operate the same way they always had, with all 28,000+ historical records migrated cleanly into the new ledger.

No. The goal is to own the revenue engine, the operational truth, and the AWS infrastructure underneath them. SaaS tools that still do their job stay connected (Stripe, DocuSign, Klaviyo, Google Merchant Center, GA4) via first-party typed adapters with Zod validation, retry semantics, and dedicated tests. There is no plugin marketplace dependency.

SaaS layers that no longer pay for themselves at this order velocity (heavy commerce platform fees, redundant marketing apps, parallel back-office tooling staff has to alt-tab to) are systematically retired.

Roughly three to six months for the initial production cutover, depending on integration scope, data migration complexity, and the number of operational workflows being absorbed. The Fine's Gallery platform shipped its first production cutover in that timeframe and has continued to evolve through a Fractional CTO retainer since.

Platform builds from $80K. Final scope depends on integration count, payment workflow complexity, legacy data migration, and the staff operating surface's depth. Discovery happens in a paid Architecture Sprint ($15K, 3 weeks) so the build estimate is grounded in your actual revenue workflow, not a generic intake form. Architecture Sprint output is a written architecture decision document — current-state risk register, target architecture, migration plan, and cost model — that you can run the build against with anyone, including yourself.

Solo delivery by Peter T. Conti, AWS Solutions Architect Professional + Associate, lead architect on a national-scale energy-sector certificate registry and the sole engineer behind the Fine's Gallery commerce platform.