Ross says one thing, the warehouse says another

You run Ross for the things Ross is the system of record for — orders, finance, lot tracking, compliance — and a separate WMS for the floor, because the WMS is genuinely better at the floor. Reasonable setup. The problem is that the two never quite agree. The inventory in Ross doesn't match what's actually in the building, and every morning someone with a spreadsheet and a good attitude reconciles the gap by hand, because the alternative is not knowing what you have.

That daily reconciliation is the tell. Ross was never designed to hand off cleanly to a modern WMS, so the two drift — and in a process manufacturer, drift between the floor and the system of record isn't a tidiness issue. It's a compliance exposure and a cash problem wearing a calm face. The good news: getting them to agree in real time is a solved problem. It's just not an easy one, and "not easy" is exactly why it keeps not getting done.

Why this specific integration is hard

WMS-to-ERP is a hard integration everywhere. WMS-to-Ross has its own additional flavor, and it's worth being honest about it.

Ross carries a lot of history in its schema, and process manufacturing piles complications on top: lots and sub-lots, unit-of-measure conversions that have to be exactly right or your inventory value quietly goes wrong, quality holds that change whether stock is even allowed to move. Mapping the WMS's tidy view of physical reality onto Ross's — and back — means understanding not just where the data lives but what it means on each side, including the cases where the two systems have a genuine difference of opinion about the same physical pallet. This is where DML and a real understanding of Ross's model stop being optional. You can't paper over the edges, because the edges are where process manufacturing lives.

Inventory is the flashpoint

If this integration is going to embarrass you, it'll do it through inventory.

Relief timing is the classic one: the WMS relieves stock when the pallet physically moves; Ross relieves it when the transaction posts; and in the hours between, the two numbers disagree and someone promises product that's already spoken for. Unit-of-measure conversions are the quiet assassin — pounds to cases to pallets, and if the math is off by a rounding rule, your on-hand and your valuation drift a little further apart every day until an auditor finds it for you. And quality holds have to be respected on both sides, or inventory that isn't cleared to move gets shipped anyway, which is the kind of mistake that turns into a phone call from a regulator.

Lot traceability, captured not reconstructed

The other thing that has to be right is traceability, and the difference between doing it well and doing it badly is when it happens.

Done well, lot and sub-lot relationships are captured as production happens — the raw lots that went into the finished lot, recorded in the moment, flowing into Ross as system of record. Done badly, traceability is reconstructed after the fact, from memory and spreadsheets, the week an auditor asks or the day a recall starts. One of those is a system. The other is a group project you never wanted to be assigned. In a food, supplement, or chemical plant, "we can trace any lot in either direction, on demand, through Ross" is not a nice-to-have — it's the thing the whole regulatory apparatus is checking for.

The discipline that makes it stick: reconciliation

Here's the design decision that separates an integration that holds from one that slowly rots. Don't trust the stream to be perfect.

A naive integration just pushes events between the systems and assumes they all landed correctly. They won't — one message in ten thousand will go missing, arrive twice, or land out of order, and process manufacturing does not grade on a curve. So the integration should treat reconciliation as a first-class feature: continuously verify that Ross and the WMS still agree, resolve the drift it safely can, and clearly surface the drift it can't instead of silently picking a number. A system that says "these two don't match, here's why, here's what I need" is worth ten that quietly guess and let you discover the guess during inventory. (This is one of the core integration archetypes — worth its own read.)

What "done" looks like

You'll know it's working because the morning reconciliation ritual disappears. Ross and the WMS agree in real time, across receiving, production, QC, and shipping, without a human standing between them holding a clipboard. Inventory in the system matches inventory in the building. Lot traceability is auditable through Ross, end to end, because it was captured correctly the first time. And the business stops depending on the one or two people who happen to understand both systems well enough to keep them in sync by hand — which is a lovely thing right up until one of them takes a vacation.

I've built exactly this on a cheese factory floor, where a line down costs money by the hour and "close enough" is not a number the FDA accepts. The shape of the outcome is always the same: the physical world and the system of record finally telling one story, at the same time, with nobody refereeing.

The bottom line

If Ross and your WMS disagree, you don't have a data-entry problem — you have a missing integration, and a person absorbing its absence every morning. It's harder than the WMS vendor implied and entirely doable with someone who knows Ross's model cold. Do it once, correctly, and you get your mornings, your accurate inventory, and your audit confidence back.

Reconciling Ross against your warehouse by hand every day? That's a solvable problem, and it's one I've solved on a live plant floor. Let's talk →. Related: the integration archetypes nobody names.