Several years ago I was tasked with coming up with a way to identify destinations based on discovery or a theme. This turned into Discovery Themed Navigation, and a way to automatically curate 1 million pages of distinct content based on destination, lifestyle, economics, terrain, etc…
Here are my Evernote notes and what I came up with.
Discovery Themed Navigation provides the user the ability to easily navigate to a destination of choice using keywords, keyword combinations, phrasing, and keyword phrasing all surrounding their interests, a category, a theme, a location, or any combination therein.
The user will have the ability to select by..,
Cancun, Beach, Luxury…
Cancun, Snow Skiing, Luxury., because there isn’t any snow skiing in Cancun.
they also wouldn’t be able to select..,
Cancun, Mountains, Luxury., because there aren’t mountains in Cancun, the closest mountain is Victoria Peak, Belize., 517 km or 321 mi away.
They will have the ability to navigate by:
Cancun, Beach, Luxury
Cancun, Beach, …
Cancun, …, Luxury
Cancun, …, …
…, Beach, Luxury
…, Beach, …
…, …, Luxury
To get accurate results.
Another way to accomplish the sane selection would be..,
Spring Break, Singles, Surf
Spring Break, Beach, …
Singles, Beach, Adult Only
How can you curate a million pages?
Targeting the user’s requests
A traditional search approach has the user enter a search term in a specific area, then return as many results as possible in order to provide as many options as possible, for the most part within relevancy, in the hopes that the user will find the result that he/she wants and then converts. This approach works well for one reason.., price. All things being equal the user will shop price, then convenience, with convenience sometimes winning over pricing, but only if that price is eclipsed by perceived value.
The categories are Events, Interests, Terrain, Lifestyle and Economic. This image shows their relationship.
As you can see categories intersect each other in accordance with commonalities with Economics being an over-arching consideration.
Beneath the categorical relationships are the relationships between categories and themes. These relationships begin to define a road map from where characteristics of a destination or experience can be identified.
The image below shows the relationship between categories and themes with the inter-relationship between some of the themes.
The image below shows the relationship between locations and categories and themes. There are some additional attributes of locations with Seasonal, Localized Ads or Promotions, and Activities. These attributes are further inter-relationships with themes and provide refined attributes to categories.
Within the relationships some are conditional, as indicated by the dashed line, and some are dependencies, while others are shared relationships. In the shared relationships there may be a conditional on one side and a dependency on the other or a dependency on both sides.
An example of a relationship condition as it applies to the image above would be “seasonal”. In this case both Interests and Lifestyle share Adventure. If the customer wanted to book a trip to snowboard Half Dome mountain at Yosemite there would be a seasonal condition.
The condition would be that snowboarding Half Dome is seasonal. You wouldn’t snowboard during the summer, but you could climb or base jump.
The relationship dependencies would be Interests and Lifestyle and Adventure. Additional dependencies might be Parks, Sports, and Activities.
The page should be fluid enough that it shouldn’t have to be a template or even a series of templates. Think of tetris.
Content should flow on to the layout, displaying as it’s available and truncate as necessary.
The content blocks should possibly be a template?
The page will have the ability to;
- Scale according to content
- Degrade according to content
- Be consistent in expression given a wide set of variables
How will the page degrade if there is more to do at one location and very little to do at another?
There are two different aspects to take into consideration when discussing how a page degrades.
The first is the degradation of content.
As the editorialized content for the page wanes, the social content begins to become more prominent.
The second aspect is content density.
As the availability of content becomes limited the opportunity to include strategically placed localized promotions becomes more of a consideration.
A faceted navigation allows users to make their initial selection from a list of facets and then refine as more targeted information or options become available until they’ve refined their search to the options they want.
What might the navigation scheme look like?
Facets will be displayed as they are available. Nothing will display a null value.
Since economics have proven to be an overarching consideration for all but the luxury purchasers, the slider should accurately reflect changes in the refinement or establishment of a trip profile. This means that beginning and ending values reflect the offerings, and the step values will reflect accurate pricing. This will eliminate having a pricing range that returns no results.
What will the page(s) look like?
The initial results would be provided in a similar fashion as today with the exception being that the primary result, which is the most targeted based on the entry, will be the largest content block shown. The secondary result being smaller and the tertiary results being smaller still.
A limited number of results will display for the user to select from. The reason is that the results should be so targeted and accurate that introducing a large number of results would skew the targeted offerings. However permutations of the targeted result could be of benefit if they are focused on the periphery of the user’s interest that has been introduced, and they have relevancy.
There should be a “browse” capability that allows users to peruse through a higher level of values based on upper level facets.
As the user adds additional details to the trip, it is reflected in the primary content block. Effectively building an itinerary that they can easily refer to.
From this “itinerary” the user can modify or edit the parameters without having to go back through the entire workflow.
Real World Capabilities
The real-world capabilities that are available spans platforms seamlessly. The underlying technology, Lucene, allows for access to results in as little as 90ms. The Solr addition provides faceted navigation capabilities in an integrated fashion without jumping technological solutions.
The ability to utilize the NLP (Natural Language Processing) functionality of the technology puts the ability to enter natural language search phrases into the site to return “relevant” results at an extremely high level of accuracy. From there it’s a very short step to add IVR (Interactive Voice Recognition – Nuance) into the solution.
This is today’s technology.., not something that will be available in the future.
If the user has the ability to enter in natural language and that includes the spoken language their request for travel, and on the device of their choosing, then they should also be able to make the transaction as well.
There’s a lot at stake with mobile payments. Juniper Research estimates the mobile payment market will reach about $630 billion worldwide by 2014. Wilcox predicts 20% of mobile phones will be NFC-enabled in the next three years.
In January 2011 IDC (International Data Corporation) released the numbers for 2010 fourth quarter computer and smartphone sales. For the first time, smartphone shipments surpassed computers with 100.9 million smartphones to 92.1 million computers shipped.
Adding Augmented Reality
Augmented reality allows the user to virtually look through their phone at their surroundings using the phone’s camera. During the scan, balloons pop up over brick and mortar vendors. These can be further explored by the user when they touch a balloon.
Using the Wikitude API an augmented reality application can be built in hours and not weeks.
Once the user selects a vendor they can be presented with an assortment of content to include;
- Listing of products or services
- LTOs (limited time offers)/Flash Sales
- Rewards and Loyalty
This can be an opportunity for additional revenue.
Customize your trip – This will add additional items like car, activities, flight, etc.. based on what the user has selected and what they’ve filtered.
Prediction – Based on limited information about the user’s trip make certain assumptions or predictions as to what they might be interested in. This would be backed up by trending information from like users.
For instance, if a user has indicated that their attributes are;
And they are looking at flight itinerary..,
The assumption or a prediction can be made that they might need ;
- a hotel
- a car
- tickets to a theme park
- tickets to a water theme park (if not by a beach)
This information can be presented to the user with great confidence based on their initial attributes. This prediction can be even further refined. Using the listing above.,
- a hotel
- Family indicates there is at least one child
- Hotels that are kid-friendly
- Kid menus
- Play area
- On-site laundry facilities
- Game room
- Near theme park
- a car
- mid-sized sedan
- tickets to a theme park
- tickets to a water park (if not by a beach)
Yet to be discovered semantic clusters
As different associations become identified they can easily be added as an additional cluster.
The would is full of current events that happen in different places and sometimes with little notification, while in other instances there is plenty of forewarned knowledge.
Parameter Error Pages
When a user enters in parameters that would violate the destination’s attributes, content could be displayed that would take this page from an error status.
San Diego, skiing, luxury
First the identification would be made if there is skiing in San Diego. If so, what kind?
There is water skiing in San Diego, so information about water skiing, wake boarding, and bare footing would all be included.
There isn’t any snow skiing in San Diego and instead of dropping the user on to an error page or just a San Diego destination page, display content options that would allow them to book a skiing trip from San Diego. Or skiing destinations that are popular from San Diego.
Let’s look at another scenario. This time the user has been a little more ambiguous with their destination identification.
In this case the user has identified;
Europe, Summer, Adventure…
We could return results like this;
“The San Fermin Festival which occurs in Pamplona Spain in early July and most commonly identified with the running of the bulls.”
“Tour du Mont Blanc. The Tour du Mont Blanc is a 10 day circular walk through the Alps that starts and ends in France but passes through both Italy and Switzerland along the way.”
Each of these selections uses the same relationships and conditions that we’ve mapped out in the diagrams. The location is Europe, the theme is Adventure and the condition is Summer.
How do you define Romance?
How do you get the site to recognize romance? Or romantic?
Almost any location can be romantic because of it’s subjectivity. Mining social data would help to narrow and identify locations and activities.
Add to Gallery
The ability for users to build a trip and save that to a gallery that can be accessed by others. They would also be able to post that trip package to their network.
There could also be incentives that would allow them to promote to their friends and if they get (x) number of their friends to get the same trip they, the originators, would get a special discount or incentive of some type.
Local Ad Management
Add an AdWords kind of functionality for local advertisers. Each local business could create an account and log in to place their ad and ad artwork.
A cross between booking.com and hotels.com
Get Content from;
- trip adviser
Contains free content that includes
Summarization and Indexing
Content can be pulled in from the above sources and will be seen as non-original content. However, if the content is summarized and some additional content added or keywords changed, it can be seen as original content by the search engines, thereby performing higher in the rankings.
The use of controlled vocabularies in the indexing can identify terms that can be linked to its source. This source can be another page or a document or even a series of pages and or documents.