Back to top

A Modern Search Results Presentation

Paul Nelson
Paul Nelson
Innovation Lead

There was a time in the history of search where we were so happy to get the long list of results from our enterprise search solution. “Wow I got 5,025 results for "oil well safety". I didn’t know all this was available, way cool”.

These days, we are more likely to say. “Wow, 289,000 results for "oil well safety". and then get overwhelmed. We'll look at the first page or two of results hoping (or worse thinking) those actually are the best results.

The truth is the most “relevant” documents as defined by the search engine and shown in a single flat list, often do not give you the best overall picture of what is known by your organization. Issues with these simple flat lists include:

  • Document types are undifferentiated - wouldn’t it be nice if publications, web pages, people’s CVs, tools, data sets, etc. were somehow visually differentiated?
  • Document sources are undifferentiated - different sources have different purposes. For example, HR documents serve a very different purpose than support tickets or documentation wikis.
  • Documents are not grouped in any useful fashion to aid in navigation - a simple list of documents doesn’t provide navigational aid.
  • The initial first page or two of results do not represent the complete view - the best documents for certain sources or types are often hidden on later pages.
  • It is difficult to easily search within a type of document - search navigators help with this, to some extent, but there are other methods which may be even more effective.

Internet search engines such as Google and Bing have recognized this challenge and changed the way they present results. You may have noticed results are grouped by location, images/videos, news, social media...


Above are the results from a Bing search for Jimmy Buffett. You see a standard list of sites and articles, however in the list is a video section. Bing also provides separate results sections for social network content, Wikipedia and albums from Xbox Music.

Google uses a similar presentation style, often with a mixture of traditional text results, images and mapped results, each in their own section, and served depending on the search topic.  This presentation model provides a more complete view of what is known about the subject. This model may be even more applicable for the enterprise, because the content types tend to be better established, and the audience is more limited and defined in terms of their scope of interests.

Search Technologies is currently completing a Search Assessment for one of our clients that includes a proposal to adopt the Google/Bing presentation model. This blog post is intended to introduce the high level concepts, follow-on posts will go into more detail.

As with many customers, the current search results interface in use by this client is a single list of results sorted by relevance. With a flat list, you get all types and varieties of document in a single list – there is no differentiation.(see image below left). It is up to the user to say:  “Oh, that’s a policy document, and those are addendum notes, and that’s a spprofit/loss spreadsheet, etc”.  The users of the client's search system are mostly people  doing product development, financial analysis in their sector, or developing related policy issues. The content under index can be classified under one of the following, shown in the table below and to the right.



Recently, new features have become available in some advanced search architectures which can produce more than one list of documents from a single query. For example, in  Solr this is known as “Grouping”. It’s a deceptively powerful feature which we think will lead to a revolution in search user interfaces.

With multiple-top-lists, an enterprise  search application can understand that the world is made up of different types of documents, and then provide the top results from each type when presenting results to the user – making it easier to see “what’s available” at a glance, and then to focus in on the document type of interest. This presentational model guarantees that you will get samples from all types on the page. 

This gives the user an “overview” of what’s available. In a traditional representation model your top results might all come from one or two types, and the fact that some other types are missing will not be known to the user. With this approach, you get a sampling of documents from all types. Once the user has focused on the document type that best fits their use case (hopefully the document they seek is already visible). then with one click they can drill deeper.

Applying this new model to search interface design could result in the following example interface. The goal of this interface is to provide a “visual organization overview” To achieve this goal, documents are grouped by the following types:  Publications, Images, Maps, News, Projects, Tools and Staff.

This grouping achieves several important goals:

  • It clearly identifies types of the documents  in the search results
  • It trains the user as to what types of assets (maps, images, publications, people, etc.) are available for searching from this particular search engine
  • It prevents each type or source of documents from being “swamped” by results from other sources or types
  • It provides “jumping-off” points for more detailed searches via  “Search xyz >>” links which are shown above each type
  • It organizes the search results so that the eye is drawn to the most valuable information first.

Let’s look at the structure of an example interface using this new model, what each section is, and how they operate. 

At the very top of the search results are two documents which represent the “starting point” for new users who wish to educate themselves about the topic of the search, A maximum of two documents are presented in this section. In this case they are a basics” article for users with no prior background knowledge, plus the closest “home page” for the topic on the organization’s web site.

Next comes “organic” search results, straight from the search engine and based on the search engine's relevancy model. In this example, the organic search results are presented with standard “Google-like” styling (large blue titles, green URLs / navigators, light-grey teasers, bold highlights). Note that the organic results include web pages from, and nothing else. Other collections (publications, images, etc.) will be shown in separate sections.

In the above example, the traditional URL has been turned into a navigation bar to improve navigation. Each of the green items can be individually clicked by the user. This will allow, for example, a click on “pv” to jump directly to the Photovoltaics home page, or on “” to go directly to the that home page. Further, standard teasers have been improved to prefer showing a “first paragraph” from the document where this is appropriate – rather than three random snippets. Each of the sections, except for the “starting point” section, has a Search ABC>> link, where ABC is the content type, to transition to a detailed search only within that content type.

The next section provides publications and reports that are a common focus for search within this community. The best three matching publications are shown. To preserve real-estate on the screen, highlighted teasers are not served. This works for the intended audience, and hit highlights are shown in the title field. The author’s name is click-able, and shows all reports written by that author.

The next two sections are Images and Maps, presented side-by-side, as thumbnails in a miniature “light-box” view. This presentation style can be used for video too.
Information each can be presented as an “alt text” hover-over pop-up. Clicking on “Search images” or “Search maps” takes the user to a light-box style set of search results similar to Google's image search results.


For news, display of the meta title and publication date provide a sufficient summary. Results are sorted by date (rather than relevance).

After the news section, further “organic” results are displayed in descending relevance order, with pagination options at the bottom.

Over to the right we find Projects, Tools & Data, and Staff sections.  The purpose of these is to highlight and segment other important aspects of the ABCEnergySystems organization. These items are represented with appropriate icons to make them easy to spot in the search results. Even though the presentation is substantially different from the main body of the search results, they are produced by the search engine and from the same search request


No organization that I have ever known is flat and uninteresting. They are all made up of multiple groups, some small, some large, with different goals and purposes. Further, the amount of variety in type, size, and purpose of content within the organization is large, from employee directories and news snippets to 500 page technical manuals.

Given this variety it seems frankly dishonest to represent the results in a single, undifferentiated list. It’s dishonest because two documents are placed next to each other as if they were equal in every way, when clearly they are not. They have different relevance, purpose, and metadata. And even more important, they are often targeted at very different use cases.

As the quantity and the diversity of content types within our organizations continues to grow it is not only important that the search engines keep up with the ingestion of this “Big Content” explosion, the models and methods for presenting results must also keep up. 

The next generation of search presentation interfaces must help us to not get overwhelmed and to see the total picture.