The commercial problem
What a business can deliver is often larger than what it can sell online.
Businesses can often deliver more than they show online. They may support different materials, dimensions, configurations, modifications, production methods and supplier options. Yet the webshop presents only the products that have already been defined, priced, photographed, described and published.
When a customer cannot find an exact fit, the interface commonly concludes that no answer exists. The business itself may never have made that decision.
Traditional ecommerce is organised around products that already exist in a catalogue. That works well when the customer wants one of those products. It works less well when the customer begins with a requirement, a room, a technical constraint, a combination of needs, or an outcome nobody has published yet.
What follows is not one conversion problem. It is a connected set of commercial failures across discovery, customer intent, merchant knowledge, custom sales, digital visibility and fragmented fulfilment, and each one makes the others harder to solve.
One
Ecommerce is shaped by how products are represented, and a product has to travel a long way before it becomes commercially visible.
Before a customer can buy something online, it usually has to be:
Every one of those steps is work, and work has to be justified by expected volume. So a business publishes the combinations it expects to sell often, and leaves the rest unpublished. That is a rational decision. The problem is what it looks like from the outside: identical to not being able to do it at all.
Deliverable capability
Published catalogue
Catalogues are useful. They give customers clarity, comparability, speed and confidence, and most purchases should stay exactly that simple. The problem is not that catalogues exist. It is that the catalogue gets treated as the complete boundary of what the business can answer.
Nor is everything outside the published set possible. Capability is governed by real constraints: what a machine can hold, what a material tolerates, what a supplier stocks, what an engineer will sign off. The gap between published and deliverable is not a list of free options. It is a set of possibilities that nobody has evaluated.
A missing product page does not necessarily mean a missing capability.
Two
Customers do not always evaluate products one category at a time. Somebody furnishing a room needs furniture, lighting, textiles, storage, delivery, installation and occasionally advice, and they would rather solve that once than seven times.
For many multi-product purchases, completeness and convenience influence the decision before individual product quality is fully considered. A larger platform can offer one search, one account, one checkout, coordinated delivery, simpler returns and a clearer sense of progress. A specialist merchant may hold the better product in one category and still lose the broader purchase.
Competing with a large platform is not only a matter of price. It can mean competing with:
None of this makes large platforms harmful. Their convenience solves a real customer problem and they earned it. The open question is whether specialist merchants can offer comparable completeness without surrendering their identity or turning themselves into generalists.
A specialist product may lose before its specialist advantage is considered.
Three
Discovery works by matching customers to products already in the catalogue. The customer is expected to pick a category, use the merchant's terminology, select filters and choose from supported configurations.
When the combination they need does not exist, the interface says no results, this combination is unavailable, out of stock, or contact us. Those messages describe the state of the interface. They do not describe a decision the business has made.
The merchant might have been able to:
But the interface rarely asks any of the questions that would establish that:
What happens now
What the request needed
The second sequence is what the first one is missing. It is a description of the problem rather than of a shipped feature.
High-intent customers are often lost at the moment their needs become most informative.
Four
When the webshop cannot answer, the request moves into email, phone calls, contact forms, spreadsheets, shared documents and individual employees' memory.
What that produces, reliably:
Human involvement is not the inefficiency. Expert judgement is frequently the most valuable part of the process and some requests should never be answered without it. The waste is that experts are repeatedly required to reconstruct information that could have been structured the first time somebody established it.
The objective is not to remove expertise. It is to use expertise where judgement is required rather than where information is merely missing or scattered.
This has a commercial consequence that is easy to miss. A request can be perfectly deliverable and still get declined, because the cost of understanding, pricing and coordinating it exceeds what the order is worth. The refusal looks like a capability limit. It is actually a processing-cost limit.
Some custom demand is rejected not because it cannot be fulfilled, but because it is too expensive to process manually.
Five
Ecommerce analytics are good at what happened inside the interface and weak at what the person wanted.
| Search terms entered | Tells you the words used, not the requirement behind them. Somebody may search a product name while trying to solve a much broader problem. |
|---|---|
| Filters selected | Tells you which options were available to click, not which dimension, material or date was actually essential. |
| Zero-result sessions | Tells you a search failed. Not what was needed, whether a near match would have been accepted, or what the customer did next. |
| Abandoned sessions | Tells you somebody left. Not whether they left because the price was wrong, the lead time was impossible or the product simply did not exist. |
A clickstream shows what the customer did inside the interface. It does not necessarily show what the customer was trying to achieve.
Without structured unmet demand, a merchant makes decisions using only the demand the catalogue was capable of receiving. That can quietly distort:
The demand that disappears without a transaction can still contain valuable commercial information.
Six
A merchant's real capability is usually distributed across founders, salespeople, craftspeople, engineers, suppliers, drawings, price lists, old emails, spreadsheets and production systems.
That knowledge typically covers:
The knowledge is usually accurate. It is simply not reachable. A customer cannot query it, employees have to explain it again each time, and when the person who holds it is unavailable, the business looks less capable than it is.
Making that knowledge usable is not a matter of copying notes into a database. To be worth relying on, it has to be:
The webshop sells what has been published. The business may be able to deliver what has only ever been known.
Seven
Customers increasingly research and describe what they need through AI assistants, AI-enhanced search and conversational interfaces. That is becoming another route by which merchants and products get found.
An AI system cannot represent a business accurately when the information it needs is missing, vague, contradictory, out of date, buried in an image, spread across unrelated pages, or expressed in terminology nobody else uses. A merchant can have an excellent product and still be difficult to describe correctly.
The facts most often absent from a product page, and most often needed to answer a real question:
When those are absent, an assistant may:
Access is a separate question from description. A site can be hard to interpret when:
This is not an argument that JavaScript blocks AI systems. It is a question of whether the information a merchant intends to be found stays accessible, stable and interpretable across the channels they want to sell through.
The goal is not to manipulate AI systems. It is to make the merchant's real products, capabilities and limits easier to understand accurately.
Different AI systems obtain and use commercial information in different ways, most of them undisclosed and all of them subject to change. Clear, accessible, structured information improves the possibility of accurate representation. It does not guarantee inclusion, citation or recommendation, and anybody selling that guarantee is selling something they cannot deliver.
There is a harder point underneath this, and it is the reason the section exists at all. Even a perfectly described catalogue exposes only the products already published. A merchant can improve every page and leave the rest of the business invisible. So there are two separate questions:
AI visibility is not only a distribution problem. It is also a capability-representation problem.
And a boundary worth stating plainly, because it is where this gets dangerous:
Making a business understandable to AI is useful. Making the answer trustworthy requires authoritative capability, constraints and approval.
Eight
Every independent merchant has to support some version of the same list: product information, photography, content, search, navigation, filters, configuration, checkout, analytics, integrations, support, security, machine-readable data and continuous maintenance.
A large platform spreads that investment across many merchants, enormous catalogues, high transaction volumes and dedicated engineering teams. An individual specialist cannot match it alone, and mostly should not try.
This is not a claim that small merchants have less to offer. They frequently have:
Specialist capability does not automatically produce specialist discoverability.
The waste here is cumulative duplication. Many businesses separately fund similar infrastructure, each arriving at something less complete than the platforms they are competing with, while the thing that actually distinguishes them stays invisible through the interface.
Nine
One merchant may satisfy one part of a broader request. A table, or the lighting, or the textiles, or the installation, or the bespoke component, or the technical review. Rarely all of them.
Today the rest of that demand is usually:
What merchants lack is not goodwill. It is a commercial system for:
Complementary merchants often compete separately because they lack the infrastructure to fulfil demand together.
None of this is simple, and it would be dishonest to present it as a matter of recommending another merchant. A workable system eventually needs agreed rules for:
The opportunity is not to refer work elsewhere. It is to make complementary capability commercially actionable while letting specialists stay specialists, which is a governance problem at least as much as a technical one.
One request
I need a table for twelve people, made for this room, in a durable material, delivered before September.
The catalogue
Six, eight and ten seats are published. Twelve is not, because it sells rarely enough that nobody photographed it.
Search and filters
The filter combination returns nothing. The interface reports the absence of a product, not the absence of a capability.
The contact form
The customer writes a paragraph. Nothing in the form asks about the room, the material tolerance or the September deadline, which are the three facts that decide the answer.
Sales
Somebody reconstructs the requirement by hand. They know which materials hold at that length and which supplier could meet the date, because they have done it before.
Afterwards
Whether the order lands or not, the requirement, the constraints that governed it and the reasoning behind the answer stay in a mailbox.
An AI assistant
It reads the published page, finds a ten-seat table, and has no way of knowing the business makes longer ones on request.
The complete purchase
The customer also needs seating and installation. A larger platform covers more of the job, and the specialist loses a sale it was better qualified to fulfil.
Every step in that sequence is a reasonable decision by somebody. The failure is structural rather than anybody's mistake, which is exactly why it does not get fixed by trying harder.
The connection
Each of the nine makes the others harder to solve. That is what makes the whole thing durable, and why fixing any single one of them tends to disappoint.
A sequence rather than a cycle. Each step makes the next more likely; none of it is automatic, and the chain can be broken at several points.
Each failure makes the others more difficult to solve.
What it costs
High-intent requirements reach dead ends.
Demand beyond the catalogue is lost rather than declined.
Flexible production capability stays commercially invisible.
Published products get treated as the limit of possibility.
Businesses can end up selling only a fraction of what they are capable of delivering.
The distinction
That is the single sentence underneath all nine problems. The interface is organised around the products that exist, which is why it fails whenever the customer starts somewhere else.
A commercial interface built the other way round would start from a different set of questions:
What that looks like as a product, and which parts of it exist today, is the subject of the Vision page rather than this one.
The commercial interface should represent not only products, but demand, capability, constraints and decisions.
A customer can fail to find a product even when the business could have answered the need. A specialist can lose a sale even when its product is better suited. A merchant can receive valuable custom demand and gain no structured knowledge from it.
An AI assistant can overlook or misrepresent a business whose products and capabilities are hard to interpret. Several complementary businesses can each hold part of the answer while the complete purchase goes somewhere else entirely.
None of these are isolated website defects. They are what happens when commerce is organised around a fixed published catalogue while productive capability, customer intent and digital discovery all become more dynamic than the catalogue can express.
The catalogue is smaller than the business behind it.
That gap is where the commerce goes.
What follows
The Vision page sets out what Specify is building in response, and where the boundary is between what exists today and what does not. The thesis explains why the commercial model is changing underneath all of it.
Or start with one difficult request. The complimentary Enquiry Audit takes a real enquiry the catalogue could not answer and shows what a structured reading of it produces.