When the stock figure depends on which screen you are looking at

Ask four people in the same business what the stock on hand is and you can get four answers. What is telling is that nobody is surprised. Everyone knows which system to believe for which question, and that knowledge lives entirely in their heads.

That is the actual state of a lot of otherwise well run operations. Not a broken system. Several systems, each holding part of the truth, and a workforce that has absorbed the discrepancies as normal.

How it gets to that point

It is almost never one bad decision. It is a sequence of individually sensible ones.

A business buys accounting software, because it needs to do accounts. Then production outgrows the spreadsheet, so someone buys a system for that. Then a customer requires despatch data in a particular format, so a third tool goes in to produce it. Each purchase solved the problem in front of it. None of them was asked to agree with the others.

The joins get made by people. Someone exports on a Monday and imports on a Tuesday. Someone re-keys the despatch numbers into the ledger. Someone maintains the spreadsheet that reconciles the two, and that spreadsheet becomes the most important system in the building without ever appearing on an asset register.

The costs that do not show up as an IT problem

The obvious cost is the time spent re-keying, and it is the smallest one.

Decisions get made slowly, or not at all. If the number needs a day of reconciliation before anyone trusts it, then questions that deserve a quick answer do not get asked.

The month end reveals things too late to act on. A margin problem found six weeks after the job finished is a history lesson.

Nobody can be new. A person who has been there eight years knows the system does not include consignment stock and adjusts accordingly. A person who started in March does not, and their error looks like carelessness rather than a design fault.

The reconciliation is a single point of failure with a name. Every business like this has someone who is the only one who really understands the joins. That is not their fault and it is a genuine risk.

Why “we need a new system” is usually the wrong first move

The instinct at this point is to replace everything with one platform, and sometimes that is right. But a replacement chosen before the disagreement is understood tends to reproduce it, because the mapping decisions get made under delivery pressure by whoever is available.

The question worth answering first is narrower and much cheaper: where exactly do the numbers diverge, and why?

That is a mapping exercise, not a software purchase. Follow one product from receipt to invoice and write down every system it touches, every place a human retypes something, and every point where two records could legitimately disagree. It takes a few days. It is boring. It is also the single highest value piece of work available in this situation, because it turns “our systems do not talk” into a specific list of about a dozen things.

Some of those dozen turn out to be a genuine integration. Some are a process change that costs nothing. Some are a definition argument that has been running for years without anyone noticing it was a definition argument, usually about when a thing counts as sold.

What good looks like afterwards

Not one system. One place where each fact lives, and documented paths between them.

The test is not whether the reports are pretty. It is whether someone can be told a number and act on it without first asking where it came from. When the answer to “is that right?” stops being “let me check”, the problem is solved, whether you replaced anything or not.

Sometimes that lands as a full ERP implementation. Sometimes it is three integrations and a decision about terminology. We would rather find out which before anyone commits to a build, and we price the finding out on its own so that the answer is not shaped by what we would prefer to sell.