Packaging Manufacturing Software: Build vs Buy

What the off-the-shelf products actually do well

One plant, one substrate family, under about 60 jobs a week, fewer than 500 active dies: buy. A well configured Amtech Imaginera, EFI Radius or NetSuite with a competent implementation partner will serve you, and a custom build would be an expensive route to the same result plus maintenance you now own forever.

These products are not weak. Amtech and Radius understand converting. They carry the customer master, the order, purchasing and invoicing, and a scheduling view that works when your mix is stable. Kiwiplan is deeply established in corrugated. Tharstern and Cerm do real work in labels and narrow web, and Cerm in particular knows what a repeat job looks like in that world. Esko ArtiosCAD is the standard for structural design and there is no case for rebuilding it.

Buy ArtiosCAD or Arden whatever else you decide. Structural design is not the problem you have.

The boring parts matter too. Someone else keeps the tax tables current. Someone else owns the accounting close. Someone else answers when your controller cannot post a batch on the last day of the month. A build gives you none of that, and buying it back later is not cheap.

The wrong reason to build is disliking your enterprise resource planning (ERP) interface. Keep that system for receivables, payables and the general ledger, which it does properly. What follows is about the layer it was never going to give you.

Where they stop: the die is an invoice line, not an asset

Ask your plant manager how many active dies you have. You will get a range rather than a number, and that range is the whole problem in one answer.

Every packaged system stores a die number and a customer, because the record exists to support a tooling charge on an invoice. Your plant treats a die as something else: a physical object with a rack location across three buildings, a rule height, a board caliper range it was cut for, an up count on each specific press deck, a hit count since last rework, and a link to the drawing revision actually cut rather than the one sitting on the shared drive. None of those attributes fit an accounting record, and every one of them drives a decision.

So Tuesday happens. A buyer wants a repeat on a 24 point carton, same as last April. The order shows a price. It does not show which die ran it, which press, which board lot, or that the job ran at 11 percent waste instead of the 6 percent in the estimate because the board came from a different mill and cracked at the score. The April price goes out. The job loses money and nobody knows until the close, by which time two more repeats have been quoted from the same number.

The second gap is the estimating workbook. Fourteen tabs, a macro nobody will touch, and twenty years of knowledge about make ready on the 40 inch and which board runs three points over specification. No enterprise system absorbs that, because it wants standard routings and standard costs, while your cost is lot behaviour, die condition, press and customer interacting.

The arithmetic: licences and modules versus a build

Systems in this category price per named user with module fees layered on top: estimating, scheduling, shop floor, quality, each sold separately. Ask for your renewal broken out by module, then ask the harder question. Which of those modules is genuinely in use, and which was bought because the plant said it would replace the spreadsheet and then did not.

Most converters find at least one module they pay for every year and do not run. That line, rather than the total, starts the honest comparison.

The crossover sits around 60 jobs a week, or 500 active dies, or the moment a second plant opens, whichever arrives first. Plant count matters more than volume, because two sites force a decision about whether dies and substrate are shared or local, and no packaged tooling module answers it either way.

Then price the variance, which is the number that actually decides this. Take one part you run repeatedly and pull every job of it from the last twelve months. Compare realised margin job by job using actual material and actual hours. If the spread across the year runs more than about six points on the same part, you are not pricing, you are guessing, and at moderate volume that guess costs roughly a first release every year.

Add the dead tooling. Walk one rack, count the dies not hit in three years and the duplicates cut because somebody could not find the original. That is a one time recovery a die library pays for on its own.

What a custom build actually costs

Bands. A focused first release, meaning the die library as a real asset record plus the estimating rules engine, integrated read only against your existing enterprise system so nothing about invoicing changes, runs $60,000 to $130,000 and ships in 12 to 16 weeks. That is the release that pays for itself, because it fixes the quote. A full platform adding scheduling against substrate constraints, shop floor data capture, live job costing and a customer portal runs $150,000 to $400,000 phased across 6 to 12 months. Nobody should buy the second as one block.

Data migration adds 10 to 25 percent and here it is physical work. Digitising drawings and job history is three to five weeks of engineering. Walking the racks with a scanner to tag steel and confirm what exists is two to four weeks per plant, and it is where converters discover a large share of the library is dead or duplicated. Budget it as a line item rather than assuming it happens alongside development.

Year two runs 15 to 20 percent of build cost annually. Presses get replaced, rules get revised, and every new substrate is a factor somebody has to seed and then correct from actuals.

What pushes the number up: press integration, where a modern line with an open data layer is a fortnight and four legacy lines with proprietary counters and no network drop is six weeks plus hardware and an electrician. Plant count, for the shared or local decision. Food contact requirements. And the one that genuinely breaks budgets, an enterprise system with no clean interface access, where you end up reading at database level on a nightly sync instead of live.

The four situations where building wins

• Regulatory fit. If you supply food contact packaging under a recognised packaging materials standard, you will be asked to demonstrate traceability in a mock recall against a clock, commonly inside four hours, from finished pallet back to substrate lot and forward to every customer who received it. A quantity of a stock item cannot answer that. Serialised rolls and skids carrying a lot, tied to the job that consumed them, can. Fibre chain of custody certification asks the same thing from another direction, and pharmaceutical packaging adds electronic records obligations on top. That scope belongs in phase one, because retrofitting it means rebuilding the inventory model.
• Scale economics. Module and seat pricing across multiple plants, where you pay annually for capability the plant does not use because the estimating still happens in a spreadsheet.
• A workflow that is your competitive advantage. In converting that is almost always quoting speed with accuracy. If you win work because you turn a complex quote in four hours while competitors take three days, that capability currently lives inside one estimator with nineteen years of service. Encoding the rules as named, versioned factors turns a person into an asset, and it is usually the reason an owner funds this at all.
• Integration sprawl across three or more systems. The enterprise system, the structural design tool, the estimating workbook, the tooling folder, the press controls and a scheduling whiteboard. The die is the key that would join most of them, and no product treats it as one.

Two of those true is a build. One of them is a module you should switch on properly first.

How to decide in a week

Test the repeat. Everything you need already exists in your own records.

Monday: pick your three highest volume repeat parts and pull every job of each from the last twelve months.

Tuesday: compute realised margin per job using actual material and actual hours rather than standard cost. Plot them. You are looking for spread, not average.

Wednesday: for the three worst jobs, find the cause. Substrate lot behaviour. A die overdue for rework. Make ready on the wrong press. A late art change. Then ask whether the estimator knew any of it at the moment of quoting. Usually they did not, and that is the finding rather than the margin itself.

Thursday: run a trace. Take one finished job from six months ago and time how long it takes to name every substrate lot inside it and every customer who received product from those lots. If your certification requires that answer inside four hours and it takes a day, you have a second reason to act that has nothing to do with margin.

Friday: put both numbers against the bands above. Margin spread under about three points and a trace completed in an hour means you should configure the module you already pay for and stop there. A wide spread and a failed trace means start with the die library and the rules engine, and leave the enterprise system exactly where it is.

What follows is a paid discovery phase rather than a proposal, and in this category insist on an integration spike inside it. Two to three weeks, fixed fee, producing a signed product requirements document covering the die and substrate model, the estimating rule structure, the confirmed read method against your enterprise system, and acceptance criteria. You own that specification whoever builds it.

Who we are wrong for: single plant commodity converters, anyone shopping purely on hourly rate, and anyone who wants a full enterprise replacement. Digital Heroes writes that requirements document before any code, with more than fifty specialists and India LLP, US LLC and UK LTD entities so intellectual property assigns under your own law. ShopScore, HeroCheckout and Section Vault are our own products, over 2,000 projects sit behind us, and you meet the named team before signing. We are listed on Clutch, Trustpilot, Fiverr Vetted Pro and D-U-N-S.

Book a 30-minute call with Digital Heroes and get a written plan and a fixed quote within 48 hours.