February, 2013 Kurt Zimmer, CTO Where Do I Want to Go?? Why Don’t You Know Me! Intent, Geography and Predictive, Behavioral Analytics in Hotel Search
Jun 14, 2015
February, 2013
Kurt Zimmer, CTO
Where Do I Want to Go?? Why Don’t You Know Me!Intent, Geography and Predictive, Behavioral Analytics in Hotel Search
Welcome to Room Key
Who We Are
• Room Key was founded by six major hotel partners: Choice, Hilton, Hyatt, InterContinental, Marriott and Wyndham
• Established to differentiate the hotel shopping experience with a new customer proposition focused on delivering value, trust, and a personalized shopping experience while providing the advantages of direct booking
Room Key Proprietary & Confidential 4
What We’re Covering Today
• What is it I Wanted to Do? - Deriving Intent
• What’s in a Location? - Geography as a Differentiator
• How Did You Read My Mind? – Predictive, Behavioral Analytics as a Optimizer for Intent, Satisfaction and Revenue
• Changing the Hotel Shopping Experience - So What Do Geography, Intent and Behavioral Analytics Have to Do With Each Other?
Room Key Proprietary & Confidential 5
Before We Dive In
• Today’s discussion is about what’s possible
• It’s about thinking differently and about exploration of complex concepts that may seem simple at first
• Some of this is working today, some in the future
• Core Challenge – Knowing and Anticipating the Customer
Room Key Proprietary & Confidential 6
What was it I Wanted to Do?
• Ah, so it turns out intent is a tricky thing to define and draw out
• Simply, today’s approaches force people to already “know” what they want – sometimes yes, but often, no.
• The fun is often in the possibilities and with the confidence that you ended up where you hoped to
• Understanding intent does not have to be hard, but it just has to happen
• Food for thought – Do you start with a place or a theme?
• More food for thought – Gamification is perhaps a key
Room Key Proprietary & Confidential 7
What’s in a Location?
• Today’s hotel search approach and use of geography is, well, ugly!
• Virtually all sites today use “location from center” as starting point
• Booking site geography is most often about what you’re supposed to see not what you need to know
• Geography and locations should not be about words but about visualization – I want to “see” where I am
Room Key Proprietary & Confidential 8
Room Key Proprietary & Confidential 9
Room Key Proprietary & Confidential 10
Room Key Proprietary & Confidential 11
Room Key Proprietary & Confidential 12
Room Key Proprietary & Confidential 13
Room Key Proprietary & Confidential 14
Room Key Proprietary & Confidential 15
How Did You Read My Mind? (Ahhh, the perfect hotel room)
• Reading minds is not quite as hard as it sounds now that we know what someone wants to do and now that they have the ability to visualize the possibilities
• Just a huge dose of technology and some basic marketing concepts combined with data and real time behavior
• Well, perhaps it’s not that easy….. but enter predictive, behavioral analytics in a real time decision setting.
Decisioning with Behavioral Analytics 101 - RTD
1. Interstitial page makes an advisor call to RTD via web services to request properties
2. RTD sources nearby properties from the exit property from a database using a JBDC connection
3. RTD builds the visitor session in the cloud by pulling geographic data for the destination, recent hotel activity in the area, and information known from previous visitor sessions
4. RTD decision service processes business rules, scores predictive models, and processes decisions to rank properties
5. RTD returns an ordered set of candidate properties back to the page from the Advisor call
6. Once availability and pricing is complete, the page sends an informant with a listing of recommended properties
7. If there is a response (engagement or lead), an informant notifies RTD of this event via web services
8. RTD updates models in learning service
9. Hotel activity and visitor data for the session are updated in data sources
Engagement Lift Today, Returning Customers Tomorrow
RTD uses metadata relative to the exit property (distance, price, star rating, rewards program, Smith Travel assignment, and user rating), geographical
information (hotel density & current hotel demand) for the exit property area as well as limited visitor information to predict the best hotel choices
RTD Context Decisioning – High Level
For example, the recommendation set for a visitor from Hyatt is very different than the recommendation
set for a visitor from Choice hotels
RTD will have a very different recommendation set based upon the real-time context of the interaction
Sample RTD Analytical Insights
• Visitors are two times more likely to book a property with the same rewards program as the referring brand than a property that has a different rewards program
• Visitors are 30% more likely to book a property that has a very similar price as the referring property than a property that is slightly lower priced
• Close proximity to the referring property is much more of a factor in rural areas versus urban areas
• A higher star rating has more of a weight on increasing booking rates than having a higher Trip Advisor score
Room Key Proprietary & Confidential 20
More Than Insights and Recommendations
As a small, aggressive startup, we have:
• Complex Decisioning
• Large Amounts of Data
• Strong Need to Differentiate in Crowded Market.
But also a desire to take advantage of:
• Shrinking Time to Market and Time Between Test Iterations
• Technical Leverage
• Straightforward, Low Cost Operations.
And perhaps even changing the game in hotel search
Real Time Decisions (RTD) in Room Key
• Started RTD analysis and design in mid-January 2012 with an initial deployment early April 2012
• Average response time is about 80 ms
• RTD has made more than 600 million Room Key decisions and is currently making over 40 million decisions a week
• Have made more than a two dozen tech deployments since April enhancing decisioning, multivariate testing and analytical reporting functionality
• Marketing operations currently making about 40 changes a month to decisioning strategy
• RTD has generated visitor profiles for 35 million unique visitors tracking travel behavior
• Making worldwide recommendations from more than 50,000 consideration hotel properties which will increase over the next year to well over 100,000 properties
• And this is just the beginning……
Room Key Proprietary & Confidential 22
Changing the Hotel Shopping Experience
• Things change drastically as we get to know you – intent, history, preferences, family make up, geography approach, loyalty programs, etc
• Hotels recommendations can now be tied to predictive, self optimizing models and learnings about you and others like you, not to a simplistic distance from center starting point
• In short, we’ll be able to optimize ‘offers’ (properties and ads) based on the traveler and their situation at the time
• Even a tie to mobility using location awareness and surrounding opportunities when on the hotel property
• And if we’re lucky, we’ve changed the game
Room Key Proprietary & Confidential 23
Questions?