Own the platform,
not the portal.

Your ERP vendor will offer to become your analytics platform. The pitch is real convenience today. The cost is a dependency your company — and its next owner — carries for years. There is a better pattern, and other industries settled it long ago.

The pitch you will hear

Embedded dashboards. Native analytics. One vendor, one contract, everything integrated. The demonstration is genuinely good, and the first reports arrive fast because the vendor controls both sides of the pipe.

State the condition plainly: when the analytics live inside the ERP platform, your measure definitions, report logic, history, and your team's skills accumulate in a system you rent. Then the consequence: every year of use raises the cost of ever changing ERPs, and the analytics — the layer that was supposed to give you leverage over the ERP — becomes the strongest reason you cannot leave it.

Why this is an exit-price problem, not an IT problem

A buyer values what it can verify and prices what it must fix. Analytics welded to the incumbent ERP hands diligence two findings at once: the numbers cannot be verified outside the vendor's portal, and any post-acquisition systems change — consolidation onto the buyer's ERP, a roll-up integration, even a version migration — means rebuilding the reporting layer from zero. Both findings move the price in the same direction.

The reverse also holds. A reporting layer that stands apart from the ERP, documented, in a mainstream environment, is an asset the buyer keeps. It shortens diligence and survives the integration. This is the same arithmetic as the exit math: infrastructure that makes the business verifiable is one of the few investments that supports the multiple instead of only the EBITDA line.

The pattern other industries settled

Industries that live or die on data — banking, retail, pharmaceuticals — converged on the same architecture years ago, and none of them let an operational system own the analytics:

Pattern	What it means	What it buys
Source-system independence	The ERP is one source feeding the platform — never the platform itself. Banks treat the core banking system this way; retailers treat the POS this way.	Change or add an ERP without rebuilding reporting.
Staged refinement	Raw extracts land untouched, are conformed to shared definitions, then published for reporting — the lakehouse "bronze / silver / gold" discipline.	Every number traceable back to the source transaction.
A semantic layer	Margin, on-time delivery, inventory turns — each defined once, in one governed place, used by every report.	The controller, COO, and board read the same number.
Master data	One customer list, one item master, one calendar — conformed across systems that each spell them differently.	Cross-plant and cross-ERP reporting that actually adds up.
Data products with owners	Each published dataset has a named owner, a definition sheet, and a refresh contract — not a mystery extract.	Reporting your own team can operate and extend.

None of this is exotic. In a 50–500 employee manufacturer, the whole pattern fits in the Microsoft environment you already license: Fabric holds the staged data and the semantic model, Power BI reads it, and the ERP goes back to doing its actual job — running transactions.

The cross-ERP dividend

The pattern pays twice for manufacturers because mixed ERPs are normal, not exceptional. One plant runs Syteline, an acquired shop runs Global Shop Solutions, a division still closes the month in something older. Embedded vendor analytics can only ever see its own system. A corporate platform conforms all of them to one model: one margin definition, one delivery measure, one board view across every plant — and the next acquisition becomes a new source to map, not a reporting rebuild.

AI belongs on the platform you own

Every AI capability worth having — asking questions of your numbers in plain language, anomaly flags on margin or scrap, forecast assistance — consumes governed, defined data. Run AI on the vendor's embedded analytics and you get the vendor's AI, on the vendor's terms, deepening the same dependency. Run it on your own semantic layer and the models inherit your definitions, your lineage, and your access controls — and switching AI providers is a configuration change, because the data never moved.

The ordering matters and is unforgiving: defined numbers first, trusted model second, AI third. AI pointed at unreconciled data automates the disagreement.

What to ask before the next renewal

• If we left this ERP in three years, what part of our reporting would survive?
• Can our controller see the query behind the number, or only the number?
• Where do our measure definitions live — in a document we own, or in the vendor's configuration?
• Could this analytics layer read from a second ERP if we acquired one?
• Who can export the full model — data, definitions, and logic — and what does it cost?

A platform you own answers all five without a meeting. An embedded portal answers none of them well — which is the answer.