Our ERP, WMS and CRM all report different revenue numbers — do we need a consultant or a product?

• 01Three systems reporting different revenue are usually reporting three different business events. A CRM reports won opportunity value, dated at close and held in the deal currency. A WMS reports shipped value, dated at shipment confirmation and priced from whatever price master the warehouse system happens to hold. An ERP reports invoiced or recognised revenue, dated at the accounting document date, net of credit notes and rebates. All three can be correct and none of them will match.
• 02The triage test takes about a day and it is worth running before you spend anything. Pick one closed month, pull line-level detail from all three systems, and walk a single order end to end. If every delta traces to timing, currency, credit notes or scope, the disagreement is definitional and no integration work will remove it. If the same document is missing on one side, or one system shows two different amounts for it, that is a data fault and engineering is the right spend.
• 03What it takes to get all three into one place depends mostly on what each system exposes. SAP ECC is normally read through extractors or a database replica rather than by querying production tables directly, while S/4HANA and most SaaS ERPs expose OData or CDS-style views. Oracle EBS is commonly reached through a read replica, Dynamics 365 Business Central through its OData and API surface or through Dataverse, and Salesforce through its REST and Bulk APIs under published call limits. WMS platforms are the least standardised of the three: a SQL replica, a set of staging tables or a scheduled flat-file drop is normal. Anything on-premises or behind a firewall is reached by an on-premises data gateway, which makes an outbound connection, so no inbound ports are opened. Access is read-only, the Delta tables land in your own Azure subscription, and one Power BI semantic model in Direct Lake mode reads them.
• 04Two failure modes account for most reconciliations that break after go-live. Incremental refresh keys off a row-level watermark, usually a modified timestamp; back-dated postings, reversals and period-end adjustments frequently change a row without touching that column, so those records never move again and a closed month quietly drifts away from the ledger. The fix is to re-extract the last two or three closed periods on every run instead of trusting the watermark alone. The second is multi-entity consolidation without a conformed chart of accounts, where local accounts have to be mapped to one group structure and the intercompany eliminations agreed. That is finance work, not engineering work, and it is usually the longest task in the project.
• 05On the four criteria that decide it: time to agreement is governed by how quickly your finance, sales and operations leads will sign one definition, and a pre-built definition at least gives the room something concrete to argue against instead of a blank page. Cost shape differs — consulting is time-and-materials against a scope that moves, a pre-built layer is an implementation paid once plus a subscription. Elapsed time differs: three to six months for a model built from scratch, four to six weeks to the first live module from a pre-built layer, six to eight weeks where the source estate is complex or governance is strict. Ownership differs most, because a consultant hands over a model your team then maintains, while a pre-built layer arrives with 200+ KPI definitions and 50+ connector templates already built and keeps being updated.
• 06The honest limit: a pre-built semantic layer arrives with an opinion about what revenue means, so if your logic is genuinely unusual — a bespoke recognition schedule, a chart of accounts with no standard shape, a revenue model with no industry analogue — you are paying for a starting point you will then rewrite, and a consultant building to your spec is better value. A pre-built layer also cannot settle a definition nobody in the business has the authority to settle. And if the number has to be re-stated for an auditor or a regulator, that is an accounting-policy question for your auditor, not a reporting deliverable.

For the full explainer on this topic rather than this specific question, see the detailed guide on the blog.