Next page or endless scroll? Choose how people browse your catalogue
Pagination and loading more serve different tasks. Design for orientation, returning to an item, the end of a list and loading errors.
Article language: English

In this article
A "Show more" button takes up one line in a mock-up. After someone has looked through forty models, it carries a much larger commitment: another batch of results, a way back to items already viewed, and an end they can recognise. If opening a product means starting over, the saved click costs the person another search.
There is no need to begin with a vote on the most modern pattern. Start with the task. Someone seeking a manual for a specific device, a shopper comparing several lamps and a reader looking through photographs may need different ways to browse, even on the same site. The mechanism should support their work, not merely make the list longer.
Pagination divides a collection into pages of results. "Show more" adds content after a deliberate action. Infinite scroll loads more as the person moves down the page. The differences concern control, orientation and returning to a place; the presence of a button alone does not settle all of them. The examples below are design proposals, not claims of measurements that have not been made.
Choose a browsing method for what needs to be preserved
In a catalogue of manuals, a person might want to check several consecutive items and share a particular part of the results. Pagination can provide a reference point: a page in a defined order. It only helps, though, if the current page is clear and there is a way back. A row of numbers alone does not explain the extent of the list.
Do not give numbers a false precision. If the system does not yet know the total number of results, the design should not invent a last page. It can offer a way forward without declaring an end. The reader needs to understand that this is the next portion, not the last one in a counted total.
USWDS distinguishes pagination with a known last page from an unbounded set with no calculated page count. It recommends indicating the current position and making the navigation range recognisable. Your design should reflect the information actually available about the results, rather than copying a complete set of numbers from a component example.
A load-more button can instead keep earlier items on one screen. It is an option when people compare successive items and want to move back up a little. It does not remove the need to decide what happens after several batches. You need to know whether earlier content still exists and whether the person can find its beginning.
With automatic loading, the reader does not make a separate decision to request another batch. This can suit casual browsing, but may be less clear for a task such as "check the whole specified set". Do not assume that continuing to move down means someone wants an unlimited number of items loaded. The footer or information below the list should remain reachable too.
Also consider what happens if someone later changes the number of items per page. A person switches from twenty to fifty. The old page number no longer identifies the same range. The design can return to the beginning or preserve a recognisable item in another agreed way. The new count and contents must agree, and the change should not look like part of the catalogue has disappeared. Record the chosen outcome in the specification, with an example record that can be used to test it.
A catalogue does not need the same mechanism on every screen. A team might use pages for a document list and load more for an inspiration gallery. Consistency here means recognisable actions and predictable state. A uniform appearance does not require forcing the same route on every task.
When the task is to find a specific item, the way a list is divided cannot replace internal search (in Polish). Someone who knows a code should not have to work through seventeen pages to prove their patience. Navigating results and narrowing them accurately solve different parts of the task.
Return to a place the person can recognise
In a hypothetical test, a shopper views a third batch of lamps, opens one model and comes back. They should be able to identify the item they viewed and continue comparing. Restoring the vertical distance from the top is not enough. Different image sizes or changed content can put a different model at the same height.
Define exactly what returns: list conditions, order, the range already loaded and a reference point. Only together do these describe the expected result. The implementer chooses the technical mechanism. The design should state an outcome that can be checked, rather than assuming the browser's back button will recover everything automatically.
MDN describes the scrollRestoration property, which distinguishes automatic restoration of scroll position from a manual mode when navigating history. It concerns scrolling. On its own it does not promise to restore an application's data, filters and previously loaded items.
The test should therefore use both the browser's Back button and any "Back to results" link on the product page. These are two actions that may take different paths. If one returns to a retained set while the other opens the category from the beginning, its label should make that outcome recognisable.
Returning through history is also different from opening a shared address in a new tab. The person who received the link has no earlier browsing history. If the address is meant to identify a specific part of the results, the page should be able to reconstruct it without that history or clearly show a different agreed range. Separate the promise of continuing a session from the promise of an address that can be saved and shared. Testing one path does not replace testing the other, even if both show the same catalogue title.

For saved table views (in Polish), restoring a personal layout is not enough to restore the place where browsing stopped. Likewise, remembering a category selection does not prove that the third batch of a list has returned. These pieces of information can be stored separately, but from the person's perspective they need to form a comprehensible continuation.
If the product being viewed has disappeared or changed position, do not pretend the state is unchanged. The design can retain context and show an update on return. The aim is an honest explanation of the difference. A person should not have to search for a withdrawn item while the system already knows about the change.
This is where list and interaction design with Aeneia Design can help. We can map the move to details, the return and the results update using the same example records. The prototype's behaviour then needs to be connected to the data retrieval and navigation history in the agreed implementation.
Another batch does not prove the list has ended
Several states can appear below a list: loading, an error fetching the next portion, or an actual end. An empty area below the list does not distinguish them. When a request fails, keep the content already available and design a retry for the missing portion. The reader should not have to search again for a product they just saw.
Quickly triggering loading again should not append the same batch twice. The design needs a visible waiting state and a response consistent with what the application can do. After an error, retry should request the missing portion, not automatically the one after it. In a test, compare item identifiers before and after, because two similar photographs can make a duplicate hard to notice. Check the same issue when a late response arrives after a category change: it should not append old content to the new list.
"All results shown" needs support from the system response. No further rows on screen are not enough if a request is still running or a limit has been reached. When a process also uses an agent, a separate check of a fetched list's completeness (in Polish) concerns the technical scope of the retrieval. The interface design should faithfully communicate the confirmed state to the user.
Check what the counter means as well. "60 of 120" may mean sixty products have loaded, but not that sixty have been deliberately viewed. If the list removes earlier portions from memory, the same wording may suggest availability that no longer exists. The label should describe the right quantity, and the behaviour should allow a return in keeping with the promise.
With pagination, a failed move to page four should not imply that the person is already viewing its results. The previous page can remain with an error, or the interface can clearly indicate that the new page is pending. The current page number, its contents and the address should describe the same state. Highlighting "4" above rows from page three creates a false reference point.
A new product added at the beginning of a collection can move other items. The application needs a stable way of traversing its data. The designer does not have to choose a particular query technique, but should request a test that adds and removes a record between batches. The result should expose duplicates and omissions that would make the task harder.
Similarly, catalogue filters (in Polish) change the set to which further browsing applies. After a change of conditions, define the effect on browsing: whether a new list begins, what loaded range is cleared and what the results label says. The old third page may no longer make sense after narrowing the set to seven products.
A working demonstration is not the end of acceptance
A button should be discoverable and operable with a keyboard. After more content loads, the person must be able to recognise it without stepping through the whole earlier list. The design needs agreed focus behaviour, meaning where keyboard interaction currently takes place, and feedback about the result of the action. An automatic list change should not also take control away from the passage being read.
On a phone, test moving down the list, opening an item and returning, using longer titles. Repeat after enlarging the text. This tests more than the appearance of the bottom button. Different card heights reveal whether the application retains a recognisable record or only a previously stored scroll distance.
A public catalogue has another audience: an external search engine. Google explains that its crawler generally discovers addresses through href links and does not click buttons that load content. Convenient manual browsing therefore does not establish that later addresses are accessible to the crawler.
This is not a promise that one pattern ranks better in search results. Acceptance also needs a separate check that the right addresses can be discovered. Pagination can be implemented incorrectly, and loading more can coexist with an accessible route of links. The visual decision and technical conditions for reading need to be brought together for the particular site.
For a shopper returning after several days, a saved-products list may be useful. The guide to saved product lists (in Polish) discusses how to design one. It is a different tool from remembering a place in browser history. Do not promise durable storage of a selection merely because the person could return to the third batch of the catalogue a moment ago.
Compare two mock-up variants using the same products and the same instruction. Ask the person to find a specified model, look at the neighbouring one and return to the first. Then give a separate task of browsing freely, without requiring a specific item to be found. The observations may lead to different choices for those tasks. Fewer clicks do not by themselves prove that an item was found more efficiently, and longer scrolling need not mean satisfaction. Record mistakes and places where the person lost their bearings.
The choice should follow the variant in which the person can find an item and recover their place without error. In a manual catalogue that may be a clear results page with a durable address; in a gallery, deliberate loading with earlier items retained. Each variant needs evidence that the person can return and recognise the end. The reader should be able to continue choosing rather than take a memory test on the list from the beginning each time.










