ask someone to picture an e-commerce project and they picture the shop: catalog, cart, a nice checkout button. that is the smallest part of the system.
behind the button sits a chain: payment provider, carriers, erp, crm. each link is a different vendor, a different api, a different support queue. the storefront takes a month. the chain takes the rest of the project, and it is where launch dates go to die.
the failure mode is almost never code. it is a handoff. the psp passes its sandbox tests, the carrier api is documented, and then the erp exports orders in a format nobody wrote down - a week before launch. i once watched a payments brief specify the solution before anyone checked the need. we challenged the requirement, built the business case for a different option, and the client changed the brief before build. that version shipped, and it made money for the agency. challenging a brief is cheap. rebuilding after launch is not.
so here is the chain, link by link.
swipe the diagram sideways on small screens.
- storefront to psp - authorize, capture, refund
- psp to storefront - webhook with paid, failed and refunded events
- storefront to erp - orders
- erp to storefront - stock, prices, order status
- storefront to crm - customers and orders
- crm to storefront - segments and consent
- storefront to carriers - rates and shipments
- carriers to storefront - webhook with tracking events
the storefront
the part everyone argues about, and the part that matters least. prestashop, woocommerce, shopify - all of them render a catalog, a cart and a checkout. pick the one your team can actually maintain. the interesting questions start at its api: how products, stock and orders get in and out.
reference: woocommerce rest api documentation ↗
read more: how to choose an e-commerce platform: prestashop, woocommerce or shopify →
the payment provider
the psp moves the money and carries most of the compliance weight, but not all of it. stripe, adyen, przelewy24 - the integration is never just a pay button. it is authorization, capture, refunds, webhooks and reconciliation.
the order is paid when the psp's webhook says so, not when the customer clicks pay. projects slip because someone treats a green sandbox as done and discovers refund handling a week before launch.
reference: stripe webhooks documentation ↗
read more: e-commerce payment integration: webhooks, refunds and reconciliation →
the carriers
the carrier turns an order into a parcel. dhl, ups, inpost - you integrate rates, labels, tracking and pickup points, and every carrier describes the same concepts in a different api.
the classic failure: the checkout promises a delivery option the warehouse can't fulfill that day, or a pickup-point list that stopped updating in march. test with real labels and real addresses.
reference: dhl api developer portal ↗
read more: e-commerce shipping integration: carriers, labels, tracking and pickup points →
the erp
the erp is the source of truth the business runs on: stock, prices, orders, invoices. usually sap s/4hana or one of its cousins. the storefront displays stock; the erp owns it. if the sync runs nightly, by noon you are selling products that left the warehouse at nine.
agree early which system wins on every field. 'we will figure out the mapping later' is how you get an order export in a format nobody wrote down.
read more: e-commerce erp integration: stock sync, order export and field ownership →
the crm
the crm knows the customer: segments, consent, history. salesforce, hubspot, sap c4c. the shop pulls segments for promotions and sends orders back. both directions need an owner per field, or marketing emails a customer about the order that was just refunded.
reference: salesforce developer documentation ↗
read more: e-commerce crm integration: customer data, consent and segments →
so the rule i work by: map the integrations before you pick a theme. every system, its owner, its test environment, its failure mode. that map is the real project plan. the storefront is just the part customers are allowed to see.
and test the whole chain early, end to end, with real orders. a green sandbox badge is a screenshot, not a launch.
e-commerce architecture is mostly dependency management with a checkout button on top. plan accordingly.