Pulp and Paper Mill Production Software: Custom Build or Off-the-Shelf

What Greycon, ABB, Honeywell and Valmet actually do well

Nobody in this category is selling snake oil, and the mills that assume otherwise usually end up commissioning a worse cutting stock solver at four times the price.

Greycon is the genuine specialist. opt-Studio and X-Trim solve the cutting stock problem properly, with people who have spent careers inside paper and board mills and who will recognise your order book from the first meeting. If your trim is high because nobody is optimising it at all, buying that product is the correct decision and it will pay back faster than any build.

On the process side, ABB, Honeywell with Experion MX, and Valmet with IQ and DNA cover quality control, sheet measurement and production reporting to a standard no bespoke project should try to reproduce. Basis weight, moisture and caliper control against TAPPI and ISO test methods is their ground and it is not yours. SAP for Mill Products handles the commercial layer for larger groups, including the variant configuration and catch weight handling that ordinary enterprise systems fumble.

Most mills should buy. A machine running long campaigns of a handful of standard widths is already close to its trim floor, and the money belongs in reliability or in reducing quality variability rather than in an optimiser. We say that in scoping calls and it costs us projects, but a mill whose planner spends an hour a week on sets rather than a day will not see a return.

Where they stop: the constraint your winder has and their model does not

Two gaps, and they compound.

The first is that trim and grade sequencing are one problem solved separately. A better trim solution may require running two orders together at different basis weights, which forces a grade change, which costs broke, machine time and a period of off specification production. Optimise trim alone and you can produce a schedule with excellent millimetres and terrible economics. Most tools address one side, the scheduling discussion happens in a meeting, and the interaction between the two lives in the planner's judgement, which is fine until the planner is on holiday.

The second is your own constraint set. Maximum knives on the winder. Minimum roll width the winder handles reliably. Maximum set count. Roll diameter limits at the customer. Whether a particular account accepts a splice. Which orders share a core size and can therefore go on the same set. Whether the sheeter downstream takes a given width without a changeover. Generic optimisers accept a handful of these as parameters and leave the rest as rules the planner applies afterwards, which is exactly the manual work that makes the optimiser only partly useful.

Then there is tolerance. Most paper and board orders carry a delivery tolerance agreed with the customer, commonly around plus or minus ten percent. That tolerance is an optimisation asset. Producing slightly over on one order can enable a far better set, and slightly under can avoid a grade change. Because tolerance is filed as a contractual footnote rather than a variable, planners use it accidentally rather than deliberately, and no packaged tool asks them to do otherwise.

Finally, certification. If you hold Forest Stewardship Council or Programme for the Endorsement of Forest Certification chain of custody, and increasingly if you sell into the European Union under its deforestation due diligence rules, the certified fibre claim has to follow the order through production and onto the delivery document. That is a data lineage requirement, and it sits outside the trim solver entirely.

The arithmetic: licence per machine against a millimetre of trim

This category is not priced per seat, so do the sum in tonnes.

Take your deckle and your typical trim. A hundred millimetres of trim on a 6.2 metre deckle is about 1.6 percent of production. A machine at 300 tonnes a day over 350 days makes 105,000 tonnes, so that trim is roughly 1,700 tonnes a year going to broke, repulped, carrying fibre, steam, energy and machine time that produced nothing sold. Multiply those 1,700 tonnes by your own contribution per tonne. That is the annual prize, and it is a number nobody in the mill would approve as a line item if it were presented in a meeting.

Now put the two paths beside it. A specialist licence plus annual maintenance for one machine is a modest fraction of that prize, which is why buying is usually correct. A custom first release runs $90,000 to $190,000 with support at 15 to 20 percent a year, so a $140,000 build costs roughly $259,000 across five years.

The crossover is order book variety rather than tonnage alone, but tonnage sets the floor. Below roughly 100,000 tonnes a year on one machine with a narrow width range, buy the specialist product or do nothing. Above roughly 150,000 tonnes a year with a varied order book, frequent grade changes and tolerance that is never used deliberately, the combined trim and sequencing gain typically clears a build inside two years. The cheapest diagnostic: count how many minutes your planner spends redoing downstream work after trying an alternative set. If trying a second option is too expensive to attempt, you are accepting the first one every week.

What a custom build actually costs

A first release covering order intake from your enterprise system, the full constraint model for one machine, combined trim and sequencing optimisation against a money objective and a planner interface with alternative comparison runs $90,000 to $190,000 in 14 to 20 weeks. That is a system your planner uses for next week's schedule, not a study. A full platform adding converting and sheeting constraints, roll and parent reel inventory, multi machine allocation, order promising, quality system linkage and plan versus actual attribution runs $250,000 to $600,000 phased over 9 to 18 months.

Then the lines nobody quotes. Migration runs 10 to 25 percent of build cost, and here it is not really data movement. It is capturing the constraint set, which lives in the heads of two people on the machine floor and in the exceptions a planner applies without thinking. Budget it as facilitated workshops with the winder crew present, not as an extract. Year two onward runs 15 to 20 percent of build cost annually, and in this category that budget has a standing job: constraints change when you rebuild a winder, add a customer or requalify a grade, and a stale constraint set produces schedules the crew quietly ignores.

What drives the number up: the count of machines and whether orders can be allocated between them, since multi machine allocation is materially harder than single machine trim; owned converting operations, which roughly double the model; grade structure complexity, meaning how many basis weights, coatings and colours run and how constrained the transitions are; and how deep the enterprise integration goes, particularly if order promising is handed back to the sales system.

The four situations where building wins

• Regulatory fit. Chain of custody claims that must follow a specific order through trim, winding and dispatch, plus energy reporting under ISO 50001 if your site is certified. When the compliance record is produced by the same plan that runs the machine, it stops being a quarterly reconstruction.
• Scale economics. Above roughly 150,000 tonnes a year on a varied book, the combined trim and grade sequencing gain clears the build inside two years. Do that sum with your own contribution per tonne rather than a vendor's percentage claim.
• A workflow that is your competitive advantage. If you own converting, scheduling the mill and the converting plant as one system is the thing your competitors cannot copy by buying a licence. Scheduling them separately means one of them permanently absorbs the other's inefficiency.
• Integration sprawl across three or more systems. The enterprise system for orders, the quality control system for actual production and off specification tonnes, the warehouse for roll inventory, and dispatch. When four interfaces each need a vendor change request, the interfaces become the project.

One signal alone is rarely enough. Two together is the tipping point, and the most common pairing is owned converting plus a tolerance policy that never reaches the optimiser.

How to decide in a week

Do not commission a study. Run a bake-off on one real week.

Take next week's order book exactly as it stands. Have your planner build the schedule the way he always does and record how long it takes and how much trim it carries. Then ask every vendor and every developer on your shortlist to produce an alternative from the same data, and require each to state the objective they optimised. If anyone answers minimise trim, they will hand you a schedule the mill cannot run, because the objective has to price grade changes, lateness, tolerance usage and plan stability in money alongside trim.

Second, hand over three constraints that are not in any standard model. Your minimum reliable roll width, the customer who refuses a splice, and the core size rule. Ask how each is maintained after go-live. If constraints live in code rather than in data your planning team can edit, the model is stale within a year and the crew will start ignoring it.

Third, ask what happens when Wednesday's break wipes eight hours from the plan. A rebuild that takes minutes lets you re-plan daily. A rebuild that takes an afternoon means you will not, and the schedule will drift every day it is not refreshed.

Then buy a paid discovery phase that ends with a written specification you own: the full constraint model, the objective function with its cost terms, the integration points named by system, the stability penalty, and acceptance criteria measured against a real week. Take it to every other firm on your list. That document is what makes four quotes comparable and what keeps a fixed price fixed.

Digital Heroes will not write code without a signed product requirements document. We contract through India LLP, US LLC and UK LTD entities so intellectual property assigns under law your own counsel already reads, and you own the repository, the infrastructure and the constraint model from the first commit. Over fifty specialists, more than 2,000 projects, a named team you meet before signing, checkable on Clutch, Trustpilot, Fiverr Vetted Pro and D-U-N-S.

We are the wrong firm for a mill running long campaigns of a few standard widths, and wrong for anything inside the quality control loop, which belongs with your automation supplier. Buy from ABB, Honeywell or Valmet for that and do not let anyone tell you otherwise.

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