MICES US 2022

Registration

Welcome to MICES! - Welcome to the e-commerce search community!

Welcome by the organizers and logistics

A model for understanding e-commerce search

René Kriegler
Co-Founder of MICES / OpenSource Connections

When we started MICES in 2017, one of our main motivations was to meet with other people who work in the field of e-commerce search. We felt that not much had been understood about search in e-commerce. It didn't even seem to be a domain on its own right, especially compared to search in other domains, such as Enterprise Search or general web search.

Five years on, e-commerce search has become much more recognized. Partly, as there is a growing awareness that search is a key factor to a successful online shopping experience, but probably also because the commodization of machine learning and its application to search seems to promise great business yield.

On the other hand, at least in the open source search engine world, the available implementations for ranking are still the ones that were made for document retrieval and not for products. And search optimization work still seems to focus on search relevance but hardly on consumer experience.

In this talk, I shall present a model for understanding e-commerce search that starts from the consumer perspective and that can serve as guidance for search optimization work.

René works as Director E-commerce at OpenSource Connections, leading the company's e-commerce search team and practice.

He has worked in search for more than 15 years, including on projects for some of the top 10 German e-commerce sites. René is co-founder and co-organiser of MICES. His technological focus is on Solr, Elasticsearch and Lucene. He created and maintains the Querqy open source library for query rewriting and he is co-initiator of the Chorus project – an open source software stack that combines Querqy with other powerful tools to build e-commerce search and to measure and improve search quality.

Slides
Video (Youtube)

Break

Focus session: E-commerce search teams and processes

A Search Divided Cannot Stand - Shopify's Data-Engineering Partnership

Doug Turnbull & Chen Karako
Shopify

Imagine being a carpenter with one hand tied behind your back? That’s how it can feel when building e-commerce search without embracing both data and engineering roles.

In this talk, we share a core principle of Shopify’s search team - data and engineering collaboration. In past lives, we’ve seen what happens if you don’t take this ideal to heart: slow, broken machine learning models developed without engineering input crash and burn, we’ve seen engineers tweak relevance uninformed by experimental data – just because the results seem to ‘look good’.

At Shopify search data and engineering collaborate so fluidly, you would be hard pressed to tell who’s a data scientist and who’s an engineer. Rather than search being an arcane specialty, we’ll argue the best, modern relevance specialists, must actually be ‘professional learners’ - eager to embrace both the data scientist and engineer roles.

We go beyond the philosophical, to the practical, organizational work to build this fluidity – whether it’s how we recruit for our team, ensure an environment of safe collaboration, or handle alignment between our organizations, we will share the nitty-gritty, real-life story of how we arrived at a mature, highly collaborative search organization.

In 2012, Doug saw search relevance would be central to user experiences. Sadly, search relevance was a topic clouded in esoteric mystery. He began to democratize this daunting field through blogging and speaking. In 2016, Doug’s book Relevant Search woke the industry up to the importance of search quality. Doug now works at Shopify, working to make great commerce search+discovery possible for everyone. Doug co-created Elasticsearch Learning to Rank with the Wikimedia Foundation and contributed to Trey Grainger’s upcoming AI Powered Search. You can find Doug at his website where he blogs about search and data.

Chen works as Data Science Lead of the Discovery Experience Data team at Shopify, building search engines and recommendation systems to improve product discovery.

Slides
Video (Youtube)

Discussions and breakouts

All Participants

Suggested topics for your inspiration:

• What roles should be present in an e-commerce search team?
• Getting funding for a search team - what's the right size of the team?
• Challenges and best practices for growing and scaling search teams
• Get to know your peers and learn from each other: Who works as product manager, data scientist, searchandizer, ... - and what problems are they solving?
• How to communicate about search with stakeholders outside the e-commerce search team?

Lunch

Focus session: The technical implementation of e-commerce search

Own your e-commerce search with Chorus

Daniel Wrigley
OpenSource Connections

Successful online retailers need competitive advantages over their contenders. Owning search is a key to success as it enables you to tailor search results to the buyer’s and seller’s interests. But owning a search is not child’s play. Owning your search means taking care of indexing your data, establishing processes around active search management, providing an offline testing framework, generating judgments, … Search teams need to cover a lot of ground to be successful!

But worry no longer, Chorus is here to support these efforts! Two years ago, Chorus was initially released with Solr as its search engine. Now, it’s also available for Elasticsearch and combines best-in-class open source tools to shorten the amount of time you need to set up an e-commerce search stack based on Elasticsearch.

This session will introduce you to this stack containing production-ready tooling for query rewriting, search management, assessing and tuning search relevance. You will learn how you can use this stack to not only own but also optimize your e-commerce search.

Daniel has worked in search since 2012 and is now part of OpenSource Connections’ e-commerce team where he can follow his passion for search and natural language processing. His experience as a search consultant paved the way for becoming an O’Reilly author co-authoring the first German book on Apache Solr. He enjoys combining multiple open source tools like Apache Zeppelin or Apache NiFi to build powerful search stacks to deliver relevant search. In his free time Daniel supports the local fire brigade as a volunteer firefighter and serves as the sports director of the local shooting club in the village he lives in.

Slides
Video (Youtube)

Survey: Your toolstack for E-commerce Search

All Participants

During the day we will ask MICES attendees about the toolstack that they use for e-commerce search and what is the one thing that they like the most about it and what is the one thing that they definitely want it to change. In this short session, we will share the results of the survey.

Relevance is more than just algorithms

Johannes Peter
Principal Search Consultant

Relevance algorithms do not end in themselves. They are developed to serve customers and to convey the offering of a company to them. There is a simple recipe to continuously address the needs of customers: keeping improvement iterations as short as possible while testing as much as possible. Algorithms can simply be considered as a transient state in these iterations. This is why the essence of relevance is the continuous improvement, whereas algorithms and techniques are merely details.

In addition to serving customers, companies require relevance engineering to be organizationally scalable in order to accelerate development on demand, e.g. by implementing more features simultaneously or in shorter time. This has become critically important in the last years as many companies have increased their investments into online retail in response to changes of customers' behavior in the context of the pandemic.

The main reason as to why keeping iterations short and organizational scaling are both very challenging in the context of relevance is the complexity that comes with the diversity of expertise that is required to implement search in general. Furthermore, the open source context has been historically grown in a way that relevance is centralized at a single place. However, the only way to tame these challenges is to introduce well-defined interfaces within and especially across teams in order to strive for an independent and efficient relevance engineering.

The talk will evaluate relevance from an end-to-end perspective (including pre- and post-processing of queries) on the basis of real-world examples. This will include discussions as to why machine learning should always be driven by API-first thinking, how serverless technology is able to accelerate relevance engineering significantly, and how re-ranking helps to achieve a better abstracted relevance architecture.

Johannes is a passionate leader, consultant and engineer in the areas of search, data and cloud. He is experienced in driving and implementing solutions from initial ideas towards systems running in productive environments and measurably contributing to business KPIs.

Throughout his entire career, Johannes has been passionate about open source. In his early days, he contributed several features to Apache NiFi to improve its integration with Apache Solr. Later he became a main contributor and committer for Querqy, a query rewriting solution for Apache Solr and Elasticsearch used by various big retail companies worldwide.

Slides
Video (Youtube)

Break

A story of auto-completions

Lucian Precup & Radu Pop
all.site & Adelean

There are auto-completion and auto-completions. We can count at least five different types of use cases and implementations. For most of them, a specific dictionary and dedicated index needs to be constructed. In this presentation, we will show the different possibilities Elasticsearch and Solr offer on different use cases of this search before search that we call auto-completion.

Lucian Precup is CTO of all.site - the collaborative search engine developed at Station F in Paris. With his colleagues at Adelean, Lucian develops solutions for indexing, searching and analyzing data. Lucian regularly shares his knowledge in specialized conferences and organizes the Search & Data Meetup in Paris.

Radu Pop provides Consulting Services as Solutions Architect at Adelean. He handles projects around Elasticsearch and Adelean’s A2 search technology. He oversees the integration and evolution of search engines within large e-commerce platforms, marketplaces or organizations' data lakes. Prior to joining Adelean, Radu acquired a solid experience in Web archiving, operating large scale crawling systems in the context of several European research projects. He holds a PhD in Computer Science and a MSc in Distributed Systems.

Slides
Video (Youtube)

Self-organizing sessions / Barcamp

All Participants

Closing session