Build the store, or rent the boring parts of it
Custom is usually argued as control. Most of what a store does is identical everywhere, and paying to rebuild it is where store budgets quietly go.
Use a platform unless your catalogue, your pricing or your fulfilment cannot be expressed in one. Checkout, payments, tax handling and order management are commodity problems that never stop needing maintenance, and building them yourself buys ongoing work rather than advantage.
Separate the parts that differ from the parts that do not
A store is two things stacked together. Underneath is machinery every store needs: a cart, a payment integration, order records, refunds, tax rules, stock counts. On top is whatever makes your business unusual: how the catalogue is structured, how a price is calculated, what happens after an order is placed.
The underneath is where custom builds spend their budget and where they age fastest, because payment providers change, tax rules change, and fraud handling is never finished. The top is where a business actually differs, and it is usually the smaller piece.
When a platform genuinely cannot hold you
There are real cases, and they share a shape: the constraint sits in the data model rather than in the design. A catalogue where a product is configured rather than chosen. Pricing that depends on contract, volume or approval. Fulfilment split across branches with different stock and different lead times.
If you can state your constraint in one sentence and point at the platform concept it does not have, custom is the honest answer. If the sentence keeps turning into a list of preferences about layout, it is not.
- Configured products rather than selected variants
- Prices that depend on who is asking, not on what is being bought
- Approval or credit steps between order and fulfilment
- Stock that lives in several branches with different rules
The arrangement that usually wins
Keep the commodity machinery on a platform and build only your unusual part against it. Modern platforms expose their catalogue, cart and checkout through an interface, so the storefront can be entirely your own while payments and orders stay somebody else's problem.
This is not a compromise so much as a division of labour. You own what makes you money and you rent what would otherwise consume an engineer forever.
Cost over five years, not at signature
Platform fees are visible and recurring, which makes them easy to argue about. Custom costs are invisible and also recurring: the payment integration that needs updating, the tax change nobody scheduled, the person who has to be available when checkout fails on a Friday.
Compare the two over several years with maintenance included on both sides. A comparison that puts a build price against a monthly fee is comparing a purchase with a subscription and will always flatter the purchase.
What to take from this
- 01Most of a store is commodity machinery that never stops needing maintenance
- 02Custom is right when the constraint is in the data model, not the design
- 03A custom storefront on a platform back end is usually the correct split
- 04Compare over five years with maintenance on both sides
Ask the question directly
- Does a platform limit how the store can look?
- Not meaningfully any more. The storefront can be built from scratch against the platform's interface, so the visual limits people remember from older platforms are largely gone.
- What about transaction fees?
- They are a real cost and worth calculating at your actual volume. Weigh them against the engineering time that would otherwise maintain payments, which is a salary rather than a percentage.
- Can we move off a platform later?
- Products, customers and orders export cleanly. What does not move is the integration work and anything built with a platform-specific extension, so keep your own catalogue source of truth if a move is plausible.
Related answers
Web development
New builds, rescues and re-platforms, measured against what the site earns rather than against how it looks. The brand system that carries it is built in the same engagement.
Bring the decision you are stuck on
A call, forty five minutes, no deck. We will tell you if we are the wrong firm.