When is MACH architecture a bad idea for ecommerce?
The costliest IT mistake is building for hypothetical scale that sales cannot back. GMI Software advises against MACH when the goal is a fast experiment on a small catalogue without ERP integration or a serious omnichannel plan.
What problem does MACH actually solve?
MACH earns its keep when you must iterate on shopping UX, wire multiple channels (web, app, marketplaces) and backends (ERP, WMS, PIM) without freezing the whole platform. Without that cadence you pay operational complexity with no return.
In practice you need a team that can run CI/CD, observability and API contracts. Without that discipline even solid MedusaJS v2 code struggles in the first promo season.
When are Shopify, Shoper or IdoSell enough?
If you sell finished goods with simple tax rules, skip contract B2B pricing and do not need a mobile app sharing the same cart as web, SaaS often wins on time-to-market and predictable fixed fees.
At high scale SaaS take rates or API ceilings can bite. Our TCO ebook on gmi.software walks through the break-even versus MedusaJS headless - run the numbers before renewing a three-year SaaS deal.
Can a small IT team run MACH?
Yes, with a production partner such as GMI Software and clear SLAs and runbooks. It fails when one generalist owns everything and PTO blocks releases. Prefer a simpler stack or well-supported SaaS.
Staff augmentation without a strong product owner on the client side often ends with purchased hours but no decisions.
Which signals mean MACH in 12-18 months?
Rising tickets that start with “the template cannot…”, ERP bridges built on spreadsheets, mobile conversion drops despite traffic, and pressure for dedicated B2B partner apps are common thresholds.
When they appear, rerun DDT: domain map, current stack TCO and a strangler-style migration plan.
Does “MACH first, business idea later” ever work?
Rarely for mid-market firms. Well-funded startups sometimes pre-build for scale, but we see more wins when a simpler first version proves revenue and MACH follows proven PMF.
GMI Software prefers engagements where the client names three KPIs the new stack must move (B2B order cycle time, oversell rate, integration maintenance cost).
What should you do instead of MACH today?
Limit tech debt: clean module boundaries even inside a monolith, regression-test checkout, monitor core paths. Prepare product and pricing data as if headless arrives tomorrow to cut future migration cost.
If you are unsure where to start, book a consultation with GMI Software - we often return within 48 hours with guidance on whether to move toward MedusaJS v2 or first tidy the SaaS footprint.
Frequently asked questions
- Does GMI Software ever say no to MACH projects?
- Yes, when product ownership, data or KPIs are missing. We then propose a lighter discovery phase or SaaS hardening instead.
- Does MACH pay off at 50 SKUs?
- Rarely, unless those SKUs hide heavy configuration, CPQ or manufacturing integration. For a simple retail catalogue SaaS usually wins on a 24-month TCO view.
- How long does a typical SaaS stack last before migration?
- It depends on business change velocity: stable models may stay on SaaS for five to seven years, fast-moving B2B may need MACH after 18-30 months. Track integration tickets, not the calendar.
- Will migrating later destroy customer data?
- Not if you keep order IDs and consent records GDPR-clean and maintain API exports from day one. Cleaner identifiers early mean cheaper MedusaJS migrations later.
- Where can I read about SaaS TCO vs headless?
- gmi.software publishes five-year platform cost materials; in sales we tailor the model to your transaction volume and integration count.
Content updated: April 6, 2026