Conférence Smile au Bargento 2014
Intervenants : Farid Aït Abdeslam, Consultant ecommerce Smile et Aurélien Foucret, expert technique Smile.
Lorsqu’un moteur de recherche interne à un site fonctionne, les visiteurs deviennent plus facilement des acheteurs : le parcours de navigation est raccourci, la réponse à leur problématique est immédiate.
Smile met en œuvre aux côtés de Magento une technologie de recherche innovante basée sur ElasticSearch, améliorant les performances et ouvrant également la voie vers de nombreuses fonctionnalités telles que la recherche contextuelle, la recherche floue, la percolation, et des capacités de merchandising avancées… le tout Open Source.
2. 2
SPEAKERS
FARID AIT ABDESLAM
eCommerce Consultant
farid.aitabdeslam@smile.fr
@faitabdeslam
AURÉLIEN FOUCRET
R&D Manager
aurelien.foucret@smile.fr
@afoucret
Bargento
October 2014
3. 3
Bargento
October 2014
SMILE IN A FEW FIGURES
Leading European open source integrator
More than 20
YEARS EXPERIENCE
IN OPEN SOURCE
More than 35
PARTNERS
IN OPEN SOURCE
More than 50
MILLION EUROS
IN TURNOVER IN 2013
700
OPEN SOURCE
EXPERTS
25%
INCREASE IN TURNOVER
PER YEAR
4. 4
17 OFFICES
IN 7 COUNTRIES
Bargento
October 2014
5. 5
WHAT DO WE DO FOR YOU ?
Bargento
October 2014
CONSULTING
Project Framework / Audits / Benchmark
DIGITAL
Visual identity/ Ergonomics
Accessibility / Editorial
Strategy / SEO
HOSTING
Hosting / Corrective and
upgrade Maintenance /
Support
TRAINING
Change management
Intra and inter company training
ENGINEERING
Conception / Development /
Configuration
A 360°
SERVICE FOR YOUR
PROJECTS
6.
7. 7
OUR EXPERTISE AND OUR BELIEFS
Bargento
October 2014
ARE DETAILED IN OUR WHITE PAPERS
Web Culture
System
E-business
Content Management and DMS
ERP et Business Intelligence
OVER
100 000
DOWNLOADS PER YEAR
8. 8
OUR REFERENCES
70%
OF CAC 40
COMPANIES HAVE TRUSTED US
Bargento
October 2014
9. 9
SMILE IS THE LEADING EUROPEAN
SMILE IS THE BIGGEST OPEN SOURCE PLAYER IN EUROPE
700 employees, 17 offices, 300 new projects per year
MULTI-TECHNOLOGIES, A UNIQUE EXPERTISE
More than 35 technological partnerships, the highest level of recognition for most
Bargento
October 2014
OPEN SOURCE INTEGRATOR
10. 10
SMILE & MAGENTO
PARTNER OF THE YEAR
150 trained engineers
15 certified developers
80 projects
6 years of experience
45 000 man days produced
Gold Partner
Bargento
October 2014
11. 11
SMILE OPEN SOURCE SOLUTIONS
LAND OF INNOVATION
We did it first :)
Integration of key solutions to Magento
Bargento
October 2014
https://github.com/Smile-SA
13. 13
ELASTICSEARCH
WHAT IS IT ?
Distributed RESTful search & analytics server
Build on top of Apache Lucene
Drop in replacement for SolR
Open Source License
Apache 2
Unique features
Percolation
Aggregations
Hadoop Integration
Quick learning curves & deployment
Full text search
Bargento
October 2014
Schemaless
Real time indexing
Real time analytics
Distributed & High availability
14. 14
ELASTICSEARCH INC.
THE COMPANY BEHIND ELASTICSEARCH
Bargento
October 2014
2 HQ
US : Los Altos (California)
EU : Amsterdam (Netherlands)
Offices in France
Other Open Source projects (ELK) :
Big momentum on the market
8+ million downloads since 2010
> 500 000 downloads per month avg.
90 employees
Revenue growth : 400% last 6 months
Business Model
•Sales support and
training for
ElasticSearch
•3 levels of support
(Silver, Gold,
Premium).
Per node
•Up to 24/7 SLA
Investors
•Benchmark Capital
$10M Series A, 2012
•Index Ventures
$24M Series B, 2013
With participation
from Benchmark
Capital
•New Enterprise
Associates $70M
Series C, 2014
With participation
from Benchmark
Capital and Index
Ventures
ElasticSearch Logstash Kibana
17. 17
SEARCHANDISING
LESSONS LEARNED
What is Searchandising ?
1 - Search
o Up to 70% of the visitors use the search bar
2 - Merchandising
o new
o special offer
o promotion
o destocking
Bargento
October 2014
18. RELEVANCE = FULLTEXT RELEVANCE + EXTERNAL SIGNAL RELEVANCE
18
SEARCHANDISING – LESSONS LEARNED
WHAT IS RELEVANCE FOR AN ECOMMERCE WEBSITE ?
Fulltext relevancy
What field is used into search ?
How are field analyzed ? What are their respective weight ?
Fulltext is only a small part of what matter when dealing with an
eCommerce website
Other parameters of relevance are :
Business rules
Customer behavior
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
19. 19
SEARCHANDISING – LESSONS LEARNED
FOCUS ON CUSTOMER BEHAVIOR
Data can help you to make better decision from insights on customer
behavior
Find new relevant optimization based on trends, sales, …
Is your search engine able to apply rules on insights ?
Search optimization is an Iterative process :
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
Collect
data
Analyze
behavior
Optimize
engine
20. 20
SEARCHANDISING – LESSONS LEARNED
OTHER IMPROVEMENT AREA
Rich autocomplete
Get the user to the right content directly
Multiple content type : products, category, …
Rich faceting features
Multiple select facets
Range sliders
…
Category built on search engine rules
Personalization / Contextualization
Experiment optimization through A/B testing
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
21. 21
SEARCHANDISING – LESSONS LEARNED
WHAT ABOUT EXTERNAL SEARCH ENGINES ?
Use self-hosted search engine instead of SaaS solution
Weak network introduces latency and less security
Customization
The editor have his own roadmap
No way to have your own
Indexing
Most of the platforms does not provide real time indexing leading to inconsitencies between database
and search engine
Inconsistencies lead to the « holes into product list syndrom »
Retroaction on the platform not smoothly managed (Magento caches,Varnish ,…)
Unconsistent features
Most external searchandising are not well integrated into Magento
2 admin interface ; one for Magento, one for the searchandising
Useless feature : not consistent with the way Magento works.
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
22. 22
MAGENTO
& ELASTICSEARCH
Bargento
October 2014
23. 23
MAGENTO & ELASTICSEARCH
OUR MODULE GOAL
An Open Source module
Smile Magento ES a.k.a Smile Searchandising Suite
Integration of ES into Magento
Drop-in replacement for SolR
Combining our long experience with OSS Search Engine
High quality implementation
Does not break Magento standard feature
Built for extensibility
Upgradable : our module is upgraded with every new Magento version
The ultimate goal : give merchant tools to improve conversion
with state of art features
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
24. 24
MAGENTO & ELASTICSEARCH
KEY FEATURES – FULLTEXT SEARCH
Fulltext Search improvements
Fine control on what fields are used in search and their respective weight
Fuzzy search : search for « Frankenshten » and get the results for « Frankenstein »
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
Rich faceting toolkit
Multiselect facets implementation
Price ranges, rating filters, …
Easier customization of the facets logic and templating by developers
SEO optimizations
Autocomplete framework
Autocomplete popular searches, products and categories
Customizable templates (e.g. : display product images and prices)
Pluggable models allowing developpers to easily add their own autocomplete (store,
CMS, …)
25. 25
MAGENTO MODULE
KEY FEATURES – MERCHANDISING
Behavorial data collect : get richer from your clients
Our module collects all your clients behavior and creates a knowledge database you can query at will
Real time data refresh : don’t miss the trends
Use it into merchandising optimizations rules
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
Smart categories
Build categories by defining search engines queries and not by
picking products one by one
o Ex : « All discounted products that are in stock »
Nice UI for rules administration
Optimizations tunning – Rules based relevancy
Append business rules to your catalog to modify relevancy
o « Boost all products which are in stock »
o « Boost all products of category HighTech »
Pluggable system : developers can append new models easily
26. Many extension hooks shipped by default (templating, custom optimizations models, and new features)
26
MAGENTO MODULE
KEY FEATURES
Built for Magento, into Magento
Deep Magento integration.
Operate everything from Magento admin. No additional back-office.
Optimized data indexing. Get your catalog updated in real time and reduce your time
to market.
Your business is unique. Your search engine should be too.
Index what you need in custom indices (stores, cms content, …)
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
Designed for extensibility
ElasticSearch : a scalable and secured backend
Scale over many servers when your catalog or your business grow
Avoid downtime with : automated failover when a server fails or during maintenance
Snapshot and restore in minutes the whole cluster. Easy to backup.
Monitoring tools available from ElasticSearch Inc. (Marvel)
28. Smile_SearchOptimizer
Product Data Indexing User Behavior Indexing
28
ELASTICSEARCH & MAGENTO
GLOBAL ARCHITECTURE
Technology stack :
Website Usage Collect Data
Smile Webtracker
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
Website Users
Smile_Modules :
Search Queries
Catalog Index
Behavorial Data Index
Smile_ElasticSearch
Smile_VirtualCategories Smile_Tracker
29. 29
ELASTICSEARCH & MAGENTO
OPEN SOURCE RELEASE
First release : technical core containing
Indexing, search and autocomplete
Available as soon as Friday on our GitHub
Compatibility :
Magento EE 1.13.x & 1.14.x and Magento CE 1.8.x & 1.9.x
ElasticSearch : 1.2.x & 1.3.x
Later this year release of more modules :
Virtual categories & Search optimizers
New features (see roadmap)
Bargento - Boost your searchandising with ElasticSearch
07 Oct. 2014
http://github.com/Smile-SA/
31. 31
ROADMAP HIGHLIGHTS
RECOMMENDATIONS
Collected data will allow us to build recommenders into our
module :
Cart recommendations
User recommendations
People who viewed this products are also interested by this product
Recurring bought products
…
Machine learning algorithms :
From Spark Mlib
Customizable
Other :
More like this into ES (avoid cold start)
Manual recommendations by the merchandiser
Bargento
October 2014
32. 32
ROADMAP HIGHLIGHTS
SEARCH CONTEXTUALIZATION
Search relevance is matter of context
Two kinds of context :
User related :
o A customer looking only at « Women products »
o A customer buying only products with size « M »
Navigation related :
o An user has launched a search from the « Book » category
In both case we should be able to influence relevancy from these
contexts buy boosting product matching the context
The more an user is using your website, the better his shopping
experience is
Bargento
October 2014
33. 33
ROADMAP HIGHLIGHTS
PROMOTION TOOLKIT
SalesRules in Magento :
Defined by rules (just like VirtualCategories)
Can be associated to banners
Goals :
Better relay of your promotional operation
Trigger the right banner when an user is « near » a promotion
Based on the defined rules
Display tags and labels on products if it match a promotion
To provide this features we will use an feature that is unique to
ElasaticSearch : the percolator
Bargento
October 2014
34. 34
ROADMAP HIGHLIGHTS
CROSS CHANNEL / MARKETPLACE
Magento catalog is products centric
It’s great for your eCommerce main
offer
Need for an additional layer into
Magento : the offer
You want to sell produts for other merchants
You are a retailer with many point of sales
ElasticSearch will be central in our
effort to integrate the cross-channel /
marketplace logic and present offers
Product
(eCommerce)
Bargento
October 2014
Offer Paris
Store
Offer
Merchant 1
Offer London
Store
Offer
Merchant 2
35. 35
ROADMAP HIGHLIGHTS
MAGENTO 2 SUPPORT
Will the module be outdated with Magento 2 ?
Our module will be updated with Magento
This include Magento 2
Bargento
October 2014
Notas do Editor
Vous connaissez Magento et nous venons de vous présenter ElasticSearch,
Faisons maintenant un pas en arrière pour entrer dans le vif du sujet… « Booster le Searchanidising avec ElasticSearch »
Searchandising : terme maketing qui regroupe deux choses
Search : la recherche
Merchandinsing : l’art de mise en avant des produits afin qu’ils se vendent « tout seul » sans intervention d’un conseiller afin d’augmenter la rentabilité / booster ses ventes
Business Rules Does the product have an image, Is the product in stock, Do we have an agreement with the manufacturer to highlight the product ?
Rating, Products sales, Products viewed, Trends, … ?
Search engine feature ?
Merchandising team size ?
Need for automation ?
Performance evaluation
Data driven
Quick wins ? Plan for better relevancy
Magento et ElasticSearch
Vous devez surement vous posez la question « pourquoi ? »
Magento, moteur de recherche basic, full text en plus le fait très mal