Ecommerce search is broken. If a problem cannot be solved, enlarge - - PowerPoint PPT Presentation

Ecommerce search is broken.

If a problem cannot be solved, enlarge it

• Dwight D. Eisenhower, 34th President of US

Understanding users is too slow and too complicated

1

2

The cost of experimenta:on is too high and too manual

Why are you buying your own data back from your vendors?

3

The AI gap is widening

4

Based in San Francisco Offices in Bangalore, Bangkok, New York City, Raleigh, London Over 300 customers across the Fortune 1000 Fusion, a Solr-powered plaOorm for search-driven apps Lucidworks powers retail and consumer search

Next-Gen Results. No PhD Required.

• Cogni&ve: machine learning to

automa<cally tune relevancy and drive recommenda<ons out of the box.

• Rules when you need them: 


Point-and-click query pipeline configura<on and rules module allow fine-grained control of results.

• Science, not guessing:

Experiment management and bandits do the heavy liEing of systema<cally tes<ng new ideas

Simplified App Dev

• Over 50 connectors and a

robust parsing framework to seamlessly ingest all your data

• Powerful pipeline stages:

Customize fields, stages, synonyms, boosts, facets, and dozens of other powerful search stages.

• Point and click Indexing

configura&on and itera<ve simula<on of results for full control over your ETL process

• Your security model

enforced end-to-end from ingest to search across your different datasources

Trusted Technologies

SQL

• Built on the most widely deployed

search engine on the planet

• Enhanced with scalable machine

learning and analyRcs

• Well understood and documented

APIs and client support designed to work with industry standard tools like SQL and Tableau

• Enriched with deep experRse and

partners and a large community

Trusted Technologies

SQL

UI should be beau&ful

• Modular library of UI components to create

prototypes in hours, not weeks.

• Stateless architecture so apps are robust, easy

to deploy, and highly scalable.

• Supports over 25 data plaIorms including Solr,

SharePoint, Elas<csearch, Cloudera, ANvio, FAST, MongoDB, and many more - and of course Lucidworks Fusion.

• Full library of visualiza&on components for

charts, pivots, graphs and more.

• Pre-tested re-usable modules include

pagina<on, face<ng, geospa<al mapping, rich snippets, heatmaps, topic pages, and more.

UI should be beau&ful

Retail Highlights

“Have a good weekend”

— Head of search at top 5 US Retailer on Black Friday to our support team

Mee'ng the Ecommerce Challenge

Conversion & Revenue Business Support Intelligence Open & Scalable

Personalization Recommendations Head/Tail Landers/Boost/Block Query Simulations Category Management Analytics Business Rules UI Proven Search Extensible Simplified DevOps Real Time Signal Processing Machine Learning Query Intent Smart Suggest

Signals: the Key to Retail Success

• Fusion has naRve support for capturing and processing a wide

variety of user feedback data, including:

• Clicks, Add To Cart, Conversions
• Easily integrates with 3rd party analyRcs systems like

Omniture/SiteCatalyst and Hadoop systems

• We ship with ready to go models for clickstream boosRng and

recommendaRons built on signals

• Deploy and access just like any other Fusion content!

Rules, Modernized The Fusion Rules Module extends Fusion with:

• Support for:
• Keyword, Tags, start/end date matching
• Boost/block, filters, promoRons/landers/

redirects, arbitrary request parameters

• UI
• Query Pipeline integraRon
• Extensible
• AudiRng and security built-in

If the Machines Ain’t Learning, You Ain’t Earning

• CapabiliRes include:
• Clickstream auto-tuning relevance and analyRcs
• Query Intent and Document ClassificaRon
• RecommendaRon and PersonalizaRon
• Synonym Discovery and Expansion via Word2Vec, Cluster Labeling, Topic DetecRon
• A/B/MulRvariate TesRng (MulR-arm bandit experimentaRon)
• Learning to rank
• Custom algorithms
• Fully integrated into Fusion index and query pipelines

BI Ready

• Out of the box search reports and dashboards
• Full SQL 2003 compliant interface
• Seamless integraRon with any SQL-based

Business Intelligence tool (e.g. Tableau)

• Numerous pre-defined jobs for common tasks
• Want to do your own math? Scala, Java and

Pyspark all supported

• Data Science Notebook support (Zeppelin,

Jupyter)

Fusion Architecture

SECURITY BUILT-IN Shards Shards

Apache Solr Apache Zookeeper

ZK 1

Leader Election LB

ZK N

Shared Config Management

Worker Worker

Apache Spark

Cluster Manager

REST API | SQL

Admin UI

HDFS (Optional)

Core Services

• • •

ETL and Query Pipelines Recommenders/Signals NLP Machine Learning Alerting and Messaging SQL and Catalog Scheduling Connectors/Federation

LOGS FILE WEB DATABASE CLOUD

ElasticSearch 3rd Party APIs Fusion App Studio Search Devs Client APIs/SQL Web App MOBILE ANALYTICS

Demo

Demo Details

• Ecommerce Data Set
• Product Catalog: ~1.3M
• Signals: 1 month of query, document logs
• Fusion 3.1 Preview showcasing:
• User-Item, Item-Item, Query-Query RecommendaRons
• Clickstream boosRng
• Query Intent classificaRon + Learn to Rank
• Rules Module
• Lucidworks App Studio UI (aka Twigkit)

Resources

• hnps:/

/lucidworks.com/online-retail/

• hnps:/

/lucidworks.com/fusion/ (3.1 very soon!)

• hnp:/

/lucidworks.com

Thank You