On Monday, a food processing machinery manufacturer gets an inquiry: a packaging line for a client's new plant, output of 8,000 units per hour, integration with the existing labeling system. The salesperson has to quote a machine made of several hundred components, half of them from external suppliers. By Friday the client will pick a vendor. Most likely the one who replied first.
In theory it's a simple process. In practice, at most Polish manufacturing companies this quote takes five days, involves four people and ends with an email attachment the client will open once. Conversion on such inquiries rarely exceeds 25 percent, because by the time the quote is sent, the competition has already won the client over.
This article describes a system architecture that solves the problem at the source. Not by hiring another salesperson. By orchestrating four layers that have to work together: the Product configurator, the CPQ engine, the supplier catalog and the quote engine. Each of them exists separately on the market. The challenge is integrating them so the whole chain, from inquiry to accepted quote, runs in hours, not days.
Why custom (non-series) manufacturing is a different world from off-the-shelf sales
In an online store, the product has a price on the label. The customer clicks buy, the warehouse ships. In a B2B wholesaler, prices are in the price list, availability in the warehouse system, everything is calculated in seconds. We wrote about this in the context of automation in wholesale trade.
In custom manufacturing none of that works. There's no catalog with prices. Every inquiry is a unique configuration that nobody has priced exactly before. Materials still have to be ordered, and last quarter's price is no longer valid. Subcontractors have lead times that change every month. The client gets a quote, then asks for a variant with a different housing, and the whole process starts over.
This isn't limited to classic machine building. The same pattern shows up with manufacturers of steel structures, custom B2B furniture, specialty packaging, made-to-order electronics, mid- and large-scale solar PV systems, and prefabricated building elements. Anywhere the product is made for a specific inquiry and its cost depends on a dozen to several dozen suppliers.
Four layers that have to work together
An effective quoting system in custom manufacturing consists of four layers arranged in a chain. Each has a different job, a different interface and a different user. Their integration is the foundation. If any one of them works in isolation, the whole process slows down to the speed of the weakest link.
Quoting architecture in non-serial manufacturing
Four layers, one process. Without any one of them, the system falls apart.
Product Configurator
What the client wants. 3D visualization, parameters, variants, exclusions. The input is a question, the output is a specification.
CPQ engine and BOM generator
Business rules, calculation, breakdown into materials and components. The input is a specification, the output is a list of items to order with quantities.
Supplier catalog
A database of 30+ suppliers with price history, lead times, MOQ. Automatic RFQs to selected suppliers, response aggregation, selection of the optimal option.
Quote engine and presentation
A finished quote as an interactive page or PDF. The client accepts online, and you see who opened it, when and how many times.
Below I break each layer down: what it does, who uses it, what its inputs and outputs are, and what specifically changes when it works well.
Warstwa 1: konfigurator produktu
The configurator is the front end for the client or salesperson. It's where an abstract inquiry turns into a concrete technical specification. The client chooses parameters, variants, options. The system won't let them pick a technically impossible combination, because the rules are built into the configurator. No more calls to the engineering team asking “can this be done”.
In industry, the configurator often has a 3D visual layer. The client sees the machine they're looking for, adds modules, changes the housing, rotates the model. It's not cosmetic. Visualization reduces configuration errors by tens of percent, because the client sees what they're ordering. We wrote about configurator ROI in a separate article on return on investment.
The input to this layer is the client's inquiry in the form of a business need. The output is a list of technical parameters that can be passed on for further processing. In a well-built configurator, the client can get from “I need a packaging line” to “X-450 packaging line, Y-12 labeling module, QR code option” in a few minutes on their own, without talking to a salesperson.
Layer 2: CPQ engine and BOM generator
CPQ, or Configure Price Quote, is the brain of this system. It takes the specification from the configurator and converts it into two things: a component breakdown (BOM, bill of materials) and a base price. This is where the business rules live: minimum margin for a client from a given industry, tiered discounts, transfer prices between companies, payment terms depending on order size.
If you're thinking about CPQ for the first time, start with the introductory article on what CPQ is. It includes an interactive calculator showing how much a company loses each year on manual quoting, and five signs it's time for an implementation.
In custom manufacturing, the CPQ engine has to do more than calculate a product price as the sum of its components. It has to generate a list of materials and components that need to be ordered from external suppliers. That's the input to layer three. If the BOM is incomplete or contains errors, the whole chain falls apart: suppliers get inquiries about the wrong components, come back with quotes that don't fit, and the salesperson has to make manual corrections.
Layer 3: supplier catalog and price history
This is the layer that, at most manufacturing companies, doesn't exist as a system. It exists as tribal knowledge in the heads of experienced buyers. One knows that structural steel is best sourced from mill A, because they offer a 5-day lead time and an 8 percent lower price on orders above 2 tons. Another knows that the electric motor for a particular line only comes from company B, because they have service in Poland. A third remembers from last year that supplier C raised prices by 15 percent in Q3, so it has to be factored into the forecast.
This knowledge is a valuable company asset. And at the same time its biggest risk, because it disappears with the buyer when they leave. A supplier catalog as a system turns that tribal knowledge into a database every authorized employee can access. Not just a list of companies with tax IDs and emails. A database with price history, lead time history, MOQ, payment terms, quality ratings, complaint history and certificates.
Automatic RFQ runs on top of this database. The system receives the BOM from layer two and sends inquiries on its own to the three to five best suppliers for each component, based on history and current terms. Responses come back in a comparable format, because suppliers reply through a form or API instead of email. Aggregation and selection of the optimal option happen automatically, with the option of manual correction by the buyer for unusual cases.
This is the most underinvested layer in Polish manufacturing companies. It's usually where the biggest process leaks get sealed.
Layer 4: quote engine and offer delivery
The last layer is the interface through which the client sees your quote. In the classic model it's a PDF attached to an email. That's a 2005 solution that in 2026 costs companies lost orders. The client opens the PDF once, maybe prints it, you don't see what caught their interest, you don't know what stage of the decision they're at, and you have no way to interact other than another “just following up” email.
A modern quote engine delivers the quote as a dedicated, interactive page on the client's domain or a subdomain of your company. The client gets a link instead of a file. They open it on any device, see the quote with visualizations, can expand sections, accept line items one by one or all at once, and comment on specific points. You see when they opened it, how many times, how much time they spent on which section, and where they stopped.
JSON Hub: layer 4 ready from day one
Instead of building your own quote engine from scratch, you plug in JSON Hub as the presentation layer. Every quote becomes a dedicated, interactive page on the client's own domain. You see who opened it, when, how many times, and which section they stopped at. The client accepts with one click, no PDF attachments.
See how JSON Hub worksThis layer is most often overlooked in discussions about quoting automation, because it seems cosmetic. It isn't. Inquiry conversion in custom manufacturing depends largely on how the quote is presented to the client and how much visibility you have into their decision process. A PDF attachment gives zero visibility. An interactive page gives full visibility.
What this changes in numbers
The table below shows the same type of inquiry handled in two ways. The left column is a typical process at a company that uses Excel with supplier price lists and email as the main communication tool. On the right, the same process handled by four orchestrated layers.
The same inquiry: 30 suppliers in Excel vs 4 orchestrated layers
The biggest jump isn't in the quoting time itself. It's in conversion. Going from 20 percent to 45 percent isn't a cosmetic improvement, it doubles effective revenue with the same number of inquiries. For a manufacturer handling 200 inquiries a month with an average value of PLN 80k, that's the difference between 3.2 million and 7.2 million a month. The system pays for itself within a few months.
Where to start if you have 30 suppliers and Excel today
The biggest mistake companies make when trying to implement this kind of architecture is trying to do everything at once. A year-long project that simultaneously builds the configurator, maps CPQ rules, migrates the supplier database and deploys the quote engine. After a year it turns out salespeople are still using Excel, because the new system is too complicated to beat old habits.
The sequence that delivers results in a reasonable time looks different. It starts with layer three, the supplier catalog. Not because it's the most important, but because it delivers value immediately, even without the other layers. Simply having a supplier database with price history, where someone can search “who did we buy S355 steel from in Q2”, changes how procurement works within a week.
Only then, once the catalog is live and people are using it, do you add layer four, the quote engine. It's the fastest way to increase conversion on the quotes you already send. You still send them manually, but the client gets them as an interactive page, and you see what happens with it.
Once these two layers are stable, it makes sense to build layer two, the CPQ engine with a BOM generator. It's the most demanding part, because it requires formalizing business rules that at many companies exist only in salespeople's heads. But once layers 3 and 4 have already shown value, the case for investing in CPQ is much easier to make. We wrote more about implementations themselves in the article on why AI deployments in companies fail, but the same principles apply to CPQ.
The configurator, layer one, is the last step. Not because it's the least important, but because a good configurator needs all the other layers behind it. A configurator without CPQ generates inquiries that can't calculate a price on their own. Without a supplier catalog, it shows the client configurations that can't be built in a reasonable time. Without a quote engine, it spits out a PDF that is no different from what you have today.
Summary in three sentences
Quoting in custom manufacturing isn't a problem of too few salespeople, it's a problem of architecture. Four layers, arranged in a chain, turn a process that takes weeks today into one that takes hours, and double conversion along the way. Start with the supplier catalog and the quote engine, because they deliver value fastest, and only then invest in CPQ and the configurator.
If you want to talk about what this kind of architecture could look like for your company, write to us via the contact form. Tell us how many suppliers you have, how many inquiries a month, and what the average time is from inquiry to sent quote. We'll come back with a concrete breakdown of which layer makes sense to start with for you.
Frequently Asked Questions
Can these layers be implemented gradually, without turning the whole company upside down?
Yes, and it's the only sensible approach. The supplier catalog and the quote engine can be implemented in a few weeks, each independently, and they deliver value right away. CPQ and the configurator require more analytical work, but they can also be done modularly, starting with the most repeatable 20 percent of configurations.
What if we already have an ERP system that partly does this?
ERP usually covers part of layer 2 (calculation) and layer 3 (supplier database) well, but handles layer 1 (client configurator) and layer 4 (quote engine for external clients) poorly or not at all. A modern architecture usually keeps the ERP as the system of record and integrates it with external tools responsible for the configurator and the quote engine.
How much does this cost for a small manufacturing company, say 20 people?
It depends on the scope. Layer 4 via an off-the-shelf SaaS like JSON Hub starts at PLN 500 a month. A basic supplier catalog (a database with price history, no auto-RFQ) can be built as an internal tool for somewhere in the PLN 10-20k range. A full CPQ with a 3D configurator is an investment starting at PLN 60k, but only after layers 3 and 4 are stable.
Does this architecture only work in classic industry?
No. Anywhere you sell a product or service made for a specific inquiry and dependent on many external suppliers, this model works. Steel structures, custom B2B furniture, solar PV, specialty packaging, HVAC systems, construction prefabrication, production line assembly, IT integrators selling hardware from multiple manufacturers. The pattern is the same.
Where does AI fit in?
AI adds a layer of learning from transaction history: automatic supplier recommendations for a specific component based on historical quality and price, predicting whether a client will accept a quote based on similar transactions, and automatically generating technical descriptions for the quote engine. These are all extensions of the existing architecture, not a replacement.
