Predictive match ppw
1 like594 views
The document discusses the rise of predictive modeling and how big data science is being used by property portals. It describes how property portals are currently mining user behavioral data through web logs and various technologies to better understand users, inform product decisions, and implement personalized recommendations and targeted marketing. The document also outlines several predictive modeling applications including recommendation engines, segmentation, risk modeling, and fraud detection that can benefit property portals.
Predictive match ppw
- 1. THE RISE OF PREDICTIVE MODELING How big data science is penetrating property portals now and into the future Property Portal Watch Singapore 2014
- 2. PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014
- 3. EMERGING TECH PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014 The challenge is no longer how to answer the question, it has become what question t
- 4. BIG COMPANIES MINE BIG DATA PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014
- 5. CURRENT USES - Recommendation engines * - Sentiment analysis - Segmentation * - Risk Modeling - Marketing Campaign Analysis * - Social graph analysis * - Fraud detection - Customer experience analysis * PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014 * Apply to property portals
- 6. HOW IS THIS RELEVANT TO PROPERTY PORTALS ? -The role of a portal – connect buyers to sellers. -Consumer marketing attract users to the site -Product and tech try to keep them there and convert them into leads -Achieved through search. -Behaviourial data already exists, but is not utilised. PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014
- 7. WHAT ARE PROPERTY PORTALS DOING? -Statistical analysis and mining of historical web logs to: - Understand user behaviour - Inform product road maps - Create deeper and more accurate segmentation of your user base - Personalise and targeted marketing and display advertising PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014 Events Web Logs Tag Management Analytics Output Store reduce map elastic NOSQL Graph Visualize
- 8. WHAT ELSE? -Combine with other data sources: - Social networks - Other 3rd party sources Personalised / Curated / Suggested / Recommended WE KNOW WHO YOU ARE AND WHAT YOU WANT PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014
- 9. PROPERTY PORTAL WATCH CONFERENCE SINGAPORE 2014 http://youtu.be/rbX2DBBCrUw
- 10. WHAT WE DO IN A NUTSHELL? 1. Serve behavioural based user centric recommendations 2. Improve audience engagement through understanding data and unlocking insights COMMERCIAL IN CONFIDENCE
- 11. WHY WE EXIST Adobe Recommendationsitunes Genius Proprietary E- Commerce E- Commerce Providers Software Platforms Music Content Service s Sites and services Software Providers Coremetrics Analytic s Movies Consumer Books Big Data Analytic s PROPERTY PORTAL WATCH CONFERENCE MAY 2014 Behavioural based recommendations are common place in e-commerce, music, movies, content, books and have been proven to increase conversions, yet no off the shelf solution currently exists for classifieds
- 12. HOW IS PREDICTIVE MATCH DIFFERENT? Real-time recommendations Customizable user behaviour targeting and listing segmentation Predictive models built for classifieds domain A/B testing for all algorithms Open Platform and Modern REST API Lightweight and fast HTML5 JavaScript templates Highly available locally developed and hosted Plug and Play COMMERCIAL IN CONFIDENCE
- 13. USAGE This data is collated and used to predict listings of interest for new users Data is collected for all users via a simple asynchronous javascript placed on Search, View, Enquiries, Share and Log In. User centric predictive recommendations are returned via a simple call to be displayed on site and in email alerts to improve conversions and engagement Commercial in confidence – copyright 2013
- 14. HOW DO YOU INTEGRATE? Track User Behaviour Sync/Async JavaScript Tracking Pixel Notify Listing Changes (Create, Update, Expire) Sync/Async JavaScript Server to Server Listing API Serve Recommendations Inline Javascript Widget Native Dust.JS template Publisher CSS COMMERCIAL IN CONFIDENCE
- 15. SHOW ME THE SYSTEM ARCHITECTURE Publisher website Tracking Pixel Recoms. API Listing API CDN User behaviour (search, view, eoi, enquire, click, hide) Listing changes (create, update, expire) Recommendations (dynamic listing json) Content (template js, css) Listings Docs In-mem. Recoms Events Online Near time Offline Hadoop Machine Learning Analytics Graph Algos. Queue Event Processor COMMERCIAL IN CONFIDENCE
- 16. mnjhjhjhjhj +61 417 334 001
Editor's Notes
- #3: By definition big data refers to data sets whose size and unstructured nature make them impractical to process and analyse with traditional database technologies and tools.Machine learning deals with an area of artificial intelligence where technology can study and learn from data.Predictive modeling as the name suggest refers to machine learning algorithms that identify patterns in data which is used to predict outcomes.The Big Data Market is already huge. Estimated to be over $16B USD growing to over $50B USD by 2017
- #4: Emerging big data technologies (like hadoop and mapreduce) make it possible to run predictive models and machine learning algorithms over massive amounts of data to answer questions and discover insights.
- #6: OK great… so big data is real .. People are using it to add value
- #7: SEARCH – Text based.. Textual characteristics
- #8: SEARCH – Text based.. Textual charactrists
- #9: SEARCH – Text based.. Textual charactrists
- #12: The advancement of behavioural based recommendations was led by companies such as Amazon (people who bought this also bought) and Netflix are common place in e-commerce, music, movies, content, books and have been proven to increase conversions. (Netflix gained a $90m uplift in revenue from a 10% improvement in recommendation quality. Amazons sales increaded by 30% the first year their introduced product recommendations). So why cant these same techniques be applied to property portals?