How to structure a parts website so it ranks for the machines the parts fit
Parts sites are organised the way the stockroom is organised. Buyers search by the machine. The gap between those two facts is where most of the organic demand in a technical catalogue is lost. By Adrian Street, Founder, Arlieo / Updated 13 September 2026
The gap between your structure and their search
A parts website is usually organised the way the stockroom is organised. Filters sit under Filters. Bearings sit under Bearings. It is tidy, it is logical, and it is not how the buyer searches. The buyer searches for the machine.
Somebody standing next to a broken machine knows two things for certain: what the machine is, and what has failed. They rarely know your part number and they almost never know your category naming. So they type the make, the model, and the component. If no page on your site is about that machine, Google has nothing specific to return, and the click goes to whoever built a page that is.
Multiply that across every model you serve, every application and every superseded number, and the missing demand is substantial. It is also invisible in most reporting, because you cannot see impressions for pages that do not exist.
Four layers that a parts catalogue needs
Rather than replacing your category tree, add the layers the buyer’s search needs and connect them to it.
1. Component categories
Keep these. They serve the buyer who knows what part they want and wants to browse the range. Give them a short piece of copy that explains selection, fitment and the specifications that matter, where that helps somebody choose. Not a block of prose on every category for its own sake.
2. Machine pages
One page per make and model that matters, listing every part you stock for it, grouped by service area. This is the page that answers the search the buyer actually made. It is also the page most parts sites do not have.
Build it from a template so the range can be extended without a new design decision each time, and prioritise by demand rather than by catalogue size. A hundred well built machine pages for the models people search for will do more than a thousand thin ones generated for every model in the database.
3. Application and problem pages
Some buyers search by the job rather than the machine or the part. What they need is a page that explains the selection, names the specification that decides it, and links to the products that fit. These pages tend to be few and valuable.
4. Product pages
The product page carries the part number, the equivalents and supersessions, the specification, the fitment list and the availability. It should link back to every machine page it appears on. That interlinking is how the structure holds together and how authority moves between the layers.
The technical decisions that decide whether it works
Structure alone is not enough if the site is generating thousands of near identical URLs alongside it.
Facets
Filter combinations often create very large numbers of crawlable URLs that duplicate each other and compete with the pages you want to rank. Decide which combinations serve a real search and should be indexable, which are canonicalised to a parent, and which are kept out altogether.
Variants
Either each variant has distinct demand of its own and deserves its own page, or the variants belong on one page selected by a control. Splitting them arbitrarily divides the signals; collapsing them all can lose the specific searches.
Discontinued products
Handle these case by case. Where there is a real replacement, redirect to it. Where the old page still carries demand or links, it may be better kept with a clear message and a route to the replacement. Where neither applies, retiring it is fine. Deleting everything by default throws away both the ranking and the sale of the replacement part.
Out of stock and seasonal lines
Removing pages each season and adding them back resets whatever they had built. Keep the page, show the position honestly, and offer an alternative or a notification.
Structured data
Product and offer markup with price, availability and identifiers, so the listing carries the information a buyer is looking for before they click.
Where to start
1. Separate the machine queries
Pull your Search Console queries and separate anything containing a model, a make or a part number.
2. See where they land
Look at which pages those queries currently land on. Generic category pages with high impressions and low click through are the clearest evidence of the gap.
3. Rank the models by value
Rank the models by the value of the parts you sell for them, not by search volume alone.
4. Build the template and the first twenty
Build the template, then the first twenty machine pages, and interlink them with the products.
5. Measure by page group
Measure organic revenue by page group, including the orders that start online and finish on the telephone.
Technical fixes and architecture changes can move within a few months. New machine and application pages build over six to twelve. The order matters, and the fastest returns should come first.
Questions
Do we need a page for every machine model?
No. Prioritise by the value of the parts you sell for each model rather than by catalogue completeness. A hundred well built machine pages for models people actually search for will do more than a thousand thin ones generated from the database.
Our platform generates thousands of filter URLs. Is that a problem?
Often. It creates duplicate and near duplicate pages that compete with the ones that should rank, and makes crawling less efficient. Facets need rules: which combinations serve a real search and should be indexable, which are canonicalised to a parent, and which are kept out altogether.
What should we do with discontinued parts?
It depends on the product. Where there is a genuine replacement, redirect to it. Where the old page still has demand, links or usefulness, it may be better kept with a clear message and a route to the replacement. Where none of that applies, retiring it is fine. Deleting everything by default throws away rankings and the sale of the replacement part.
Do we need blog content as well?
Rarely as a priority. For a technical catalogue the demand sits in product, category, compatibility and application pages. Articles help when they answer a genuine buyer question and link to the products that solve it.
How long before this works?
Technical fixes and architecture changes can move within a few months. New machine and application pages build over six to twelve. The order matters, so the fastest returns should come first.