MICES 2019

Mix-camp E-commerce Search

When

19th June 2019 - main event
Starting at 8:30 am

18th June 2019 - pre-MICES class
Starting at 9:00 am

Note to participants

If you are registered as a participant but are unable to attend the main event on 19 June, please inform us as soon as possible so that your seat can be given to somebody on the waiting list.

Pre-MICES class 'E-commerce Search for product managers' (18 June - sold out)

OpenSource Connections and Tudock have prepared a full-day class on e-commerce search for product managers (beginner to intermediate level) to take place on 18 June - the day before the MICES main event.

This event is sold out. Get more information here! (external link).

MICES - Main event (19 June)

In the search community, e-commerce search has received less attention and is less established than search in other domains, such as enterprise search or general web search. But it is a very exciting domain with a lot of challenges. That's why we started MICES in 2017 as an informal one-day event to bring together experts of different backgrounds - such as IT, product managers, UX designers, search managers, information retrieval specialists, search engine vendors - to discuss challenges, ideas, best practices, and case studies in the e-commerce search domain.

Format

MICES brings together participants from a variety of backgrounds, all sharing a common interest in e-commerce search. In order to stimulate and facilitate discussion, we will start with scheduled talks in the morning. This will be followed by self-organising sessions: participants are encouraged to initiate discussions about their topics of interest or to give an ad hoc presentation.

Topics

The workshop welcomes all topics related to e-commerce search. We have compiled a list of topics that will probably feature at the workshop. This list is not comprehensive. Participants are welcome to discuss further topics at the workshop.

  • Managing e-commerce search: finding the right process and organisation for it, and how does it change over time?
  • How can we measure and improve search quality for e-commerce?
  • What are the best search relevance models for e-commerce search?
  • Personalised e-commerce search
  • Choosing the right search technology and search management tools
  • From keyword search to virtual shopping assistants: designing the e-commerce search user experience
  • Exploring artificial intelligence, visual and voice search for e-commerce
  • How to deal with poor data quality in e-commerce search

Past events

Programme

 

08:30

Registration

09:00 - 10:00

Introduction

René Kriegler
Freelancer

Search Relevance Organizational Maturity Model

Eric Pugh
OpenSource Connections

Smarter search drives value to your business. Delivering search that matches users to the right content (jobs, products, articles..., whatever) is what you care about. But organizations often get stuck getting there. Why? It turns out that you need quite a number of very different ingredients to deliver tremendous search. You need the intelligence to understand what users are searching for and whether they're satisfied. You need the domain expertise, infrastructure, and data science to extract meaningful features from your content, user personas, and user queries. More mundanely, you need to successfully install, scale, and operate a search engine!

All of this can send your head spinning! In this talk, Eric Pugh, a long-time Search practitioner, discusses his broad experience across many e-commerce organizations delivering smarter product search. He introduces the maturity model used by OpenSource Connections to help think through where your team is located on its road to smarter search from Basic to Advanced. He then dives deep into the two most technical areas of the seven areas in the maturity model: Experiment Driven and Content Driven.

This talk will give you a much stronger understanding of how to measure your capabilities in two key areas of search, and a more general understanding of how to understand your teams maturity across all seven different areas in order to drive better ROI from search.

Come and learn from Eric your next steps on the road to delivering great search!

Fascinated by the craft of software development, Eric Pugh has been involved in the open source world as a developer, committer and user for the past fifteen years. He is an emeritus member of the Apache Software Foundation and continues to be very active in the Solr and Tika projects! In biotech, financial services, and defense IT, he has helped European and American companies develop coherent strategies for embracing open source software. Eric became involved in Solr when he submitted the patch SOLR-284 for extracting text from binary files (such as PDF and MS Office formats), that subsequently became the single most popular patch as measured by votes! He co-authored the book Apache Solr Enterprise Search Server, now on its third edition. Today he helps OSC’s clients build their own search teams and improve their search maturity, both by leading projects and by acting as a trusted advisor.

10:00 - 10:15

Break

10:15 - 11:20

Offline Evaluation in E-Commerce Search: Applications and Requirements

Jens Kürsten & Andreas Wagenmann
OTTO

Making changes to existing data structures or the search logic require a mechanism to reliably evaluate those effects offline before shipping them to production. Further, tuning a search system often involves a wide parameter space, requiring high throughput to be able to find optimal settings. This led us to develop an offline evaluation framework, which enables us to evaluate both changes in result selection as well as ranking on the basis of judgements generated from user interactions with search results.

In this talk we are presenting the framework, the process of generating the data needed for the evaluation and use cases that show how we applied it to distinct search problems. We will also discuss how those are reflected in the requirements to such a framework and used evaluation metrics. Along the way, we will also provide insight into the reasoning behind and approaches to the development of the evaluated search system.

This includes the incorporation of business aspects, such as availability, delivery time, and profit margin. These vary between different types of products and assortments and are shown to improve the business outcome significantly if used for the ranking and requires optimizing weights for different business factors depending on general or category-specific goals. Further, we are looking at our first steps towards data driven interpretation of the user's intent and incorporation of this information during the query pre-processing stage along with the influence on result size and ranking.

We will also describe how we managed to ensure search quality when the structure of product data changed significantly, reflected in a change of fields and their respective content.

Jens is a search engine enthusiast working at OTTO GmbH, always pushing hard to find and implement the most relevant metric(s) to quantify search relevance. He graduated from Chemnitz University in Germany while researching methods for organizing and searching audio-visual archives of local TV stations. He is also interested in the inter-connected domains of natural language processing, computer linguistics, and machine learning.

Andreas is a developer and search specialist at OTTO GmbH, always happy to find new search problems to dive into to come up with solutions. He graduated from Saarland University with MSc in Bioinformatics, developing a model for protein folding simulations after studies in germany and abroad, spanning the fields of physics, computer science, AI and psychology. He finished a BSc in psychology besides working full time as developer. Also interested in ML, NLP, CL, functional and concurrent programming.

Measuring and Optimizing Findability in e-commerce Search

Andreas Wagner
search|hub

In this talk, we shed some light into the problem of evaluating whether results served by an e-commerce search engine for a given query are good or not. Findability is a critical question in evaluating any e-commerce search engine.

Based on our large-scale user interaction logs @search|hub.io, we will show you how using simple metrics like query CTR or time-to-click in this domain can be misleading or even result in wrong judgments. We'll introduce a new Findability-model that aims to better learn the quality of the results based on the user's interaction with the results and demonstrate the feasibility, efficiency, and accuracy of such a model in predicting query performance.

Andreas has been building and optimizing search engines for almost 12 years. He worked for FACT-Finder and Fredhopper in the past and is co-Founder & CTO at CXP Commerce Experts.

11:20 - 11:40

Break

11:40 - 12:45

Building the ecommerce dream team: Does AI enabled search need an algorithm in charge?

Peter Thomas
Attraqt

In his presentation the CTO of Attraqt, Peter Thomas, will examine what we need from the ecommerce team of the future (or the now). With more than three decades of experience watching technology evolve, Peter is a firm believer in people being the real power behind technology. Drawing on Attraqt’s Fredhopper legacy of almost twenty years pioneering search in ecommerce, Peter will ‘interview’ the ideal ecommerce candidates both algorithms and humans.

The audience will leave with a greater understanding of the following in the context of search as it becomes increasingly important in driving the customer journey and the wider ecommerce landscape:

  • The rise of the hybrid business team with obvious data skills, optimisation, business/shopper knowledge and product/project management skills
  • The right questions to ask of the algorithms being deployed to ensure results are optimised for long term goals, not just short term for ROI
  • What pitfalls to look out for when assessing the limits of algorithms, and how to overcome this
  • How to create a workforce built on cross-functional teams (humans and algorithms), and assessing the best “AI or person for the job”
  • How to design algorithms with reinforcement learning in mind

Peter has more than 30 years' experience that began with micro-electronics, then evolved through business systems into the SaaS sector where he has worked for over a decade. Peter joined Fredhopper in 2016 as CTO and continued the role with the Attraqt Group following the acquisition. A firm believer in people being the real power behind technology, his main responsibilities include product development, leading innovation teams, and driving key collaborations to ensure the best services are delivered to clients. Previously, Peter served as CTO of the cloud division for IRIS Software and Director of Development at Betfair across the UK, Romania and Portugal.

Combining BigQuery, user tracking and scoring model in the search management context

Diego José de Calazans & Georg Wolf
MediaMarktSaturn

Automatization is always a big challenge in the area of search management. Search managers have limited capacity concerning how far the long tail area can be reached and kept under control during daily business. Still there is a clear necessity to secure revenue beyond short head queries and ultimately to fulfill customer expectations.

Machine learning models and third party search technologies try to provide tools and solutions for this puzzle, but many of those models are still projects under development and therefore far away from being part of daily search management activities.

Based on the experience we gathered with such search tools and with classic search quality measurement KPIs at MediaMarktSaturn, we have developed a Search Management Dashboard in PowerBI. At the core is a BigQuery-based collaborative filter. The dashboard combines the insights from customer behavior tracking and predictive algorithms with the need of the search managers to have faster qualitative insights into the search engine performance for long tail queries.

This approach has been so fruitful and precise in generating insights, that we now intend to use its core relevance model as part of the relevance scoring in an open source based search solution that we will build in-house.

Diego has over 7 years of experience in e-commerce search during which he gathered experience with some of the leading search tools in the market. He currently works for MediaMarktSaturn, where he introduced the idea of KPI-driven search management and helped establishing the processes around it. Working very closely with the product teams, Diego is currently the business owner for search, recommendation and overall shop performance.

Georg is a data scientist at MediaMarktSaturn where he works on search relevance optimization. After completing his studies in Business Administration and Information Management in 2012, his career has focussed on topics that combine data handling and economic thinking. For the Scout24 Group, Weg.de and MediaMarktSaturn, he worked in product management, online marketing, data analysis, data warehousing and data science to identify performance differences, better understand them and increase value for customers.

12:45 - 13:45

Lunch

13:45 - 14:50

Coping with the challenge of sorting large product catalogs

Cagdas Senol
Zalando

Finding relevant documents is a challenging task. In Zalando, we have over 450k articles for our customers. Their availability, prices, and other features change over time quite frequently (over 2k per second). Sorting this amount of products with high velocity is a challenging task. In the talk, I will discuss how a small team in Search evolved into a dedicated Data Driven Sorting team. I will show how we steer sorting towards different goals and how our Sorting architecture allowed our team to conduct numerous A/B tests in a year and how we impacted financial KPIs with modern sorting architecture.

During this journey,

  • We decoupled updating articles from updating sorting signals with a stream processing solution.
  • Moved from business-driven popularity calculations to learn to rank by leveraging existing architecture.
  • Built a testing framework for testing Elasticssearch painless scripting.
  • Set up monitoring solutions to observe the quality of sorting over time.
  • ntegrated customer understanding signals into sorting such as Customer’s best suitable size, brand affinities, etc for real-time personalization.

Cagdas Senol is a senior software engineer working at search department in Zalando. He built recommendation systems, data pipelines, and information retrieval solutions. He loves listening to podcasts. His favorites are Planet Money and Darknet Diaries.

Personalizing search results in real-time

Roman Grebennikov
Findify

You type a query in a search box and get great search results back. Sounds pretty simple, right? But the same search results ranking can be great for you, but not for someone else.

You type a query in a search box and get great search results back. Sounds pretty simple, right? But the same search results ranking can be great for you, but not for someone else.

With widely known Learn-to-Rank process, you need to deduce which search item was relevant for a user in the past to get the better ranking in the future. But relevancy does not depend only on the query itself, it’s also dependant on the current context. For example, for the same “shoes” query:

  • during the winter, slippers are not that popular;
  • people of different genders will probably prefer different models;
  • if someone landed on a specific product from google search before, he may be looking for a specific brand.

In this talk Roman will share a lessons learned experience on building personalized search platform for thousands of merchants at Findify:

  • Does better search relevancy improve sales? (spoiler: yes)
  • Why does real-time reaction matter?
  • Fighting position bias while training the ML models, when people are clicking on the first results because they are just first.
  • Solving cold-start problem for fresh merchants with transfer learning.

Roman Grebennikov is a passionate software developer and ML engineer with hands-on experience in JVM, high performance computation and algorithmic research. During last years he has focused on delivery of functional programming principles and practices to real-world data analysis and machine-learning projects, focusing mostly on IR & search problems.

14:50 - 15:10

Break

15:10 - 15:40

The Mind and Body of Search

Angel Maldonado
Empathy.co

Search succeeds not only because of its capacity to function but also because of its capacity to make us feel.

When thinking Search we shall recognise that the body of Search (the Interface) has the power to turn sophisticated technology into failure and simple technology into success.

In this talk and through the following three questions, we will discuss what makes a good search (in-feeling), what type of feelings can Search evoke and finally close with a theory of Morphological Search (one that is capable of adapting its presentation or body).

  • What makes a good search?
  • Can Search evoke emotions?
  • Can Search become expressive?

The talk will resource to real samples as well as research concepts and interaction experiments from Empathy.co Labs.

Angel Maldonado has spent the last 19 years passionately developing and executing eCommerce Search and Navigation solutions. Having studied Computer Information Systems at Liverpool University, Angel started his career working for Autonomy before founding EmpathyBroker. He also recently spent a short period seconded to Rich Relevance’s Exec team to help them develop and advance their new search offering (Find). At Empathy.co, Angel drives forward the product vision, innovation and ethos of creating a company with truly organic values. Angel is an evangelist and enthusiastic blogger who regularly speaks at industry events. He likes to explore the future of Commerce search, placing the importance on people, their emotions and unpredictability rather than data and processes to create experiences that generate an emotional connection. A keen sea sport enthusiast, Angel plays the guitar and enjoys poetry and meditation. He’s lived in the US, Spain and the UK, and now splits his time between London and Gijon (Asturias).

Registration - waiting list available

Unfortunately, all seats at MICES 2019 are taken. To join our waiting list, please use this form. We will notify you as soon as a seat becomes available.

If you are already registered as a participant but are unable to attend the event, please inform us as soon as possible so that your seat can be given to somebody on the waiting list.

Fill out your waiting list registration information



 

Please fill in the form to join the waiting list for MICES 2019. You can also follow us on Twitter for the latest information about MICES.

Organisers & Partners

René Kriegler

René has been working as a search consultant for clients in Germany and abroad for more than ten years. Although he is interested in all aspects of search and NLP, key areas include search relevance consulting and e-commerce search. His technological focus is on Solr/Lucene.

Paul Maria Bartusch

Paul is a freelance consultant in the e-commerce domain. He supports his client as product manager, Scrum product owner and business development consultant, focussing on search (onsite, SEO/SEA), mobile and responsive application development, and agile transitioning.

ESEMOS

ESEMOS is a German company focusing on web and e-commerce search. Their customers range from small online shops to international web portals. They appreciate ESEMOS's hands-on approach and their long experience in the search industry.

Berlin Buzzwords


MICES is partnering with Berlin Buzzwords and takes place on the day after the main conference. See here for the Berlin Buzzwords tickets.

Sponsors

  • myToys Group
  • OpenSource Connections
  • Otto
  • ESEMOS
  • Tudock
  • reBuy

Location and Venue

Hosted by myToys (main event) & reBuy (pre-MICES class)

Venue

Main event: myToys Group

Potsdamer Straße 192

10783 Berlin

Germany

Pre-MICES class: reBuy GmbH

Potsdamer Straße 188

10783 Berlin

Germany

Nearby Public Transport

Underground Train Stop (U-Bahn)

U7 Kleistpark

Overground Train Stop (S-Bahn)

S1, S2, S25, S26 Yorkstraße

Bus Stop

106, 187, 204, M48, M85 Kleistpark

See www.bvg.de/en to plan your journey.