The most expensive sentence in e-commerce is: "We'll figure out the back office once the store is live."
I have rebuilt enough digital stacks to know how this story ends. The brand launches a beautiful Shopify or Magento store, sales come in, and within 90 days the ops team is buried in spreadsheets, the accountant cannot reconcile, inventory is wrong on every channel, and the founder is paying an agency to integrate seven SaaS tools that should have been one system.
The fix is to invert the build order. ERP first. Storefront last.
The reference architecture
Customer → Storefront (Shopify / Magento)
│
▼
Integration Layer (Make / native connectors)
│
▼
ERP (Odoo / NetSuite) ◄── single source of truth
/ | | \
Inventory Orders Finance Customers
\ | | /
▼
Channels (POS, marketplaces, B2B, wholesale)
The storefront is just one channel. The ERP is the operating system.
What lives where
| Concern | Storefront | ERP |
|---|---|---|
| Product catalog (master) | ❌ | ✅ |
| Storefront-specific copy & media | ✅ | ❌ |
| Pricing rules | ❌ | ✅ |
| Promotions | ✅ for storefront-only | ✅ for cross-channel |
| Inventory | ❌ (read-only) | ✅ |
| Orders (capture) | ✅ | ❌ |
| Orders (fulfillment, accounting) | ❌ | ✅ |
| Customers | ✅ for storefront accounts | ✅ as system of record |
If you put any of the "ERP" rows on the storefront, you will pay for it within a year.
The four integration contracts
Whatever ERP you pick, you need four contracts in place on day one:
- Product sync. ERP → storefront, one-way. The storefront never edits the master record.
- Inventory sync. ERP → all channels, near real-time. Sub-minute latency for hot SKUs.
- Order sync. Storefront → ERP, with idempotency keys. Every order must land exactly once.
- Customer sync. Bi-directional, with ERP as the merge authority on conflict.
Design these as queues, not direct API calls. A queue you can replay; a direct call you cannot.
The 90-day deployment cadence
This is the cadence I use for ERP-first transformations:
- Weeks 1–2: Process mapping. Document every flow from quote to cash to refund.
- Weeks 3–6: ERP configured against the documented flows. Real data, real users, no storefront yet.
- Weeks 7–8: Storefront cutover plan, integration contracts implemented, dry-run with shadow orders.
- Weeks 9–12: Cutover, with the old storefront still alive as a rollback path for 14 days.
That is it. Twelve weeks to an operating system that will outlast three storefront redesigns.
The payoff
Brands that run this architecture share four traits:
- They can change storefront platforms in a weekend, not a quarter.
- They never have an "inventory accuracy" project — accuracy is structural.
- They can open a new channel (POS, marketplace, B2B portal) in two weeks.
- Their CFO trusts the numbers without reconciliation calls.
You are not building a website. You are building a revenue operating system. The website is just the surface.