FP-011
Open a product page for a pair of tall riding boots.
The specifications include:
Calf: 36 cm
Nothing beside the measurement explains what is being measured or why it matters.
A customer familiar with tall riding boots probably understands the measurement. They know where the boot sits on the leg and why calf circumference matters to fit.
We bring that kind of experience to product pages all the time, but AI does not necessarily interpret the same information in the same way.
This publication investigates why product information that makes sense to us can be harder for AI to understand and use.
Product catalogues have traditionally been written and organised around people.
We already know things the product page never needs to explain.
Someone buying tall riding boots knows why calf measurement matters. Someone replacing a filter may recognise the part used by the product they already own. A customer looking at a 20 cm saucepan understands what the measurement refers to.
We also recognise familiar terminology and use what we already know about a product or category to interpret what we see.
That ability can compensate for information that is not always presented consistently.
A footwear retailer may describe width as Fit on one product and Width on another. Customers familiar with the products may understand both.
An electronics retailer may mention compatibility halfway through a description rather than maintaining a consistent compatibility field. A customer searching for a particular model number can still find the answer.
A food retailer may identify Gluten Free as an attribute on one product while another includes the same information only in the description or ingredients.
AI may not interpret the same information in the same way. A compatibility relationship buried in a description, a product characteristic described several different ways, or an intended use shown only in an image requires the reader to make connections between the information provided. People often make those connections using what they already know.
For catalogue teams, this changes how a product page needs to be assessed. It is no longer enough to ask whether someone familiar with the products can understand the page; the catalogue itself needs to state important product information clearly.
AI depends on what the catalogue actually says because it cannot be assumed to understand product information in the same way we humans do.
So why can product information that makes sense to a customer be harder for AI to understand and use?
Consider a customer looking for a replacement battery for a cordless power tool.
They know the brand and roughly when the drill was purchased. The battery currently attached to it has a model number printed on the side.
They open the retailer’s website and find several batteries from the same manufacturer.
One product title contains the battery model.
Another mentions the voltage and battery platform.
The description says that the battery is suitable for “current 18V tools in the X-Series range.”
A photograph shows packaging carrying several familiar tool models.
The customer owns one of those tools and concludes that they have found the correct battery.
Now look at the same page only through the information the catalogue has provided.
The drill model is not listed in the specifications.
The battery platform appears in the description but not as an attribute.
Some products call the platform X-Series while older pages use X System.
The photograph contains several compatible model numbers that do not appear anywhere in the written product information.
The customer was able to bring those clues together.
The customer did not make the decision from the product page alone. They also used what they already knew about the tool, the battery and the manufacturer’s product range.
Customers Bring Knowledge to the Product Page
Open the product page for a pair of riding boots.
The specifications include:
Calf: 36 cm
There is no further explanation beside the measurement.
A customer familiar with tall riding boots probably understands that this refers to the circumference of the calf and that it is relevant to choosing the correct fit.
Another customer may have bought the same brand before and already know how its boots fit.
The page does not need to teach either customer what a calf measurement is before they can use it.
Consider a replacement filter for a pet drinking fountain.
The product page gives the filter dimensions and shows the filter beside several fountains in the product photography. The description refers to it as a replacement filter for “selected models.”
A customer who owns one of those fountains may recognise the filter immediately from its shape, dimensions and packaging.
But the page never lists the compatible fountain models.
The catalogue has not actually stated which products the filter fits.
The same thing happens with homewares.
A saucepan is listed as 20 cm. The customer understands that the measurement refers to diameter because that is how saucepans are normally described.
The Same Product Fact Can Appear in Different Forms
Look across the footwear section of a multi-brand retailer.
One product lists:
Width: Wide
Another lists:
Fit: Wide
A third describes the shoe as:
Wide fit
A customer understands that all three are describing the fit of the shoe.
The same inconsistency appears elsewhere in a catalogue.
One food product carries Gluten Free as a dietary attribute. Another says Gluten Free in the description. A third includes the information only with the ingredients.
Furniture specifications show the same pattern.
One supplier provides:
Overall Width: 900 mm
Another provides:
Width: 90 cm
A third includes the measurement only in a dimensions diagram.
Some Relationships Are Visible Without Being Stated
Consider an outdoor furniture catalogue.
A customer is looking at a modular sofa. While browsing the same range, they find a matching ottoman, side table and replacement cushion.
The products use the same range name, the timber finish looks identical and the cushion fabric appears to match.
The replacement cushion description mentions the sofa, but the sofa page does not identify the replacement cushion. The side table shares the range name but no other relationship is stated.
An electronics catalogue may do the same thing with compatibility.
A wall mount description lists several television models in a paragraph.
Another compatible television links to the mount as an accessory.
A third television is compatible according to the manufacturer’s latest documentation but neither product page records the relationship.
In cosmetics, one moisturiser lists Sensitive Skin under skin type. Another says in the description that the formulation is suitable for easily irritated skin.
It would be easy to conclude that the catalogue simply needs more information.
But a saucepan page does not need to explain what a saucepan is or why its diameter matters.
A footwear page cannot reproduce everything a customer has learned from years of buying and wearing shoes.
We bring that knowledge with us.
The situation is different when the catalogue knows that a battery fits particular tool models but lists those models only in the packaging photograph.
Or when a replacement cushion fits three sofas but that relationship exists only in the knowledge of the catalogue team.
The catalogue cannot reproduce everything we know. The question is whether it already knows something important about the product but leaves the customer to work it out.
A customer can often do that.
They may know that two terms describe the same characteristic. They may recognise a compatible product from its photograph. They may understand an abbreviation because they know the category.
Someone reviewing the catalogue may do exactly the same thing.
When they read Calf: 36 cm, they already know what the measurement means. They may also recognise how products carrying the same range name are related or understand an abbreviation without needing it explained.
This is why someone familiar with the products may not notice what the catalogue leaves unexplained.
AI may not interpret the same information in the same way.
If compatibility matters, list the compatible products rather than leaving the relationship to be inferred from a photograph or description.
If a product is intended for dry skin, state that clearly rather than leaving the reader to infer it from the formulation.
If an attribute matters when products are compared, describe it consistently across those products.
We do not rely only on the information written on a product page. We also bring our own experience with products to what we read.
Catalogue teams bring even more because they know supplier terminology, compatibility, previous product ranges and which products work together. AI may not interpret catalogue information through that same experience.
If a product fits particular models, the catalogue should identify them clearly.
If a product is intended for a particular use, that information should be stated clearly.
If an important characteristic appears across many products, it should be described consistently.
This does not mean explaining everything people already know about products. It means stating important product facts and relationships clearly when the catalogue can provide them.
AI depends on what the catalogue actually says because it cannot be assumed to understand product information in the same way we humans do.
Catalogue teams spend every day working with their products.
That expertise is necessary, but it can make missing product information or relationships harder to notice.
Ask someone who has managed an equestrian catalogue for years whether a particular rug liner works with a particular turnout rug and they may answer immediately.
Then ask where the catalogue states that relationship.
It may be in the description of the liner, in a buying guide or in an old supplier document. Or everyone on the team may simply know it.
That makes it easy to read a page and fill in something the page never actually says.
Visitors to the catalogue may also bring their own knowledge. Some know the brands, some understand the terminology and others have bought the product before.
AI gives catalogue teams another reason to ask what the catalogue actually says, rather than what everyone working with it already knows.
A useful test is to look at a product page without supplying the missing context ourselves.
Does the page identify the relevant product characteristics clearly?
Does it use terminology consistently enough to understand what those characteristics mean?
Where compatibility or another product relationship matters, is that relationship actually expressed?
We already know things a product page does not need to explain.
We understand familiar measurements, recognise products we have used before and can make sense of information presented in different ways.
AI may not interpret the same information in the same way.
This does not mean a catalogue needs to explain everything we already know. But when the catalogue knows which products are compatible, what a product is intended for or which characteristics matter, it should say so clearly.
AI depends on what the catalogue actually says because it cannot be assumed to understand product information in the same way we humans do.
Keeping product information current is one responsibility. Making sure that information can be understood and used by customers, search systems and AI adds another.
Managing a catalogue involves more than publishing products and updating information when something changes. Catalogue teams also need to decide what information matters, how it should be presented and how the catalogue should be maintained over time.
What happens when managing a catalogue becomes a professional responsibility in its own right?
That is the question explored in FP-012 — Catalogue Optimisation Has Become a Profession.