Faceted navigation best (and 5 of the worst) practices

Wednesday, February 12, 2014

Selecting filters with faceted navigation can cause many URL combinations
Selecting filters with faceted navigation can cause many URL combinations, such as https://www.example.com/category.php?category=gummy-candies&price=5-10&price=over-10

Faceted navigation, such as filtering by color or price range, can be helpful for your visitors, but it's often not search-friendly since it creates many combinations of URLs with duplicative content. With duplicative URLs, search engines may not crawl new or updated unique content as quickly, and/or they may not index a page accurately because indexing signals are diluted between the duplicate versions. To reduce these issues and help faceted navigation sites become as search-friendly as possible, we'd like to:

Background

In an ideal state, unique content—whether an individual product/article or a category of products/articles— would have only one accessible URL. This URL would have a clear click path, or route to the content from within the site, accessible by clicking from the home page or a category page.

  • Clear path that reaches all individual product/article pages

    On the left is potential user navigation on the site (or, the click path), on the right are the pages accessed.

  • One representative URL for category page https://www.example.com/category.php?category=gummy-candies

    Category page for gummy candies

  • One representative URL for individual product page https://www.example.com/product.php?item=swedish-fish

    One representative URL for individual product page

Undesirable duplication caused with faceted navigation

  • Numerous URLs for the same article/product

    • Canonical: example.com/product.php?item=swedish-fish

      Canonical URL for the same product

    • Duplicate: example.com/product.php?item=swedish-fish&category=gummy-candies&price=5-10

      Duplicate URL for the same product

    The same product page for swedish fish can be available on multiple URLs.

  • Numerous category pages that provide little or no value to searchers and search engines), as demonstrated in the following table:

    URL example.com/category.php?category=gummy-candies&taste=sour&price=5-10 example.com/category.php?category=gummy-candies&taste=sour&price=over-10
    category page for gummy candies in the price range $5-10 category page for gummy candies that are over $10
    Issues
    • No added value to Google searchers given users rarely search for "sour gummy candy price five to ten dollars".
    • No added value for search engine crawlers that discover same item ("fruit salad") from parent category pages (either "gummy candies" or "sour gummy candies").
    • Negative value to site owner who may have indexing signals diluted between numerous versions of the same category.
    • Negative value to site owner with respect to serving bandwidth and losing crawler capacity to duplicative content rather than new or updated pages.
    • No value for search engines (should have 404 response code).
    • Negative value to searchers.

Worst (search un-friendly) practices for faceted navigation

Worst practice #1: Non-standard URL encoding for parameters, like commas or brackets, instead of key=value& pairs.

Worst practices:

example.com/category?[category:gummy-candy][sort:price-low-to-high][sid:789]
  • Key-value pairs marked with : rather than =.
  • Multiple parameters appended with [ ] rather than &.
example.com/category?category,gummy-candy,,sort,lowtohigh,,sid,789
  • Key-value pairs marked with a , rather than =.
  • Multiple parameters appended with ,, rather than &.

Best practice:

example.com/category?category=gummy-candy&sort=low-to-high&sid=789

While humans may be able to decode odd URL parameters, such as ,,, crawlers have difficulty interpreting URL parameters when they're implemented in a non-standard fashion. Software engineer on Google's Crawling Team, Mehmet Aktuna, says "Using non-standard encoding is just asking for trouble." Instead, connect key-value pairs with an equal sign (=) and append multiple parameters with an ampersand (&).

Worst practice #2: Using directories or file paths rather than parameters to list values that don't change page content.

Worst practice:

Where /c123/ is a category, /s789/ is a session ID that doesn't change page content: 

example.com/c123/s789/product?swedish-fish

Good practice:

The directory, /gummy-candy/, changes the page content in a meaningful way:

example.com/gummy-candy/product?item=swedish-fish&sid=789

Best practice:

URL parameters allow more flexibility for search engines to determine how to crawl efficiently. 

example.com/product?item=swedish-fish&category=gummy-candy&sid=789

It's difficult for automated programs, like search engine crawlers, to differentiate useful values (for example, gummy-candy) from the useless ones (for example, sessionID) when values are placed directly in the path. On the other hand, URL parameters provide flexibility for search engines to quickly test and determine when a given value doesn't require the crawler access all variations.

Common values that don't change page content and should be listed as URL parameters include:

  • Session IDs
  • Tracking IDs
  • Referrer IDs
  • Timestamp

Worst practice #3: Converting user-generated values into (possibly infinite) URL parameters that are crawlable and indexable, but not useful in search results.

Worst practices:

For example, user-generated values like longitude/latitude or "days ago" as crawlable and indexable URLs: 

example.com/find-a-doctor?radius=15&latitude=40.7565068&longitude=-73.9668408
 
example.com/article?category=health&days-ago=7

Best practices:

example.com/find-a-doctor?city=san-francisco&neighborhood=soma
 
example.com/articles?category=health&date=january-10-2014

Rather than allow user-generated values to create crawlable URLs—which leads to infinite possibilities with very little value to searchers—perhaps publish category pages for the most popular values, then include additional information so the page provides more value than an ordinary search results page. Alternatively, consider placing user-generated values in a separate directory and then robots.txt disallow crawling of that directory. 

example.com/filtering/find-a-doctor?radius=15&latitude=40.7565068&longitude=-73.9668408
 
example.com/filtering/articles?category=health&days-ago=7

with robots.txt:

User-agent: *
Disallow: /filtering/

Worst practice #4: Appending URL parameters without logic.

Worst practices:

example.com/gummy-candy/lollipops/gummy-candy/gummy-candy/product?swedish-fish
 
example.com/product?cat=gummy-candy&cat=lollipops&cat=gummy-candy&cat=gummy-candy&item=swedish-fish

Better practice:

example.com/gummy-candy/product?item=swedish-fish

Best practice:

example.com/product?item=swedish-fish&category=gummy-candy

Extraneous URL parameters only increase duplication, causing less efficient crawling and indexing. Therefore, consider stripping unnecessary URL parameters and performing your site's "internal maintenance" before generating the URL. If many parameters are required for the user session, perhaps hide the information in a cookie rather than continually append values like: 

cat=gummy-candy&cat=lollipops&cat=gummy-candy&...

Worst practice #5: Offering further refinement (filtering) when there are zero results.

Worst practice:

Allowing users to select filters when zero items exist for the refinement.

category page that lets users select filters when zero items exist

Best practice:

Only create links/URLs when it's a valid user-selection (items exist). With zero items, grey out filtering options. To further improve usability, consider adding item counts next to each filter.

categopry page that greys out filters if there aren't any items

Prevent useless URLs and minimize the crawl space by only creating URLs when products exist. This helps users to stay engaged on your site (fewer clicks on the back button when no products exist), and helps minimize potential URLs known to crawlers. Furthermore, if a page isn't just temporarily out-of-stock, but is unlikely to ever contain useful content, consider returning a 404 status code. With the 404 response, you can include a helpful message to users with more navigation options or a search box to find related products.

Best practices for new faceted navigation implementations or redesigns

New sites that are considering implementing faceted navigation have several options to optimize the "crawl space" (the totality of URLs on your site known to Googlebot) for unique content pages, reduce crawling of duplicative pages, and consolidate indexing signals.

  • Determine which URL parameters are required for search engines to crawl every individual content page (for example, determine what parameters are required to create at least one click-path to each item). Required parameters may include item-id, category-id, page, and others.

  • Determine which parameters would be valuable to searchers and their queries, and which would likely only cause duplication with unnecessary crawling or indexing. In the candy store example, I may find the URL parameter taste to be valuable to searchers for queries like "sour gummy candies" which could show the result example.com/category.php?category=gummy-candies&taste=sour. However, I may consider the parameter price to only cause duplication, such as category=gummy-candies&taste=sour&price=over-10. Other common examples:

    • Valuable parameters to searchers: item-id, category-id, name, brand, and others.
    • Unnecessary parameters: session-id, price-range, and so on.

  • Consider implementing one of several configuration options for URLs that contain unnecessary parameters. Just make sure that the unnecessary URL parameters are never required in a crawler or user's click path to reach each individual product!

    • Option 1: rel="nofollow" internal links

      Make all unnecessary URLs links rel="nofollow". This option minimizes the crawler's discovery of unnecessary URLs and therefore reduces the potentially explosive crawl space (URLs known to the crawler) that can occur with faceted navigation. rel="nofollow" doesn't prevent the unnecessary URLs from being crawled (only a robots.txt disallow prevents crawling). By allowing them to be crawled, however, you can consolidate indexing signals from the unnecessary URLs with a searcher-valuable URL by adding rel="canonical" from the unnecessary URL to a superset URL (for example example.com/category.php?category=gummy-candies&taste=sour&price=5-10 can specify a rel="canonical" to the superset "sour gummy candies" view-all page at example.com/category.php?category=gummy-candies&taste=sour&page=all).

    • Option 2: Robots.txt disallow

      For URLs with unnecessary parameters, include a /filtering/ directory that will be robots.txt disallowed. This lets all search engines crawl good content, but will prevent crawling of the unwanted URLs. For instance, if my valuable parameters were item, category, and taste, and my unnecessary parameters were session-id and price. I may have the URL: 

      example.com/category.php?category=gummy-candies

      which could link to another URL valuable parameter such as taste:

      example.com/category.php?category=gummy-candies&taste=sour

      But for the unnecessary parameters, such as price, the URL includes a predefined directory, /filtering/: 

      example.com/filtering/category.php?category=gummy-candies&price=5-10

      which is then robots.txt disallowed:

      User-agent: *
Disallow: /filtering/

    • Option 3: Separate hosts

      If you're not using a CDN (sites using CDNs don't have this flexibility easily available in Webmaster Tools), consider placing any URLs with unnecessary parameters on a separate host—for example, creating main host www.example.com and secondary host, www2.example.com. On the secondary host (www2), set the Crawl rate in Webmaster Tools to "low" while keeping the main host's crawl rate as high as possible. This would allow for more full crawling of the main host URLs and reduces Googlebot's focus on your unnecessary URLs.

      • Be sure there remains at least one click path to all items on the main host.
      • If you'd like to consolidate indexing signals, consider adding rel="canonical" from the secondary host to a superset URL on the main host (for example www2.example.com/category.php?category=gummy-candies&taste=sour&price=5-10 may specify a rel="canonical" to the superset "sour gummy candies" view-all page, www.example.com/category.php?category=gummy-candies&taste=sour&page=all).
  • Prevent clickable links when no products exist for the category or filter.

  • Add logic to the display of URL parameters.

    • Remove unnecessary parameters rather than continuously append values. Avoid:

      example.com/product?cat=gummy-candy&cat=lollipops&cat=gummy-candy&item=swedish-fish

    • Help the searcher experience by keeping a consistent parameter order based on searcher-valuable parameters listed first (as the URL may be visible in search results) and searcher-irrelevant parameters last (for example, session ID). Avoid: 

      example.com/category.php?session-id=123&tracking-id=456&category=gummy-candies&taste=sour
  • Improve indexing of individual content pages with rel="canonical" to the preferred version of a page. rel="canonical" can be used across hostnames or domains.

  • Improve indexing of paginated content (such as page=1 and page=2 of the category "gummy candies") by either:

    • Adding rel="canonical" from individual component pages in the series to the category's "view-all" page (for example, page=1, page=2, and page=3 of "gummy candies" with rel="canonical" to category=gummy-candies&page=all) while making sure that it's still a good searcher experience (for example, the page loads quickly).
    • Using pagination markup with rel="next" and rel="prev" to consolidate indexing properties, such as links, from the component pages/URLs to the series as a whole.
  • Be sure that if using JavaScript to dynamically sort/filter/hide content without updating the URL, there still exists URLs on your site that searchers would find valuable, such as main category and product pages that can be crawled and indexed. For instance, avoid using only the home page (or one URL) for your entire site with JavaScript to dynamically change content with user navigation—this would unfortunately provide searchers with only one URL to reach all of your content. Also, check that performance isn't negatively affected with dynamic filtering, as this could undermine the user experience.
  • Include only canonical URLs in sitemaps.

Best practices for existing sites with faceted navigation

First, know that the best practices listed above (for example, rel="nofollow" for unnecessary URLs) still apply if/when you're able to implement a larger redesign. Otherwise, with existing faceted navigation, it's likely that a large crawl space was already discovered by search engines. Therefore, focus on reducing further growth of unnecessary pages crawled by Googlebot and consolidating indexing signals.

  • Use parameters (when possible) with standard encoding and key-value pairs.
  • Verify that values that don't change page content, such as session IDs, are implemented as standard key-value pairs, not directories.
  • Prevent clickable anchors when no products exist for the category/filter. Don't allow clicks or URLs to be created when no items exist for the filter.

  • Add logic to the display of URL parameters. Remove unnecessary parameters rather than continuously append values. For example, avoid: 

    example.com/product?cat=gummy-candy&cat=lollipops&cat=gummy-candy&item=swedish-fish

  • Help the searcher experience by keeping a consistent parameter order based on searcher-valuable parameters listed first (as the URL may be visible in search results) and searcher-irrelevant parameters last. For example, avoid: 

    example.com/category?session-id=123&tracking-id=456&category=gummy-candies&taste=sour&

    in favor of:

    example.com/category.php?session-id=123&tracking-id=456&category=gummy-candies&taste=sour

  • Configure Webmaster Tools URL Parameters if you have strong understanding of the URL parameter behavior on your site (make sure that there is still a clear click path to each individual item/article). For instance, with URL Parameters in Webmaster Tools, you can list the parameter name, the parameters effect on the page content, and how you'd like Googlebot to crawl URLs containing the parameter. The following table demonstrates how different settings affect parameters:

    Parameter name Effect on content? What should Googlebot crawl?
    trackingId None One representative URL
    SortOrder Sorts Only URLs with value = 'LowToHigh'
    SortBy Sorts Only URLs with value = 'Price'
    FilterByColor Narrows No URLs
    itemId Specifies Every URL
    page Paginates Every URL
  • Be sure that if using JavaScript to dynamically sort, filter, and hide content without updating the URL, there still exists URLs on your site that searchers would find valuable, such as main category and product pages that can be crawled and indexed. For instance, avoid using only the home page (or, one URL) for your entire site with JavaScript to dynamically change content with user navigation—this would unfortunately provide searchers with only one URL to reach all of your content. Also, check that performance isn't negatively affected with dynamic filtering, as this could undermine the user experience.
  • Improve indexing of individual content pages with rel="canonical" to the preferred version of a page. rel="canonical" can be used across hostnames or domains.

  • Improve indexing of paginated content (such as page=1 and page=2 of the category "gummy candies") by either:

    • Adding rel="canonical" from individual component pages in the series to the category's "view-all" page (for example, page=1, page=2, and page=3 of "gummy candies" with rel="canonical" to category=gummy-candies&page=all) while making sure that it's still a good searcher experience (for example, the page loads quickly).
    • Using pagination markup with rel="next" and rel="prev" to consolidate indexing properties, such as links, from the component pages/URLs to the series as a whole.
  • Include only canonical URLs in sitemaps.

Remember that commonly, the simpler you can keep it, the better. Questions? Please ask in our Webmaster discussion forum.

Written by Maile Ohye, Developer Programs Tech Lead, and Mehmet Aktuna, Crawl Team