Published: 29 May 2026 by Dan Williams
Series: Notes on eBay & Discovery
Notes from selling stamps on eBay and reading a few engineering papers.
Prefer a condensed version of this article in video format? View eBay Search Still Matters – But It’s Not the Whole Story at YouTube
For many years, online visibility was largely discussed in terms of search.
Whether you ran a website, a forum, a blog or an online shop, the basic model was familiar. Someone searched for something, the platform matched those words against its database, and the most relevant result appeared.
Search still matters and in some situations it does so enormously.
As a stamp dealer, many of my buyers arrive through highly specific searches. Sometimes that might be a Stanley Gibbons catalogue number, a particular variety, a watermark error or an obscure item. Those buyers know exactly what they are looking for and search remains the most efficient way to connect them with the right item.
But over the last couple of years I’ve found myself paying increasing attention to something else.
Not search alone.
Recommendations.
From Search Results to Discovery Paths
One thing I’ve noticed while buying and selling on eBay is that buyers don’t always seem to follow a simple path.
The traditional model might look something like:
Search → Click → Buy
In reality, many collector journeys probably appear more like:
Search → Click → Browse → Compare → Watch → Return → Buy
Or:
Search → Click → Browse → Buy several related items
The first click often arrives via search. What happens afterwards increasingly appears to be driven by recommendations.
Sections such as:
- People who viewed this also viewed
- Customers also considered
- Similar items
- Frequently bought together
are no longer decorative additions to a listing page. They are part of the discovery process itself.
The more I observe what I think I see, the more I suspect these recommendation pathways are becoming just as important as search results.
The Curious Case of Specialist Inventory
This becomes particularly interesting in niche categories.
A collector looking at a Machin variety may not simply want that one item. They may be interested in related varieties, different denominations, phosphor errors, paper types or other specialist material.
Once a buyer enters an area of interest, the platform appears to work hard to keep them there.
I’ve noticed this repeatedly when checking listings in private browser sessions with no account history, prior cookies or personalisation. Certain groups of related items seem to reinforce each other. Buyers enter through one listing and are repeatedly exposed to other similar items – a sort of echo chamber effect?!
The exact mechanisms behind this are obviously not public.
But the behaviour itself is observable.
A Possible Explanation
Part of what led me down this route was reading some of eBay’s engineering articles.
One article in particular, eBay’s Blazingly Fast Billion-Scale Vector Similarity Engine, describes systems that use embeddings, vector similarity and nearest-neighbour retrieval across billions of listings.
I’m not an engineer and don’t claim to understand every detail.
However, reading these articles helped explain something important.
Modern marketplaces are not simply matching keywords against databases.
They are also trying to match relationships.
Items may be linked because:
- buyers interact with them together
- images appear visually similar
- titles and attributes occupy similar semantic spaces
- recommendation systems repeatedly observe the same behavioural patterns (like co-visitation or item co-purchases)
In other words, items may be related for reasons beyond exact keyword matching.
Search still exists. But it increasingly appears to sit alongside recommendation systems built around similarity and behaviour.
Read more at eBay:
eBay’s Blazingly Fast Billion-Scale Vector Similarity Engine
Why This Matters
Search Google for eBay SEO or How to increase eBay sales and you’ll find lots of advice which still sounds like it belongs to an earlier era:
- Use more keywords.
- Fill all available title space.
- Refresh listings constantly.
- List every day.
Some of that advice may still be useful.
But this is treating listings as isolated objects.
My observations make me suggest the opposite.
Listings exist within clusters.
Buyers exist within clusters.
And recommendation systems appear to spend a lot of time connecting the two.
A Thought About Low-Value Items
One of my different ways of thinking now is that not every listing’s value will come from the price it sells for.
Occasionally a low value item appears to have functioned as an entry point.
A buyer discovers one item and then goes on to purchase several related items. Likely with increased browsing and sometimes with adding to watchlist.
In that scenario, the value of the first item may be larger than the sale itself.
That is a subject I’ll likely explore further in a future article.
Closing Thoughts
Search is not dead!
If anything, search remains critically important for buyers with clear intent.
But the more I think about how collectors browse, compare and discover items, the more I suspect recommendations play a larger role than many sellers realise.
Perhaps the most interesting part is that these recommendation systems are no longer hidden away in the background.
They are increasingly shaping how buyers move through a marketplace.
Once you start noticing this type of behaviour, it becomes surprisingly difficult to stop seeing it.
In my next article, I try to explain Why a £1 Stamp Can Be More Valuable Than a £50 Stamp 😄
— Dan Williams
Read More
This article forms part of Notes on eBay & Discovery: a collection of my observations about visibility, recommendations and buyer behaviour on modern marketplaces.