Dewulf builds agricultural machinery, more than a thousand configurable machines a year. Sales configured each machine in Elfsquad. Engineering then re-typed that same configuration into Windchill to get it built. Two teams, two systems, one product, and a person in the middle acting as the integration layer by hand. Hundreds of engineering hours a year went into copying decisions that had already been made correctly once.
Nobody had designed it that way. The sales configurator arrived to solve a sales problem. The PLM system arrived to solve an engineering problem. Both worked. The gap between them was never anybody's job.
We are the.good.code;, a founder-led team that builds custom software, CPQ (configure, price, quote) and configurator systems, and ERP-integrated operational layers for manufacturers. We removed that manual step for Dewulf by building the layer between the two systems, which is the sort of work that teaches you where the real asset in a configurable business sits. It sits in the rules.
In Tacton's own 2026 State of Manufacturing survey, an online survey of 280 manufacturing leaders fielded in April 2026 with Researchscape, 7% of manufacturers said they define configuration rules once and reuse them [1]. Everyone else re-synchronises at every change, order and handoff.
Configuration logic lives in four systems at once
Configuration logic is the set of rules that says what your product can be. Which options are compatible. Which parts a feature requires. What a combination costs to make, and whether the parts can be sourced at all.
In most manufacturers that logic exists in four places, in four different forms. In the CPQ or configurator it lives as product rules, written so a salesperson can reach a valid quote. In engineering it lives as design constraints, written so the machine can be built at all. In the ERP it lives as bill of materials structures, written so the thing can be costed and purchased. On the shop floor it lives as routing and capacity rules, written so it can be scheduled.
Four systems, four native forms, four teams maintaining them. None of them is wrong. They are copies, and copies drift.
What the research says about how common this is
Vendor surveys, an analyst brief and the academic literature all point at the same shape. Who produced each one matters when you quote them.
Tacton's 2026 survey (n=280, fielded April 2026, US 62% and Europe 38%) found 33% of manufacturers using consistent configuration logic across sales, engineering and production [1]. The larger group, 55%, manages complexity through a mix of guided quoting and downstream engineering adjustments, which means the quote goes out and engineering reconciles it later. Tacton sells configuration software, so read those numbers as a vendor's own research rather than an independent market benchmark.
An IDC InfoBrief covering 511 discrete manufacturing companies across North America and Europe, sponsored by Configit, reports that manufacturers use an average of three different configuration tools [2]. Different sample, different sponsor, same finding: the logic is not held in one place.
The academic literature treats this as unresolved rather than merely unimplemented. A systematic review of product configuration for mass customisation published in the International Journal of Production Research in 2026 examined 122 articles from 2000 to 2025 under the PRISMA framework, and it names four problems the field has not solved: keeping models workable as portfolios grow, the fatigue users hit when a configurator asks too much of them, uncertainty in customer requirements, and environments that keep changing underneath the model [3]. Worth knowing before you assume the companies without a single source of truth are simply slow adopters.
Why the copies drift
Duplication on its own is survivable. Drift costs money, and drift has three ordinary causes.
Each copy has a different owner with a different deadline. A product manager adds an option and updates the configurator, because that is what stops sales quoting the old range. The design constraint in engineering gets updated when engineering next touches that assembly. Between those two events the two systems disagree, and every quote produced in that window carries a small risk forward.
The copies are written in different languages, so no automated check can compare them. A compatibility rule in a configurator and a design constraint in a PLM system describe the same physical truth in forms that cannot be diffed. When they diverge, nothing throws an error. Somebody finds out later, usually in production.
Maintenance effort scales with the number of copies, and that effort is already high. In Tacton's survey, 93% of engineering teams report moderate to very high effort maintaining configuration rules across systems, and 81% report the same for CPQ model maintenance [1]. Multiply an expensive activity by four and the rational response is to defer it, which is what happens.
Where the cost shows up
The cost never appears on a line item called configuration drift. It appears in four places usually treated as separate problems.
Quoting takes the first hit. Tacton found 43% of manufacturers citing customisation as their top quoting challenge, and third-party CPQ adoption in that same survey series rose 19 points since 2022 [1]. More tooling, same problem. Speed improved and accuracy did not, because a faster path to a quote does not help when the rules behind the quote disagree with the rules behind the build.
Engineering absorbs the second hit as rework. In the same survey, 72% of manufacturers say they are confident in their quoting, while 18% report frequent rework and the concern sits highest among engineers. That gap between stated confidence and observed rework is the most useful diagnostic in the whole report. Confidence is measured in the room where the quote is made. Rework is paid in the room where it gets built.
Delivery takes the third hit as a credibility problem. Forty percent of manufacturers are only somewhat or not very confident in their delivery commitments at the time of quote. A promise made without the constraint check behind it is a promise the factory may not be able to keep.
Margin takes the fourth hit, spread thin enough to become invisible. Sixty-two percent describe margin erosion between quote and delivery as at least a moderate issue, and the reported causes split across people, tools and process with no single dominant factor. That split is what a problem living in the handoffs looks like in a survey: each function sees its own fragment and nobody prices the total.
Adding AI on top does not resolve this
Seventy-nine percent of manufacturers in Tacton's survey are investing in or exploring AI, up from 64% the year before. Only 18% rank it as their first digital transformation priority, which suggests most teams already sense that something has to come first.
They are right, for a mechanical reason. An assistant that recommends a configuration reads the rules it is given. When the configurator holds one version of the truth and engineering holds another, the assistant produces confident output from whichever copy it can reach. That confidence is the problem, because it removes the hesitation that used to make somebody double-check the number.
Capability is no longer the constraint. Rockwell Automation's 2026 State of Smart Manufacturing report, conducted by Sapio Research across 1,560 respondents in 17 countries, found 59% actively using smart manufacturing technologies in operations against 18% still in pilot mode, with 34% of operations AI-augmented today [4]. What the technology is pointed at is the constraint. We have written separately about what AI-native requires underneath an ERP's chat interface, including the permission and audit architecture that decides whether an agent can be trusted to write. Rule consistency belongs in the same category of unglamorous prerequisite.
What define-once requires
Define-once is an architectural commitment rather than a single product you install, closer to what analysts call configuration lifecycle management than to a purchase decision, and it has three parts.
Constraint-based logic instead of chains of if/then rules. A constraint states a relationship that must hold, and the system derives valid combinations from it. If/then rules enumerate paths, so every new option multiplies the branches somebody has to maintain. Fewer enumerated paths means less surface to diverge, whichever engine you run.
Automatic propagation. Defining rules once only counts if a change reaches every consumer of those rules without a human relaying it. In Tacton's data, 21% propagate engineering changes automatically and 23% auto-generate manufacturing bills of materials from a quote [1]. Those two numbers are the honest measure of how far most manufacturers are from define-once, whatever the architecture diagram says.
One system is the source and the others are consumers. This is the part that stalls, because it is an organisational decision wearing technical clothing. Someone has to accept that their team's copy is now downstream of another team's copy.
An integration layer is a legitimate answer here, and often the only affordable one. For Dewulf we built middleware that mapped configurator characteristics onto PLM attributes and moved the configuration across automatically, with the mapping methods drawn from interviews with the engineers who had been doing it by hand. The manual step went away entirely, along with the errors it generated. Neither system was replaced. Both kept doing their job, and the copy stopped being maintained by a person.
Where we sit on this
The vendor answer to fragmented configuration logic is usually a fifth system. We don't sell CPQ licences, so we have no reason to recommend one, and that's the vantage point behind everything below.
Our configurator and CPQ work for manufacturers is where this view comes from. The Dewulf integration above. A rebuilt single-page configurator for pool cover manufacturer Coverseal, running on Elfsquad logic underneath. A bespoke visual configurator for robotic cells at Coboworx, where no off-the-shelf tool could hold the product logic. Different shapes, same lesson: the rules layer is the asset and the interface is the easy part.
Open Mercato, the platform we now build operational systems on, ships a Catalog and CPQ module next to its workflow and event modules. That makes it possible to hold the rules layer in one place instead of buying another copy of it. We should be straightforward about maturity: that module is young, and our production track record on this platform is still being built. Where a proven configurator engine is the right call, we integrate one, which is what the Dewulf and Coverseal projects did.
The first move, and it is cheap
Before any tooling decision, write down where each type of configuration rule currently lives, who owns it, and how a change reaches the other copies today. One page. The compatibility rules, the design constraints, the BOM structures, the routing rules, and for each one an owner and a propagation path.
Most teams find at least one path that turns out to be a person. That is the finding worth having, because it is where the next quote will go wrong. The fuller version of this exercise for integration projects generally is in our guide on how to map your software ecosystem before any integration project, and the configuration version follows the same discipline.
If you want a second pair of eyes on that page once it exists, our technology strategy advisory work starts exactly there, before anyone writes code. Sometimes what comes out of it is that you should change a process rather than buy software, and we will say so.
The next two articles go up and down the timeline from here: how to decide which product variants are worth carrying into a rebuilt configurator, and what it takes to automate the quote-to-BOM handoff so a commitment made at quote time survives contact with the factory.
FAQ
What does it mean to define configuration rules once?
One system holds the authoritative rules about what your product can be, and every other system consumes those rules instead of storing an edited copy. Changes are made in one place and propagate automatically to quoting, engineering, ERP and production.
How many manufacturers do this today?
Seven percent, according to Tacton's own 2026 survey of 280 manufacturing leaders. A separate IDC InfoBrief of 511 discrete manufacturers, sponsored by Configit, found companies using an average of three different configuration tools.
Why does buying a CPQ system not solve it?
CPQ solves the speed of producing a quote. When its product rules are a copy of the rules engineering maintains, the copy still drifts. Tacton's survey series shows third-party CPQ adoption up 19 points since 2022 while customisation remains the top quoting challenge at 43%.
What is the difference between constraint-based logic and if/then rules?
A constraint states a relationship that must hold and lets the system derive valid combinations. If/then rules enumerate specific paths, so each new option multiplies the branches to maintain. Constraint-based models stay smaller as the portfolio grows.
Should we fix configuration logic before adopting AI in quoting?
Yes, for a mechanical reason. An assistant reads whichever copy of the rules it can reach and produces confident output from it. Inconsistent inputs give you fast, plausible, wrong answers, and the speed removes the hesitation that used to catch them.
Do we have to replace our existing systems to get consistency?
No. An integration layer that maps and moves configuration between systems removes the manual copying without a rip-and-replace. That is the approach we took for Dewulf, moving configurations from Elfsquad into Windchill automatically.
Sources
- Tacton, The State of Manufacturing 2026: Building Toward the Connected Factory. Online survey of 280 manufacturing leaders, fielded April 2026 with Researchscape, eight countries, US 62% and Europe 38%. Vendor's own research.
- IDC InfoBrief, Orchestrating Product Configuration Management: The Secret Weapon of Leading Manufacturers, sponsored by Configit. Survey of 511 discrete manufacturing companies in North America and Europe.
- Product configuration for mass customisation: a systematic literature review, International Journal of Production Research, 2026. 122 articles from 2000 to 2025, PRISMA framework.
- Rockwell Automation, State of Smart Manufacturing 2026, conducted by Sapio Research. 1,560 respondents across 17 countries, company revenues from $100 million to over $30 billion.