We considered naming this article “the sort type all users want but zero sites offer” because category-specific sorting really is one of those rare instances where an obvious feature has somehow been completely overlooked by the e-commerce community. After all, it really shouldn’t come as a surprise that users would like to sort a list of TVs by “Screen size” or a collection of hard drives by “Storage capacity”.
During our research study on product list usability the test subjects repeatedly sought out these kinds of category-specific sort types – however, to no avail since seemingly zero sites offer them. Even after benchmarking the product lists of 50 major e-commerce sites we have yet to find a site that truly offer category-specific sorting.
In this article we’ll outline our research findings on why category-specific sorting is so important to the user’s product finding abilities, and how it can be implemented.
Given that category-specific sorting is such an uncommon feature, let’s quickly define the term. By “category-specific sort types” we mean any sorting options that are only available in one or a few select categories (because they wouldn’t make sense as site-wide sorting options – they only make sense for the products within those particular categories). Examples include being able to sort suitcases by volume, fishing rods by length, pens by point size, hard drives by storage capacity and spindle speed, road bikes by weight, etc. It’s this category specificity that makes them different from common site-wide sort types such as ‘Best Selling’, ‘Relevance’, ‘User Ratings’, and ‘Price’, which are typically available for all products and categories throughout a site.
The differences between filtering and sorting may on the surface appear subtle. Indeed, many users mix up the terminology or use the terms interchangeably. However, the differences are in fact quite significant: Filters set the criteria for whether a given product is in- or excluded from the product list (i.e. what is displayed) whereas Sorting determines the sequence of those products (i.e. how it is displayed).
This makes Filters great for setting “hard” boundaries while Sorting is optimal for applying “soft” boundaries:
Our product list usability testing showed that users need both hard boundaries (i.e. filters) and soft boundaries (i.e. sorting) for optimal control over the product list. Depending on the user’s knowledge of the product domain and their own needs and restrictions they may have very strict and 100% clear criteria the product must fall within, in which case “hard” boundary filters are appropriate.
However, in many cases users don’t have that strict criteria and sometimes they don’t even know what an appropriate filtering range would be, in which case “soft” boundary sorting is more suitable. For instance, the user might care about a product attribute without having a specific criteria for it, or they may not have the necessary domain knowledge to set meaningful cut-off points.
Let’s take an example of a user with a simple purchase preference: “I would like a lightweight road bike and my budget is $2,000.” Now, that wouldn’t be an unusual or odd request in a road bike store, but it would be a very difficult product-finding task at the 98% of e-commerce sites that don’t offer category-specific sort types, because the “lightweight” preference isn’t suited for a “hard” boundary filter (but instead needs the soft boundary of a sort type).
Let’s play out the scenario to illustrate the issue. Note in the following how you can replace “bike” with your main product category and “weight” with any one of the primary numeric product attributes for that product category (e.g., capacity, power, speed, size, duration):
Now imagine in step #3, if the user had a category-specific sort type for “Weight”. If instead of applying “Weight” as a filter the user sorted the product list by weight (while keeping the $2,000 price filter in place), they would end up with a list of road bikes priced $2,000 or less sorted by lowest weight. The user would effectively have indicated their “lightweight” preference without having had to define highly precise cut-off points. Finally, the user would have a list of road bikes from lightest to heaviest costing $2,000 or below.
The “soft” boundaries of sorting essentially enable users to apply a preference to the product list without having to worry about striking some perfect magical sweet spot between an overly restrictive and excessively lax filter range. Yet filters remain relevant for things like the user’s specific budget limitation of $2,000 – this isn’t a mere preference. It’s by applying a combination of both “soft” (sorting) and “hard” (filtering) boundaries that the user is ultimately able to get what they are seeking: a list matching both the specific budget limitations of $2,000 while also having the “lightweight” preference applied.
During usability testing the subjects often preferred the “soft” boundaries of sorting over the “hard” boundaries of filters. The road bike example illustrates why users in some cases prefer “soft” boundaries over “hard” ones – it alleviates them from defining cut-off points. Further investigation reveals 3 reasons users prefer the “soft” boundaries of sorting:
All of this is typically provoked by an information dilemma: the user has to set the “hard” boundary criteria before seeing the available products. So unless the user already has very good information about the product domain and their own preferences (prior to visiting the site) they often feel unconfident in applying such strict criteria. Few users want to fine-tune a product list if they don’t feel confident in accurately predicting the consequences of this fine-tuning.
Sorting provide users a way to fine-tune the product list to their preferences without running the risk of accidentally eliminating relevant items. It thus helps solve one of the fundamental challenges users have when browsing an e-commerce site: somehow finding all the products that are relevant to their specific needs and preferences out of the site’s typically enormous product catalog.
In order to attain this highly curated list the user must narrow the product list down to a manageable size without discarding any relevant products in the process – they must at once be sufficiently aggressive in their filtering to reach a manageable product list without being so aggressive that they filter out relevant items in the process.
When the user can only filter and not sort by category-specific attributes they are limited to setting “hard” boundaries only. This leaves the user with two less-than-ideal options: either 1) apply the preferred range as filters and run the risk of accidental product elimination, or 2) loosen the preferred range to be more inclusive to avoid this elimination at the cost of ending up with a much poorer signal-to-noise ratio (due to numerous less relevant items suddenly being included).
Soft boundaries are very helpful in this regard because they allow the user to mainly browse a certain range without excluding (any potentially relevant) items outside it – enabling users to discriminate rather than eliminate.
“Soft boundary” sorting can provide users with a better understanding of the product vertical and site catalog by revealing different “spec jumps” and “product gaps”. For instance, in the road bike example, it might be that the first three bikes in the sorted list cost $1,800 and weigh in at ~17 pounds (~7.5kg), and then there’s a jump in the list where bikes four through ten are 6 to 8 pounds heavier but also cost significantly less (e.g., weigh 22 pounds but only cost $900). This would help the user better understand the relationship between price and weight within the road bikes domain.
This is obviously very helpful to novice users because their lack of domain knowledge means they don’t instinctively know the natural “spec jumps” in the product vertical and might therefore inadvertently eliminate large clusters of perfectly relevant products. These users will therefore tend to prefer “soft” boundary sorting over “hard” boundary filters in fear of missing out due to a lack of domain knowledge, as explored earlier.
However, category-specific sorting can also help expert users despite them being intimately familiar with the “spec jumps” in a product vertical because it provides insight into the site’s catalog and any “product gaps” there might be in it. After all, just because the user knows a certain range of products exist they don’t necessarily know which of those products the site carries. By sorting instead of filtering, expert users get insight into which product groups the site carries within the vertical (and which they don’t carry).
Category-specific sort types can thus help increase novice users’ understanding of the product category (and any “spec jumps” within it) while simultaneously providing expert users insight into the breadth of the site’s product range (and any “gaps” within it).
Category-specific sort types are appropriate in product categories where the products share one or more numeric attributes that users may have an interest in or preference for – such as the “Display size” of TVs or “Storage capacity” of hard drives. This is particularly true if the attribute is something where users don’t want to apply strict criteria or are unfit at doing so due to a lack of domain knowledge.
The numericality of the product attribute is important because it makes it well-disposed to sorting as the products can be ordered based on the natural progression of the attribute. Compare this to a non-numeric product attribute like ‘Color’ which can’t really be sorted in a sequence logical to most users. It simply doesn’t make sense to sort by discrete product features even if it is technically possible because there isn’t a natural progression or hierarchy for such attributes. For example, DSLR camera lenses can’t meaningfully be sorted by “Lens type” or “Camera mount” (as they have no natural progression or hierarchy), but it could make sense to sort the category by “Focal length” or “Angle of View”.
Many product verticals do, however, have such numeric attributes and category-specific sort types are therefore relevant to a wide range of domains such as consumer electronics, kitchen utilities, office supplies, sports and hobby equipment, hardware, and similar industries. There are some exceptions of course – in particular verticals that are driven mainly by aesthetics, such as home decor and apparel. These verticals generally have only a few or no important numeric product attributes and it therefore won’t necessarily be meaningful to implement category-specific sort types in them (although even in these domains, attributes such as weight and size may be important depending on the exact product type).
Determining which product attributes should be implemented as category-specific sort types is a fairly straight-forward matter: simply take the 3-5 most important numeric category-specific filters available in each category on your site and turn them into sorting options too. Let’s tackle some common questions that typically arise as a result of that statement:
So in summary: We limit the number of category-specific filters to double as sort types to 3-5 attributes to avoid unruly and intimidating sorting drop-downs. Because of this restraint we obviously want to pick the most important product attributes to cater for the most likely use cases on our site. And we implement all of this by tapping into the data structures already in place for category-specific filters.
The beauty of category-specific sort types is that they tap into investments already made in filters, effectively increasing the return on those investments. Because it’s a technical feature, it doesn’t require continuous data collection or curation (beyond what is already being done for the filters), and it’s therefore a one-time investment that permanently increases the value of the continuous investments made in collecting, curating and maintaining structured product attributes.
The most important filters within each unique product category can thus also be utilized as sort types in that category. (It should of course still be a filter as well – this is about giving users the ability to apply both “hard” and “soft” boundaries to product attributes of interest to them.)
The user needs both hard boundaries (i.e. filters) and soft boundaries (i.e. sorting) for optimal control over the product list. When the user has good domain knowledge and a clear understanding of their own needs, “hard boundary” filters are appropriate. However, as we’ve seen throughout this article there are other times where the user doesn’t have this level of domain knowledge or strict criteria for their product needs, in which case “soft boundary” sorting will often be more suitable. Both “hard” and “soft” boundaries thus have valid usage scenarios and this alone justifies their implementation.
A major bonus of implementing both “hard” and “soft” boundary controls (i.e. filtering and sorting) is that users can then combine them. This affords the user a new level of control over the product list, enabling them to wield both “hard” restrictions and “softer” preferences onto the site’s product lists. Suddenly a rather complex statement like “I would like a lightweight road bike and my budget is $2,000” can successfully be applied, with both “hard” restrictions (i.e. “road bike” and “$2,000 budget”) and “softer” preferences (i.e. “lightweight”).
If you have category-specific filters, the data is already there (and if you don’t, you really should implement them!), ready to be reused for category-specific sort types as well. Category-specific sort types are thus a great opportunity to gain even more from the structured data you’ve invested so dearly in collecting, curating and maintaining.
Indeed, this may even be taken to the next level by combining category-specific sort types with scope selection in higher-level categories and on sub-category pages. This combines category-specific sort types with faceted sorting, allowing highly relevant and popular category-specific sort types to be displayed at (higher) category levels where all products don’t actually share the attribute to be sorted by. This would furthermore make category-specific sort types possible in search results despite some of the results not having the sorting attribute.
Sadly, category-specific sort types are a good example of just how neglected sorting is on e-commerce sites – numerous users want these sort types yet no sites offer them! It’s a rare example of an obvious innovation that e-commerce sites have yet to make. (Tip: you can see 46 different sorting interfaces from the top US e-commerce sites we benchmarked for this study.)
Category-specific sort types are ultimately about empowering users with the tools they need to reach the product selection they want and have it presented in a way that suits their unique interests. It’s an obvious opportunity to increase the return on existing product data investments, empower your users and get ahead of the competition.
This article presents the research findings from just 1 of the 850+ UX guidelines in Baymard Premium – get full access to learn how to create a “State of the Art” user experience for product lists, filtering and sorting.
Join 22,000+ readers and get Baymard’s research articles by RSS feed or
Topics include user experience, web design, and e-commerce
Articles are always delivered ad-free and in their full length
1-click unsubscribe at any time
Great article. This would indeed be very helpful feature in online shopping.
A few critiques:
1 – The sort options reflect what the user can actually see on the product display itself. For example, when you sort by price low-high, you can see the progression in price as you scan the list of products. The same is true for all other sort methods. If a website adds new sort methods, it should also add that data to the product display itself. Without this, it will reduce confidence in the results of the the sort action. And presumably, if a user selects a method to sort by, they would feel that information is important enough to display.
2 – If users could sort lipstick by color, that would be an enormously popular feature. Users sort as a way of grouping, or to help them find a particular item if they know where it will be in a list (e.g. sort by name) and there doesn’t need to be a natural progression to be useful.
3 – You underestimate the investment required for this change. The Sort and filter methods are built into the e-commerce platform. It’s out-of-the-box functionality. Building a new type of sort method would require a level of tinkering under the hood that very few e-commerce businesses have the know how to do.
Thanks for your comment Sholom. I generally agree on your sentiments although I’d like to make a few clarifications / distinctions in regards to your three listed critiques:
1) Our research showed pretty much what you are saying here, with the distinction that it is simply anything that is sorted or filtered by (not that can be, but is) which must be included in the list items / product display. This information may therefore be included dynamically to avoid list item clutter / information overload.
2) I think you’re confusing ‘Sorting’ with ‘Filtering’ here. Sorting = the ordering of the products. Filtering = criteria the product must meet in order to be displayed in the list at all. It makes a lot of sense to filter by color (e.g. the user may want to see only ‘black’ and ‘green’ sweaters). However, it doesn’t make much sense to sort by this product attribute because there’s no obvious way to order them and so even if the sweaters could be sorted by some technical aspect of colors it wouldn’t be transparent to the users and therefore it wouldn’t make sense to offer it.
3) If you’ve already been able to implement category-specific filters (which, if you haven’t, you should do first, as mentioned in the article – but agreed, that change is by no means a small investment) then reusing those same data sources to add a couple of additional sort types really are modest changes in comparison.
Thanks again for the thoughtful feedback. Hopefully I’ve been able to elaborate a bit on our reasoning behind these three arguments.
Thanks for your articulate responses. I see you’re point on 2, although I do still think that sorting (not filtering) by color would be helpful to a lot of customers at the website where I work.
I had one other question, if you don’t mind. Most of our visitors search rather than navigate. And the vast majority of our search queries result in items from multiple categories. Do you recommend that the category specific sort methods only appear once the user has already refined the results to a particular category, or do you think it’s still helpful to allow users to sort a list by an attribute that isn’t applicable to many (most) of the products.
You’re right, there might be instances where sorting by color might make sense if e.g. the site has an audience that is generally comfortable with color wheels and thus be able to comfortably navigate a list sorted by such industry standard. The “natural progression” aspect is a general guideline for picking appropriate attributes for category-specific sort types but there will obviously be exceptions depending on the particular category, site and audience in question.
With regards to search, we actually briefly touch upon this in the concluding section of the article. Our recommendation is to combine “category-specific sort types” with “faceted sorting”, another sorting concept which we discuss here: http://baymard.com/blog/faceted-sorting
By combining these two, you’d have the regular sort types presented for search results and then for the 1-2 categories with very high search relevance have separate sections in the sorting widget. These sections should very clearly set the context of the category and then display any category-specific sort types below, while making it abundantly clear that choosing these will apply the category as a scope to the search results. I recommend checking out the “Faceted Sorting” for examples and a more in-depth description.
Thank you, Jamie.
typo! “which one is the lighests”
Fixed – thanks for the tip.
Exactly, thank you. It’s surprisingly difficult to shop for e.g. plastic storage bins by their (approximate) dimensions instead of by quarts/gallons!
© 2021 Baymard Institute US: +1 (415) 315-9567 EU: +45 3696 9567 firstname.lastname@example.org