Spreadsheets are not a beginner’s tool. They are extraordinarily good at what they do, and a well-built planning workbook can run a small manufacturer for years — often better than a badly implemented system would.

The problem is not quality. It is that a spreadsheet has a ceiling built into its architecture, and you do not hit that ceiling gradually. You hit it as a series of specific failures that look like people problems, or communication problems, or one-off mistakes. Here are the seven that show up most often, and a test for each.

1. Two people look up the same stock figure and get different numbers

What it looks like: sales quotes availability from a copy on their laptop. The stores supervisor works from a copy on the shared drive. Both are confident. Both are looking at different numbers.

Why it happens: a spreadsheet has no concept of a single authoritative record. Every copy is equally valid, and versions branch the moment two people need the file at once.

The test: ask two people, separately, in the same hour, how many of your fastest-moving component you have. If the answers differ, this sign is present.

2. Somebody rebuilds the bill of materials by hand when engineering changes a part

What it looks like: a component gets substituted, and someone spends a morning finding every product that used it and updating each one.

Why it happens: a spreadsheet stores the relationship between a product and its components as text in cells, not as a link. There is nothing to follow when you need to ask which products contain a part.

The test: pick a component used in more than one product. Time how long it takes to produce a complete list of everything it goes into. Anything over five minutes means the relationship exists only in someone’s head.

3. You find out about a shortage after the job is already on the floor

What it looks like: the work order is released, the operator goes to the rack, and the material is not there — because it was committed to a different order that had not been built yet.

Why it happens: a spreadsheet shows what you physically have. It does not net that against what is already promised. Free stock and physical stock are different numbers, and only one of them is visible.

The test: take your three largest open orders and list every component they consume. Add the quantities. Compare against on-hand. If nobody currently does this calculation, nobody currently knows whether those orders can be built.

4. Job costs are reconstructed after the fact instead of accumulated during

What it looks like: at month end, somebody works out what a job cost by going back through purchase invoices, timesheets and memory. The number arrives weeks after the decision it should have informed.

Why it happens: a spreadsheet has no transaction log. Cost accumulates in reality — material issued, hours booked, scrap generated — but nothing is capturing those events as they happen.

The test: ask for the gross margin on a job that closed last week. If the answer takes more than an hour to produce, or comes with the word “roughly”, cost is being estimated rather than measured.

5. One person is the system

What it looks like: the planning workbook has an owner. When they are on holiday, planning does not happen — it waits.

Why it happens: the rules live in formulas that only the author understands, and the exceptions live in that person’s judgment, which is not written down anywhere.

The test: ask that person to take a week off without checking messages. Most manufacturers already know the answer and have quietly arranged the year around avoiding it.

6. You cannot say which customers received material from a given lot

What it looks like: a supplier flags a defective batch. Establishing where that material went takes a day of cross-referencing delivery notes, and the answer is “probably these customers.”

Why it happens: lot identity is recorded on the goods receipt and then lost the moment the material is issued to production, because a spreadsheet has no mechanism to carry an attribute through a transformation.

The test: pick a lot number from a goods receipt three months ago. Trace it forward to shipped orders. If this cannot be done reliably, it cannot be done in a recall either — and a recall has a clock on it.

7. Reordering runs on memory rather than on demand

What it looks like: someone walks the stores, notices a bin looks low, and raises a purchase order. It works because that person is good at their job.

Why it happens: the spreadsheet contains stock, but it does not contain the arithmetic that turns open demand, lead time and consumption rate into a reorder decision. So a person does it, in their head, item by item.

The test: count the expedited or premium-freight orders you placed in the last quarter. Each one is a reorder decision that was made too late.

What is actually breaking

Every sign above traces to one of three things a spreadsheet does not have and cannot be made to have.

Concurrency. Only one person can hold the authoritative version at a time. Everything else is a copy, and copies diverge.

Transaction history. A spreadsheet stores the current state. It does not store the sequence of events that produced that state, which means you can see what you have but not what happened, and you cannot audit, reverse or cost anything after the fact.

Referential integrity. A component in a BOM is text, not a link to an item record. Nothing enforces that the same part is described the same way in two places, and nothing lets you follow the relationship in either direction.

Better formulas do not fix any of the three. Neither does moving the file to a cloud drive, which addresses concurrency partially and the other two not at all.

What to do before you buy anything

Crossing this ceiling does not automatically mean signing a contract this quarter. Three things are worth doing first, and all of them make a later implementation cheaper.

Clean the item master. Whatever you eventually run, it will run on your part numbers. Duplicates, inconsistent descriptions and three naming conventions from three different eras will migrate straight into the new system and cost more to fix there. Start now, in the spreadsheet.

Write down the rules that live in someone’s head. Which supplier for which part and why. What the real lead times are, as opposed to the quoted ones. Which customers get priority. This is the most valuable and least documented asset in most small manufacturers.

Quantify one of the seven signs. Pick the one that costs you most and measure it for a month — hours spent, orders shipped late, expedite charges paid. A number turns a system purchase from a preference into a decision, and it gives you the baseline to judge whether the system worked.

If you want to see what the same planning problem looks like when the three missing properties are present, book a live demo and bring one of your own products with a real bill of materials.

FAQ

Is there a company size at which spreadsheets stop working? No reliable one. Complexity matters far more than headcount. A twelve-person shop making configured products with multi-level BOMs hits the ceiling before a sixty-person shop making three simple products in long runs.

Can we keep using spreadsheets alongside a system? For analysis, yes, and most manufacturers do. For the authoritative record of stock, BOMs and orders, no — that recreates the divergence problem in a more expensive form.

We only have two of the seven signs. Is that enough? It depends which two. Signs 3 and 6 have direct commercial and regulatory consequences. Signs 1 and 5 are early warnings and often tolerable for another year. Weight them by what they cost you, not by how many you counted.

Would a cheap inventory app fix this? It fixes signs 1, 3 and 7 if it holds the authoritative stock record. It generally does not fix 2, 4 or 6, because those need multi-level BOMs, work order cost accumulation and lot genealogy — capabilities that separate manufacturing systems from inventory apps.