Offer Personalization as a Service for Hospitality June 2015 Technical Overview by Nor1 Data Science Team

Similar documents
Maximizing Guest Experiences

Business Intelligence Solutions for Gaming and Hospitality

How the right CXM solutions deliver better customer experiences

of interaction. Operate with Efficiency. Manage the Operation. Connect with Customers. Enhance with Mobility. For Table Service Restaurants

Capitalize on Big Data for Competitive Advantage with Bedrock TM, an integrated Management Platform for Hadoop Data Lakes

WHITE PAPER SPLUNK SOFTWARE AS A SIEM

BIG DATA ANALYTICS FOR HOSPITALITY AND LEISURE Learn more about your customers than ever before!

April 2016 JPoint Moscow, Russia. How to Apply Big Data Analytics and Machine Learning to Real Time Processing. Kai Wähner.

Actionable Knowledge from Refined Data with Microsoft Business Intelligence

SAP Best Practices for SAP S/4HANA Scope 1511 Cloud Marketing Edition Business Priority view. last update:

Elevate Customer Experience and Engagement in the New Digital World

Oracle Real Time Decisions

Mitra Innovation Leverages WSO2's Open Source Middleware to Build BIM Exchange Platform

Video Analytics. Keep video customers on board

Episerver Digital Experience Cloud Omni-Channel Digital Commerce for Dynamics AX

Big Data Web Analytics Platform on AWS for Yottaa

How To Make Data Streaming A Real Time Intelligence

B2C Marketing Automation Action Plan. 10 Steps to Help You Make the Move from Outdated Marketing to Advanced Marketing Automation

SAP CRM RAPID DEPLOYMENT SOLUTION. Package Overview

Configuring and Managing Microsoft System Center Essentials 2010

How B2B Customer Self-Service Impacts the Customer and Your Bottom Line. zedsuite

ezee Absolute Release Notes

Peopleclick Authoria RMS

Solving Your Big Data Problems with Fast Data (Better Decisions and Instant Action)

PIVOTAL CONNECTOR FOR MARKETO. Copyright 2015 Tokara Solutions. All Rights Reserved.

locuz.com Big Data Services

How to Enhance Traditional BI Architecture to Leverage Big Data

Databricks. A Primer

How to select the right Marketing Cloud Edition

RSA Security Analytics Certified Administrator (CA) Certification Examination Study Guide

SILVERWARE INSIGHT. BUSINESS INTELLIGENCE. REAL-TIME REPORTING, ALERTS, & ANALYTICS.

20 Quick Tips for Improving Your Marketing Programmes

DATA QUERY: ADVANCED DATA MODELLING

Social Business Intelligence For Retail Industry

Transform Inbound Contacts Into Profits: Best Practices for Optimizing Lead Management.

Intel HPC Distribution for Apache Hadoop* Software including Intel Enterprise Edition for Lustre* Software. SC13, November, 2013

How are you combating app fatigue? Has your business found a way to realize the power of Passbook?

Big Data? Definition # 1: Big Data Definition Forrester Research

CRM Analytics. SAP enhancement package 1 for SAP CRM 7.0. Gert Tackaert

Predictive Customer Intelligence

THE TOP 10 MOBILE MARKETING AUTOMATION BEST PRACTICES. Practical Methods For Driving Engagement, Retention and Revenue

Compare versions with Maximizer CRM 12: Summer 2013

Sourcing best practices SAP AG. All rights reserved. Internal

Databricks. A Primer

Customer Timeline - New in Summer Web Lead Capture - New in Summer Built-In Dashboards - New in Summer 2012

QLIKVIEW DEPLOYMENT FOR BIG DATA ANALYTICS AT KING.COM

Retail Analytics The perfect business enhancement. Gain profit, control margin abrasion & grow customer loyalty

Hadoop & Spark Using Amazon EMR

Microsoft Office Project Server 2007

IBM Global Business Services Microsoft Dynamics CRM solutions from IBM

Best Practices for Hadoop Data Analysis with Tableau

Lambda Architecture. Near Real-Time Big Data Analytics Using Hadoop. January Website:

Features Document of

Better for recruiters... Better for candidates... Candidate Information Manual

11 emerging. trends for DIGITAL MARKETING FINANCIAL SERVICES. By Clifford Blodgett. Demand Generation and Digital Marketing Manager

Digital Business Platform for SAP

customer care solutions

A financial software company

WhiteWave's Integrated Managed File Transfer (MFT)

MARKETO CHECKLIST. All users are setup within Marketo with the appropriate roles and permissions.

Lead Management CRM Marketing Automation Powerful. Affordable. Intuitive. gold-vision

RS MDM. Integration Guide. Riversand

Find, track, pipeline, and manage your highly-skilled talent.

A guide to effective testing. Best practices for campaign analysis and optimization

Vulnerability Management

OLAP Services. MicroStrategy Products. MicroStrategy OLAP Services Delivers Economic Savings, Analytical Insight, and up to 50x Faster Performance

Cloudera Enterprise Reference Architecture for Google Cloud Platform Deployments

Data Mining, Predictive Analytics with Microsoft Analysis Services and Excel PowerPivot

WHITE PAPER. Five Steps to Better Application Monitoring and Troubleshooting

Making big data simple with Databricks

Statement of Direction

30-Day Starter Guide to Marketing

CHOOSING AN SEM PLATFORM:

Cloudera Enterprise Data Hub in Telecom:

Getting Started & Successful with Big Data

Ganzheitliches Datenmanagement

White Paper April 2006

Predictive Marketing for Banking

Sweating Digital Assets Analytics Way

Next presentation starting soon Business Analytics using Big Data to gain competitive advantage

Associate Professor, Department of CSE, Shri Vishnu Engineering College for Women, Andhra Pradesh, India 2

Alcatel-Lucent Targeted and Interactive IPTV Advertising Solution

AgilOne + Responsys. Personalizing and measuring your Responsys campaigns just got a whole lot easier.

Marketing in the 21 st Century THE MID ATLANTIC CHESAPEAKE CONFERENCE

WHITE PAPER ON. Operational Analytics. HTC Global Services Inc. Do not copy or distribute.

Cisco Advanced Malware Protection for Endpoints

Centris optimises user support with integrated service desk

SAP Digital CRM. Getting Started Guide. All-in-one customer engagement built for teams. Run Simple

Continuum increases leads by 22% with BrightInfo Content Targeting for Blogs and Landing Pages CASE STUDY

Patient Relationship Management

Solution Overview. Optimizing Customer Care Processes Using Operational Intelligence

Extension of ERP for marketing: internal system + external communication Microsoft AX Dynamics. Prof.dr. Dalia Krikščiūnienė

Marketing Automation with Microsoft Dynamics

From Lab to Factory: The Big Data Management Workbook

Multichannel Customer Listening and Social Media Analytics

Demonstration of SAP Predictive Analysis 1.0, consumption from SAP BI clients and best practices

Oracle Big Data Spatial & Graph Social Network Analysis - Case Study

Customer Success Platform Buyer s Guide

Best Practice Search Engine Optimisation

Transcription:

Offer Personalization as a Service for Hospitality June 2015 Technical Overview by Nor1 Data Science Team

Offer Personalization as a Service for Hospitality Introduction Intelligent from Day One Context Based Defining Events Map Offers to Event Context Segmentation Define Segments Offers to Segments Mapping and Rules Machine Learning Enhanced Segmentation Feedback Loop Pluggable Performance Metric (a.k.a Objective Function) Segment Suggestion using Unsupervised Machine Learning Individual Guest Modeling Guest Profile PII Anonymization Machine Learning Enhanced Individual Recommender Per Guest Predictive Model Collaborative Filtering Continuous Improvement using Lambda Architecture and Bandit A/B Testing Fully Automated Reinforcement Loop with Lambda Architecture A/B Testing for Innovation and Risk Management Non-stop Improvement & Faster Conversion with Bandit Experiment Personalization Service APIs Nor1 Big Data Analytics Platform Conclusion Introduction Nor1 believes and has proven that personalized offers dramatically increase hotel revenue and guest satisfaction. Presenting a random offer to a guest will only result in random conversion. The more personalized the offers are, the greater the conversion, revenue and guest satisfaction. This white paper explains how Nor1 s cutting edge personalization service (PRiME) works and how your hospitality business can benefit from it. The Nor1 personalization approach can be broken down to three levels based on how much individual data is available at any given point in the reservation life-cycle 1. Contextual 2. Segmented 3. Machine Learning-Enhanced Segmentation.

At each level, Nor1 s PRiME personalization service automates the decision making for you, so that management time is reduced to nearly zero and can scale to present personalized offers for all guests throughout the entire travel life-cycle. Nor1 utilizes machine learning techniques to build predictive models that achieve optimal conversion on a wide variety of products and the system continuously improves with every new observation in the recent data collected. This is extremely important because Nor1 learns about individual guest behavior throughout an individual reservation life-cycle. Intelligent from Day One Nor1 s PRiME personalization service and approach provides intelligence from the very first day of use. Then gradually and over time, this gets increasingly sophisticated with a greater volume of transactions. The maturity of the intelligence and precision of offers ramps from context based offer presentation, combined with segment based intelligence, then once the appropriate volume of observations have been collected progresses to machine learning based enhanced segmentation and finally matures to true individual recommendations based on our per guest models and collaborative filtering techniques.

Example timeline to full utilization of Nor1 s PRiME personalization service Machine Learning Methods Segmentation Segmentation Contextual Contextual Contextual from day one, 1-2 Weeks 3-8 Weeks 8-12 Weeks Progression of offer personalization Contextual Level Defining Events Events during a guest s travel trigger offer presentations and communications. Which offers will be presented is based on the context of the particular event. For example, at a post reservationpre-arrival event room upgrade offers will be presented, at two days before arrival event a ground transportation offer and an early check-in offer may be presented, at check-in event a restaurant or bar offer may be presented. Nor1 has defined a set of standard events for hotels, like post reservation, two days before arrival, welcome communication, checked-in, etc. Over time more events specific to the dynamics of a particular property will be created. The key is making the right offer at the right time. Hotels can also use the ereach Publisher Module offer creation tool to easily create offers through a self-explained wizard process using predefined templates.

Assigning Offers to Events For each event, hotels can specify which offers to present to which guests, over time PRiME will determine the optimal offers throughout the guests stay on property. An offer can be assigned to one or more events. The question now is which offers should be assigned to an event? The answer will be based on determining the objectives of the property or organization.. For example, the hotel s marketing department is promoting an offer and wants to maximize the exposure. Or a different scenario is the hotel s revenue management department wants to optimize maximizing revenue or profit. Nor1 s PRiME personalization service understands these different needs from different customers well and is flexible for a hotel user to choose the objective to be used for optimization. Based on the objective, Nor1 s personalization service will automatically compute offer rank based on historical data. The rank is computed using the objective metric chosen. Offer rank is computed for each event separately. For a specific event, the offers with the highest rank are chosen to present to guests, so that the objective metric is maximized. The ranks are continuously updated in the background to adapt to offer or seasonal changes. The PRiME personalization service also provides full control for hotels to override the automated offer assignment. For example, always include or exclude an offer for a certain event. If a hotel wants to launch a promotion campaign, an offer can be configured to always present at a certain event which overrides the popularity selection. PRiME also gives the hotel user complete visibility to the performance of offers by exposing the performance metrics through an intuitive dashboard as shown in the figures below. These tools provide real-time visibility on the offer performance. Segmented Level Define Segments The next level of personalization is to segment guests into different personae and present different offers to those different personae. By first defining the different personae, then segmenting guests to those personae, Nor1 then decides which set of offers to present for each personae.

Definition of a persona (guest segment) is dependent upon many factors: 1. Basic profile information like gender, age, date of birth, home zip code, language, email domain, company, loyalty member status, first digit of credit card, etc. 2. Reservation data like length of stay, booking price, room category, arrival and departure dates, accompanies, stay day of week, etc. 3. Historical behaviors like offer purchasing or denial, total spending on offers, etc. Personae can be easily defined by hotels using the ereach Publisher Module persona creation tool, which leverages the hotels existing business insight. Next, hotels can map offers to each persona to treat different persona differently.

Offers to Segments Mapping and Rules PRiME also provides a dynamic business rule engine to make it easier for hotels to fine-tune the mapping from offers to personae, in which a hotel user can define exclusions of a certain category of offers for a certain personae. For example, do not show bar offers to family persona.

Machine Learning Enhanced Segmentation Level Feedback Loop To simplify the process of mapping offers to personae, Nor1 s PRiME personalization service provides automation by using machine learning algorithms. Performance metrics e.g. revenue or conversion are recorded per offer per persona and used as metrics to decide which offers are most effective for a certain type of persona. Pluggable Performance Metrics (Objective Functions) Performance metrics, a.k.a objective function, can be customized based on your optimization needs. Pre-defined objective functions to Maximize: 1. Exposure (for Loyalty & Marketing) 2. Purchases (for Merchandising) 3. Revenue (for Revenue Management) 4. Profit 5. Overall Conversion rate

A hotel user chooses a performance metric and the machine learning algorithm will optimize towards that set objective. Segment Suggestion using Automated Machine Learning From the reverse angle, Nor1 s PRiME personalization service uses sophisticated machine learning algorithms to suggest correlations between offers and personae (detecting small circles among personae and offers), sampling through guest population to show their distribution among different personae, so that hotels have better visibility into their personae and are able to get an intuitive sense for better targeting and expectation. Individual Guest Modeling Guest Profile The next level of personalization is true individual guest modeling. Nor1 creates a buyer behavioral profile for each guest capturing basic personal information, offer preference, current reservation, past reservations, responses to offers and most importantly, customized rules for the guest based on explicitly expressed preferences, willingness to pay trends. For example, whether a guest likes spa offers or a guest does not like bar offers, and what type of incentives resonate most effectively with this guest (2 for 1, percentage discount, free item), etc. PII Anonymization Nor1 s guest profile system collects individual guest information, which implements security and privacy policies seriously. Personal Identifiable Information (PII) are anonymized and sensitive data is encrypted. Machine Learning Enhanced Individual Recommender Per Guest Predictive Model Based on each guest profile, a predictive model is created for this individual guest to predict how interesting an offer is to this guest. Each guest can have a unique model. For new guests, a generic model based on segmentation is assigned to the guests and the model will be improved as guests respond to the offers presented. Collaborative Filtering By serving offers to hundreds of millions of guests, Nor1 s PRiME personalization service is able to leverage guest behaviors using aggregate statistics to suggest offers to a guest based on other similar guests favorite offers using techniques like collaborative filtering.

Continuous Improvement Lambda Architecture and Bandit A/B Testing Fully Automated Reinforcement Loop with Lambda Architecture Play and Learn then Play Better. Nor1 s PRiME personalization service improves itself automatically over time. This is achieved by continuously collecting data, training new models and applying those models to future offer decisions in a seamless loop. Lambda architecture is implemented to do this collect-train-deploy loop automatically and daily, hourly or even every few minutes. A/B Testing for Innovation and Risk Management Lots of observation and insights lead to lots of ideas that can improve our offers, user interface, predictive model potentially perform better and better than current version, including unknowns and brave innovations that contains risk. Question is how to evaluate which variants are better and reduce risk to make a dent in our revenue? Sophisticated A/B testing is heavily used inside Nor1 s personalization service to solve this problem. Beginning of the loop is defining variants and performance metric and goals for an experiment. Then start the A/B testing and collect data, analyze and report the results, finally picking the winner.

Non-stop Improvement & Faster Conversion with Bandit Experiment To make A/B testing converge faster and non-stop experimenting, Bandit A/B testing is another technique that allows continuous improvement, continuously experimenting with predictive models or offer making strategies against live booking offer requests in real-time. It defines a performance metric to decide which variant is the winner, then splits a certain percentage of traffic to the winner and the rest of the traffic goes to all variants evenly. Over time, the percentage adjusts automatically based on statistical significance, driving most of the traffic to the winner. When adding a new variant or changing conditions, the experiment and split percentage can be reset and run again. The split percentage will converge in a short period of time, maximizing conversion very quickly. Personalization Service APIs Integration with Nor1 s PRiME personalization service is simple through calling our APIs: 1. API for uploading room configurations 2. API for uploading rate calendar 3. API for uploading guest booking/reservation info 4. API for uploading offers 5. API for querying optimal offer set for a guest at a certain event Being true to a complete push model, Nor1 s offer algorithms provide a complete, Asynchronous API which enables scheduling and callback of offers.

Nor1 Big Data Analytics Platform The foundation of all our sophisticated analytics and machine learning work is built on top of Nor1 s big data platform, around a Hadoop eco-system. The suite of Hadoop tools used include HDFS, Hive, HBase, Mahout and Spark. Log streaming is built around these tools with Flume and batch data upload with Sqoop and database connectors. On top of the basic building blocks, Nor1 s intelligent components are event tracking and ETL, A/B testing, reporting, performance monitoring, rules engine and most importantly, machine learning pipeline.

Conclusion Offer personalization has gained greater strategic, competitive value in recent marketing, revenue management and guest relation/satisfaction management. Nor1 provides a convenient service for hotels to easily manage and maximize their overall merchandising strategies, significantly improving the guest s overall experience. 2015 Nor1, Inc. All Rights Reserved.