Consultation
We respond within 15 minutes
There is a stage in website development that is often invisible to the client. No pages have been designed yet, there are no URLs, there is no navigation menu, and there may not even be a single line of code. And yet one of the most important parts of the project is already happening: the information architecture is being designed.
For a simple website, this can look like an unnecessary exercise. For a serious manufacturing company, or a marketing platform built specifically for manufacturers, it’s something very different. It’s the foundation on which the entire information system will eventually operate.
Start with the model, not the pages
When I started designing the new Manufacturing Marketing Agency project, I deliberately didn’t begin by asking what pages the website should have. I started with a different question: what does the manufacturing marketing domain actually contain? This is a conceptual modeling problem.
In software engineering, we don’t normally begin by creating thousands of individual objects without first defining the entities and relationships that make up the system. Information architecture works in a similar way: before we create pages, we need to understand the structure of the information those pages will represent.
Manufacturing marketing is not a simple domain. A manufacturer can be described simultaneously by its industry, its production technologies, its business objectives, its products, its applications, its capabilities, and the marketing services relevant to its situation. Trying to force all of this into one hierarchical tree quickly becomes problematic, so I started with four independent dimensions.
The four dimensions
Industrial answers the question of what kind of manufacturing or industrial business we’re talking about: Automotive Manufacturing, Aerospace Manufacturing, Metal Fabrication, Industrial Equipment, Electronics Manufacturing, Medical Device Manufacturing, Construction Products, HVAC Manufacturing, Furniture Manufacturing, Contract Manufacturing. This dimension describes the industrial context.
Technology answers the question of how the product is manufactured: CNC Machining, Laser Cutting, Laser Welding, Additive Manufacturing, Robotics, Industrial Automation, Injection Molding, Sheet Metal Fabrication, Welding, IoT and Industry 4.0. This is a completely different dimension, because a technology can exist across multiple industries. CNC machining, for example, is relevant to automotive, aerospace, medical devices, industrial equipment, and many other manufacturing sectors, so Technology cannot simply be treated as a child category of Industrial.
Solutions answers the question of what business problem the manufacturer needs to solve: Lead Generation, OEM Customer Acquisition, Market Expansion, New Product Launch, Distributor Growth, RFQ Generation, Engineer & Specifier Reach, Sales Growth, Competitive Displacement, Digital Sales Growth. These aren’t industries and they aren’t technologies, they represent business objectives and problems.
Services answers the question of what the agency can actually do to address those problems: Manufacturing SEO, Technical Content Marketing, Google Ads / PPC, LinkedIn Marketing, Industrial Web Design, AI Search Optimization, Conversion Rate Optimization, Marketing Automation, Account-Based Marketing, Analytics & Attribution. Again, this is a different dimension.
These are not four levels of one hierarchy
This distinction is probably the most important part of the model. I’m not building something like Industrial leading down to Technology, then Solution, then Service, because that would imply every item belongs somewhere below another item, and it doesn’t. Instead, I treat the four dimensions as independent coordinates of the same conceptual space.
For example, a company could be described as Industrial: Automotive Manufacturing, Technology: CNC Machining, Solution: OEM Customer Acquisition, Service: Manufacturing SEO, all at once. The manufacturer exists in one industrial domain, uses certain technologies, has specific business objectives, and different services can be applied to those objectives. This is much closer to a multidimensional model than to a traditional website menu.
But the four dimensions are not enough
This is where another layer becomes important: relationships. Categories by themselves are not the system, the connections between them are.
Consider a CNC machining company serving the automotive industry. There may be a relationship linking Automotive Manufacturing, CNC Machining, Precision Manufacturing, OEM Customer Acquisition, and Manufacturing SEO, and these relationships aren’t necessarily parent-child, they’re semantic. One concept is relevant to another, and one entity can be associated with several others. A case study can demonstrate a particular technology, that technology can matter across several industries, a specific business problem can occur in those industries, and several services can address that problem. This is where the architecture begins to move from a taxonomy toward a semantic graph.
The website becomes a network, not just a collection of pages
Imagine a future website with 200 service pages, 300 industrial pages, 200 technology pages, 100 solution pages, 500 blog posts, 100 case studies, and hundreds of FAQs, applications, guides, and other resources. If these pages are created independently, the result can become a very large collection of disconnected documents. But if they’re built around a common conceptual model, the same content can participate in multiple relationships.
A case study about laser welding for an automotive manufacturer could connect to Industrial: Automotive Manufacturing, Technology: Laser Welding, Solution: OEM Customer Acquisition, and Service: Manufacturing SEO, as well as related applications, products, technical articles, FAQs, and other case studies. Now the website has an internal structure that reflects the actual domain. The content is no longer simply tagged, it’s related, and that distinction matters.
Why this matters for SEO
This is also where information architecture stops being an abstract exercise and becomes a practical SEO tool. Search engines don’t see a website exactly as a human does: they crawl pages, follow links, analyze page content, headings, anchor text, and surrounding context, and they also analyze how pages relate to one another through the site’s internal linking structure. So the architecture of a large website influences how its information can be discovered and understood. A logical hierarchy helps, meaningful internal links help, and clear relationships between related content help. But there’s another important idea here: before keyword research, there’s a conceptual space.
Building the keyword space before keyword research
Keyword research is usually described as the starting point of SEO. In practice, I see it slightly differently. Before asking what keywords people search for, it helps to first understand what concepts actually exist in the domain.
Suppose we identify Automotive Manufacturing, CNC Machining, Precision Manufacturing, OEM, RFQ Generation, Engineer Engagement, and Manufacturing SEO. We haven’t yet performed keyword research, we don’t yet know search volume, keyword difficulty, or the exact wording potential customers use. But we’ve already created something valuable: a semantic space in which keyword research can take place. I think of it as a keyword space, and keyword research then becomes the process of mapping real search behavior onto that conceptual space.
The sequence becomes concept, entity, category, relationship, search terms, content, URL. That’s very different from collecting a spreadsheet of thousands of keywords and then trying to figure out what they mean afterward.
And this becomes increasingly important for AI search
Search is changing. Users are increasingly asking questions in natural language, and AI systems and AI-assisted search interfaces are increasingly retrieving, combining, summarizing, and presenting information from large collections of documents. In that environment, simply having a large amount of content isn’t necessarily enough, the system needs to understand what the content is about and how different pieces of information relate to each other.
Consider concepts like laser welding, automotive manufacturing, OEM supplier, production capacity, RFQ generation, and contract manufacturing. A page containing these words is one thing. A website whose architecture consistently establishes meaningful relationships between them is something else.
The objective isn’t to trick AI, and it isn’t to add artificial tags everywhere. It’s certainly not to assume that a particular taxonomy automatically produces rankings or AI citations. The objective is more fundamental: to make the information structure explicit. The more complex the website becomes, the more that principle matters.
This is not just categorization
This is probably the biggest misconception about this stage of the project. Someone looking at a taxonomy spreadsheet might ask why spend so much time on categories. The answer is that categories aren’t the final product, they’re the model of the domain. They determine what concepts the website is capable of representing, create the framework future content can fit into, provide the structure for internal linking, help organize entities, create the conceptual space for keyword research, and provide a consistent framework through which content can be connected.
Without this model, hundreds of pages can become hundreds of isolated URLs. With the model, those same pages can become parts of a larger information system.
Designing for a manufacturing company is different
Manufacturing websites are particularly interesting because the underlying domain is often much more complex than a typical B2B service website. A manufacturer may have multiple production technologies, multiple industries served, multiple products, multiple applications, different buyer personas, engineering requirements, technical specifications, certifications, production capabilities, geographic markets, OEM relationships, distributors, RFQ processes, and long sales cycles.
The website therefore has to represent a multidimensional business. A simple Home, Services, About, Contact structure isn’t enough for a serious industrial knowledge base. The architecture has to be capable of growing with the business.
Designing for years, not months
This is why I’m approaching the Manufacturing Marketing Agency project as a long-term information system, not a collection of 20 or 30 pages, not a temporary SEO project, and not a pile of keywords converted into URLs. The model needs to support hundreds, potentially thousands, of interconnected pieces of information over time: new industries, new technologies, new applications, new solutions, new services, and new case studies that create new relationships between existing entities. The architecture should allow all of this without having to rebuild the entire system every time the content grows.
The technical implementation comes later
Only after the conceptual model is defined do I move toward the technical implementation. In this project, that means Next.js, React, and Payload CMS. The CMS will eventually represent the entities, content types, categories, relationships, and reusable structures defined by the model, but the technology isn’t the starting point. The starting point is understanding the domain.
The sequence is domain, conceptual model, entities, taxonomy, relationships, keyword space, content architecture, URLs, and only then technical implementation. That’s the order I’m interested in, because a website isn’t just a collection of pages. A serious website is an information system, and for a serious manufacturing company, the information system itself becomes part of the company’s digital infrastructure.
The interesting part is what happens next
Once the four dimensions, Industrial, Technology, Solutions, and Services, are connected through relationships, something much larger starts to emerge. The website is no longer simply organized, it begins to represent a model of the manufacturing domain. That model can become the foundation for content, SEO, internal linking, structured data, search visibility, AI-assisted retrieval, and eventually a much broader digital knowledge system.
The four dimensions are only the beginning. The real architecture is in the relationships between them.
Marian Polyak
Web Developer | SEO & AI Search Optimization | Digital Marketing
Manufacturing & Industrial Marketing
Consultation
We respond within 15 minutes
FAQ for Designing the Information Architecture of a Manufacturing Marketing Agency
Why do you need information architecture before creating website pages?
Because without a model of the domain, pages get built blindly. You first need to understand what entities and relationships actually exist within manufacturing marketing, and only then decide how to represent them as pages, URLs, and navigation. This reduces the risk of the website turning into a pile of disconnected documents a year or two down the line.
How are the four dimensions — Industrial, Technology, Solutions, Services — different from ordinary website categories?
Ordinary categories are usually built as a single tree: section, subsection, page. Here, the four dimensions exist independently and describe the same business context at the same time. For example, one manufacturer can belong to Automotive Manufacturing, CNC Machining, OEM Customer Acquisition, and Manufacturing SEO simultaneously — and none of these categories is a “parent” of any other.
Why not just build one large category hierarchy?
Because industry, technology, business problem, and service are different dimensions of the same object, not levels of nesting. If Technology were made a child category of Industrial, you’d lose the fact that a single technology, like CNC machining, applies across multiple industries at once. A rigid hierarchy simply can’t represent that.
What are “relationships” between categories, and why do they matter?
Relationships are what turn a set of separate categories into a system. It’s through relationships that a case study, article, or service page can belong to several industries, technologies, and solutions at the same time. Without relationships, categories remain just labels; with them, the website becomes a network of interconnected content.
How is a “semantic graph” different from an ordinary taxonomy?
In a taxonomy, every item has one path upward through the category tree. In a semantic graph, the same item — say, a case study about laser welding — can be connected to several entities at once: an industry, a technology, a solution, and a service. This lets the model reflect the actual complexity of the domain instead of flattening it into a single branch.
How does this model help with SEO?
Search engines analyze not just the content of individual pages, but also how pages are connected through internal links. If a website’s architecture is built from the start around clear entities and relationships, it’s easier for search engines to understand the site’s structure and how relevant pages are to each other than if links are placed haphazardly.
What is a “keyword space,” and how is it different from regular keyword research?
Regular keyword research usually starts by collecting search terms and their metrics. This approach starts by defining the actual concepts in the domain — industries, technologies, problems, services — and only then mapping real search queries onto them. The sequence becomes concept, entity, category, relationship, search terms, content, URL, rather than the reverse.
Why does this matter especially for AI search?
AI systems increasingly don’t just find pages, they retrieve, summarize, and connect information from multiple sources. If a website explicitly shows relationships between concepts — for example, between laser welding, automotive manufacturing, and OEM supply — it’s easier for such a system to correctly understand and use the content than if the same words are just scattered across pages with no structure.
Does this mean the taxonomy automatically improves rankings or AI citations?
No, and the article is explicit about this. The four-dimensional structure by itself doesn’t guarantee rankings or AI citations. Its value lies elsewhere: it makes the information structure explicit and easier to navigate, for both search engines and users, but it doesn’t replace content quality or other SEO factors.
Why does the author say he’s designing the architecture “for years, not months”?
Because manufacturing marketing isn’t 20 to 30 pages, it’s potentially hundreds of industries, technologies, solutions, and services that will be added over time. If a flexible model of entities and relationships is established from the start, new content can be added without rebuilding the entire site structure every time the business grows.
Consultation
We respond within 15 minutes
Comments
the site has filters by industry, by technology, by problem, and everything is easy to find.
The complexity of the architecture stays hidden inside, what shows up on the surface is just easy navigation.
The whole model runs under the hood, in the CMS data structure.
Aren’t you worried about inventing categories nobody actually searches for?
Keyword research afterward either confirms a category or shows there’s no demand for it and it can
be merged with a neighboring one. But starting from search volume means you skip concepts that
get searched rarely but matter a lot in a B2B sales cycle, specific certifications for example.
content get picked up in AI answers?
On other sites with a clear entity structure I’ve seen pages get cited in AI overviews more often,
but that doesn’t prove a direct causal link. I’m tracking it and will report back in future posts.
Solutions, Services, Case Studies and so on, with two-way connections. Payload handles that natively
and works well with Next.js. WordPress can be pushed there too through ACF, but the data layer ends up
less structured.
than one being fundamentally better for this specific task.
descriptions, that’s a few months for the first wave. Solutions and the relationships between everything
get added in parallel. A full network with hundreds of case studies and posts is a year or two of gradual
growth, not a single sprint.
while specific distributors and locators are a separate entity connected to Industrial and geography.
So it’s not a fifth dimension, it’s a layer on top of the existing four.
relationship fields, the “related content” blocks on those pages pull in programmatically. We only step in
manually in rare cases where the automatic logic surfaces something not really relevant.
dimension, Markets or Geography for example, without breaking the existing four, because relationships
live separately from the categories themselves. That’s really the point of this approach, flexibility
for scale.
to a regular services website?
similar niche structures shows faster growth across long-tail queries because of the larger number of
relevant pages, but I’ll share exact figures once there’s real data.
separate sections with no explicit relationships between them. None of the ones I looked at build this
as a single entity model with four dimensions and a relationship graph. That’s actually one of the reasons
I decided to do it differently.
not just fancy-sounding terminology. Curious to see the result in six months to a year.
result. A model without data is just theory, the real test comes once there are hundreds of pages and
actual traffic. I’ll write about what held up and what had to change.
How did you land on this, is it a standard practice in information architecture or your own approach?
It exists in SEO and IA work already, but for manufacturing I adapted it to the specific problem: industry, technology,
business problem, and service genuinely don’t nest inside each other. I checked dozens of manufacturer websites
and the same issue keeps showing up.