Inventory management for manufacturers: a practical guide

A shop buys finished products and sells the same products. A factory buys materials, turns them into something else and then sells that. Between the delivery door and the loading bay there are at least three different kinds of stock, and each one is counted, stored and worried about in a different way.

When inventory goes wrong in a factory the symptoms are familiar. A sales person promises something that is not really there. A team stops work to look for a part that is on the wrong rack. A stocktake takes a weekend and the numbers are out again by Wednesday. Nobody can say how much money is sitting on the floor.

The fix is rarely clever. It is a handful of clear rules, applied every day: know which kind of stock you are looking at, count it in a way that suits it, give it an address and scan it when it moves. This guide walks through each of those rules.

Almost every manufacturer holds some mix of the following. Treat them separately, because the questions you ask about each are different.

Raw materials and components

Timber, foam, fabric, steel, fixings, packaging, springs, glue. These are bought in bulk, measured in metres, kilos, sheets or boxes, and used up in the making. The key questions are how much is left, how long it will last and when to reorder.

Work in progress

Anything that has started to be made but is not finished: frames waiting for upholstery, cut panels waiting for assembly, a base with no cover yet. Work in progress is the stock factories track worst, because it lives on the shop floor rather than in a store. The key questions are what stage it has reached and who is waiting on it.

Finished goods

Completed products, boxed or wrapped, waiting to be loaded. These are the pieces that customers care about and the ones that tie up the most cash per square metre. The key questions are which order they belong to, where they are and how long they have been there.

Your production model decides how much finished stock you should hold at all, so settle this before you design any process.

In a make to order business, finished stock should be a waiting room, not a warehouse. A piece is made, stored for a day or two, and loaded. If something sits for weeks, that is a signal: a customer has not paid, an order is incomplete or a mistake was made.

The hybrid most factories actually run

Plenty of factories make most things to order but keep a small number of fast sellers in stock so they can promise short delivery. That is perfectly sensible, as long as you keep the two groups apart. Stock made for a named order is reserved. Stock made for the shelf is free. If the two mix, someone will eventually load a piece that was meant for another customer.

A simple rule: every finished piece should be either attached to an order or clearly labelled as shelf stock. If it is neither, it is a mystery, and mysteries cost money.

The count in your system is only useful if it matches the count on the floor. There are three ways to keep them together.

Full stocktake

You stop and count everything, usually once or twice a year. It is thorough, and it is disruptive. Production halts, people count tired, and the numbers begin to drift the moment you restart. Many auditors still expect one, so it has its place, but it is a poor way to keep daily accuracy.

Cycle counting

You count a small slice of stock every day or week, so that everything is counted several times a year without ever stopping the factory. Cycle counting finds errors close to when they happen, which makes them far easier to trace.

Prioritise what you count. A common approach is to group items by value or by how often they move:

• A items: high value or fast moving. Count them weekly or monthly.
• B items: moderate value. Count them every quarter.
• C items: low value, slow moving or cheap. Count them once or twice a year.

Counting at the point of movement

The most accurate method is not a count at all. If every piece is scanned when it is made, put away and loaded, the system knows the answer already, and a count simply confirms it. This works best for finished goods, where each piece can carry its own label.

How to run a cycle count in 15 minutes

1. Pick today’s slice: one rack, one product family or one supplier.
2. Print or open the expected list, but do not let the counter see the expected quantity if you can avoid it.
3. Count what is physically there and write it down.
4. Compare. If the numbers match, sign off and move on.
5. If they differ, recount once, then look for the cause: a wrong scan, a wrong location, a missing paperwork step, damage or a move that was never recorded.
6. Fix the cause as well as the number. Adjusting a quantity without finding the reason just guarantees you will adjust it again next month.

Keep a short log of every difference you find and the reason. After a month it will tell you exactly where your process leaks.

Accuracy is not only about how many you have, it is about where they are. A piece that exists but cannot be found is, for practical purposes, not in stock.

A good location scheme is boring and obvious. Divide the building into zones, the zones into racks or bays, and give each a short code that a new starter can read. For example, a rack called B and a bay called 04 give you B04. Put a clearly printed label at each spot.

• Use one format everywhere. Mixing names such as “back left” and “rack 3” guarantees confusion.
• Keep raw materials, work in progress and finished goods in separate zones.
• Reserve a clearly marked area for problems: damaged, returned or waiting for a decision.
• Leave some space free. A full rack is a rack where people start putting things on the floor.
• Record the location when the item is put away, not later from memory.

Recording the location at the moment of putting away is the one habit that pays back more than any other. It turns “I think it is round the back” into a ten second job.

Every number typed by hand is a chance for an error. Barcodes remove the typing. A label on the item, a scan at the moment of an event, and the record updates itself.

For internal use you do not need an official product code. You need a label that is unique, prints clearly and scans quickly. Code 128 is a sensible default for linear barcodes, because it handles letters and numbers and is read by almost every scanner. A QR code can carry the same information in a small square and is readable by an ordinary phone camera. Many factories print both on the same label.

Think about what a label has to survive. Work in progress gets handled, dragged and stacked, so put the label on a flat face that is not worn. Choose a label material that sticks to your surface, and print at a size your cheapest scanner can read from arm’s length.

• Label pieces as early as you can, ideally when the order is released to the floor.
• Make sure one barcode points to one thing. Never reuse a code.
• Test the scanner on a real, slightly dirty label, not a clean sample.
• Decide in advance what a scan means. A scan at the finishing bench means “made”. A scan at the door means “loaded”.

Scanning also gives you the data that makes everything else in this guide work: who did what, where and when.

A spreadsheet is a perfectly good place to start. It is cheap, everyone knows it and you can shape it exactly to your business. Plenty of good factories run on one for years. But there is a point where it begins to cost more than it saves.

Look for these signs:

• Two people edit it at the same time and one set of changes is lost.
• Someone is employed, in effect, to keep it up to date.
• You cannot answer “where is order 1234?” without walking the floor.
• Counts are wrong by the end of each week, so nobody trusts them.
• Retailers or customers keep phoning to ask for progress.
• Typing mistakes cause the wrong item to be loaded.

If three or more of those are true, a scanning based system will almost certainly pay for itself. Choose by the job you need done. Raw material control, purchasing and costing belong in an MRP or accounting package. Following finished pieces from the saw to the van is a different job, and it is the one Skuflo handles for make to order factories: every piece gets its own barcode, and stock is the live count of what has been scanned as made and not yet loaded. It does not manage raw materials or purchasing, so many factories use it alongside the system they already have.

1. Counting only what the accountant asks about

Raw materials show up on the balance sheet, so they get counted. Work in progress and finished goods are often guessed. Count all three.

2. Adjusting quantities without a reason

Every correction should have a cause written next to it. Otherwise the same fault repeats.

3. No owner for each area

If everyone is responsible for the store, nobody is. Give each zone a named person.

4. Letting finished goods build up

In a make to order business, aged finished stock is a warning sign. Review anything older than a few days every morning.

5. Skipping the put away scan

People scan at the start and the end of the process and skip the middle. The middle is where things get lost.

1. Write down your three kinds of stock and who looks after each.
2. Decide which products are made to order and which, if any, are made to stock.
3. Walk the building and name every zone, rack and bay.
4. Print location labels and put them up.
5. Choose one slice of stock and run your first 15 minute cycle count.
6. Pick a barcode format and test a scanner on real labels.
7. Agree what each scan means, and write it on the wall next to the scanner.

None of this needs special software, and all of it makes the software easier later. For more on tracking pieces through the factory, see our guides to production tracking and SKU tracking.