The symptom is rarely the problem
Teams usually call about one visible failure: orders arriving in the ERP a day late, stock levels that disagree between the warehouse and the storefront, a finance close that takes a week because someone is reconciling exports by hand.
Underneath, the cause is almost always the same. The integration was built to move records between two systems, but nobody defined what should happen when the two systems disagree. Without an answer to that, every edge case becomes a manual intervention, and the volume of interventions grows with the business.
What we actually build
The flows below are the standard shape of an order-to-cash integration. Which ones you need depends on where your manual work currently sits.
- Order downstream: checkout to ERP sales order, including tax, discounts, gift cards and multi-currency handling
- Fulfillment upstream: shipment, tracking and partial-shipment events back to the platform and the customer
- Inventory: available-to-promise across warehouses, 3PLs and marketplaces, with a safety-stock model that reflects real lead times
- Product and pricing: cost, price lists, customer-specific and contract B2B pricing
- Customer and account: B2B account hierarchies, credit terms and payment status
- Returns and credits: RMA through to credit memo, which is where most integrations quietly stop
- Financial reconciliation: payouts, fees and settlement matched against orders
Failure modes we design against
An integration is defined by how it behaves on its worst day. These are the conditions we test before go-live, because each one has taken a real business offline.
- Duplicate orders when a webhook is redelivered — solved with idempotency keys, not with deduplication scripts
- Silent failure, where a queue backs up and nobody learns until finance notices a gap
- Oversell during a flash sale, when inventory updates cannot keep pace with checkout
- Peak-season throughput exceeding ERP API governance limits, which requires queueing and backpressure rather than retries
- Partial fulfillment and split shipments, which many off-the-shelf connectors model incorrectly
- Drift, where both systems believe they are correct — caught by scheduled reconciliation, not by inspection
ERP systems we work across
NetSuite, Microsoft Dynamics 365 Business Central and Finance & Operations, Acumatica, Sage Intacct and Sage X3, Infor, Epicor, Odoo, and the ecommerce-oriented operations platforms Brightpearl, Cin7 and Fulfil.
We are not a reseller for any of them and receive no commission, which is what allows us to recommend keeping the system you have.
What it is worth
The business case is normally built from three numbers: hours per week currently spent moving or correcting data, the cost of the errors that reach customers, and the delay between a sale and a reconciled financial record.
We would rather quantify those with you before proposing scope. If the numbers do not justify the work, that is a useful answer and it costs you nothing to reach it.
Common questions
- Should ERP or the commerce platform be the system of record?
- Split it by domain rather than picking one. ERP is normally the record for financials, cost, purchasing and inventory ownership. The commerce platform is the record for the customer-facing catalog, sessions and checkout. Problems appear when both claim the same field, most often price and available-to-promise, so the contract has to name an owner for each field explicitly.
- Real-time or scheduled synchronization?
- By data type. Orders and payment events should be near real-time because delay becomes a customer service problem. Inventory is usually event-driven with a periodic full reconciliation to catch drift. Product and cost data is normally batched, since it changes less and the volume is higher. Making everything real-time is expensive and increases the blast radius when one system is slow.
- How long does an ERP integration take?
- For a single-region business on a mainstream ERP, a first production flow in six to ten weeks is realistic, with the full order-to-cash cycle taking a quarter or more. Multi-entity, multi-currency or multi-warehouse operations take longer, and the schedule is usually driven by finance's reconciliation requirements rather than by engineering.
- Can we keep the connector we already bought?
- Often yes. Off-the-shelf connectors handle standard flows competently and we have no commercial reason to replace one that works. They tend to fail on bundles, kits, partial fulfillments, multi-warehouse allocation and returns, so those are what we test before recommending a change.