Plain talk on search and product discovery.
Notes from a team of search engineers. Numbers over buzzwords, mechanism over hype.
Education
The Search Tax — Part 3 – How to Stop Paying It
The third and final part on the hidden cost of poor search, and the concrete steps that stop you paying it.
Education
Part Number Search in B2B Ecommerce: Why Your Product Discovery Technology Gets It Wrong (and How to Fix It)
Part numbers break most search engines. Here is why, and how to make exact-match and partial part lookups work in B2B.
Education
The Search Tax — Part 2 – The Promises That Cost You
Part two on the search tax: the vendor promises that quietly cost you more than they save.
Business
The Search Tax — Part 1
What the search tax is, where it hides in your funnel, and why most shops never see the bill.
Business
E-commerce Search Optimization: Retrospect, Outlook, and Our Community Commitment
A look back at a year of search optimization, where the field is heading, and our commitment to the community.
Business
Mind the Gap: Quantifying Vocabulary Mismatch in E-commerce Site Search
We quantify the vocabulary mismatch between how shoppers type and how catalogues are labelled, with data.
Education
The Informative Guide to Eliminating "No Results Found" Pages in Niche Ecommerce
A practical playbook for removing dead-end no-results pages in niche and long-tail catalogues.
Education
A Practical Guide to Implementing Semantic Search for Your Mid-Sized Ecommerce Store
A hands-on guide to adding semantic search to a mid-sized store without a full replatform.
Business
Simplify Your Apache Solr Configurations with Jinja2: One Template to Rule Them All
Using Jinja2 to keep Apache Solr configurations DRY, with one template to manage them all.
Business
Postmortem Incident March 31st, 2025
A transparent postmortem of the March 31st, 2025 incident: what happened, why, and what we changed.
Education
NeuralInfusion™ — A Blueprint Towards Relevant And Efficient Ecommerce Vector Retrieval — Part 3
The final part: why query enrichment beats fine-tuning, and how intent mapping lifted precision by more than 25%.
Education
The Behavioral Science Dilemma in Ecommerce
Why decode deep motivations with behavioral AI when shoppers state their intent in the search box?
Business
Merry Christmas and a Successful Start to the New Year!
2024 in review: 21 new searchHub extensions, 15 new customers, drinking-water wells for 25 villages in Zambia — and a first look at the NeuralInfusion hybrid search planned for 2025.
Education
Beyond Clicks and Scrolls: The Computational World of Online Retail Connections
What vector search actually computes, where it falls short in retail, and the business layer NeuralInfusion adds.
Education
AI-Redirects – Capitalizing on Retail Media Opportunities
How keyword-cluster performance from AI-Redirects turns into targeted, data-backed retail media ad space.
Education
Beyond Redirects: Leveraging AI to Streamline Site Search and Boost Customer Experience
Why long-tail redirects are exhausting to maintain by hand, and what keyword-level tracking tells you that hit counts never will.
Education
NeuralInfusion™ — A Blueprint Towards Relevant And Efficient Ecommerce Vector Retrieval — Part 2
How NeuralInfusion picks the queries worth improving and infuses better results into any search engine — no replatforming.
Education
Beyond Keywords: Using AI to Transform Site Search and Customer Navigation
CMS keyword pairing, SEO lists and RegEx all break down at scale. Here is why intent-based redirects don’t.
Education
NeuralInfusion™ — A Blueprint Towards More Relevant And Efficient Ecommerce Vector Retrieval — Part 1
Where hybrid search really helps, where it quietly costs precision, and why we needed a different approach.
Business
Why Librarians Are the E-commerce Site Search Pioneers
The librarian knew where your book was and what to read next. Site search forgot both. AI-Search-Redirects bring the sorcery back.
Education
TESUTO: How We Created a Unique Experimentation Stack for Search
How we built an A/B testing stack that randomizes on search trails instead of users and runs thousands of query experiments at once.
Education
Common Pitfalls of Onsite Search Experimentation — Part 2
Session-based A/B analysis attributes whole baskets to search. Why that hides real search improvements — and what to measure instead.
Education
Common Pitfalls of Onsite Search Experimentation
Two traps in search A/B tests: testing too broadly, and session-level randomization that lets carry-over effects hide the real result.
Press Release
FYAYC and searchHub Partnership Announcement
searchHub partners with foryouandyourcustomers, the commerce consultancy that puts art and the human experience at the foundation of business change.
Press Release
searchHub gewinnt renommierten K5 Commerce Award für Search & Marketing
Berlin, Juni 2023: searchHub gewinnt auf der K5 den Commerce Award für Search & Marketing — die Pressemitteilung der CXP Commerce Experts GmbH (Deutsch).
Education
Unlock Site Search Potential for Brand Searches & Retail Media Ads
If you change one thing about your site search, make it brand searches: variants, landing pages, suggest placement and retail media — in that order.
Business
Transforming E-Commerce Site Search: The Power of searchHub’s AI-Driven Approach
Efficient, flexible, transparent, automated: why an AI layer that optimizes your existing site search beats replacing it — with two customer voices.
Business
searchHub New Feature Announcement
How the new query picker chooses the best spelling for every cluster: Findability, Sellability, Reliability, Popularity and Sample Size. In English and German.
Business
A Recap of searchHub.io — Supercharging Your Site Search Engine
Five years after launch: what held up, what we learned the hard way, and where the platform went next.
Education
Towards a Use-Case Specific Efficient Language Model
Why Transformers cost 100× more per query for short, sparse ecommerce searches — and how a pruned, weighted, contextual N-Gram model reached 91% accuracy at 0.0023 ms.
Business
Y1 and searchHub Partnership Announcement
searchHub partners with Y1, the boutique agency that has been a staple of European ecommerce for over 20 years — from online retail to B2B and industrial manufacturing.
Education
How to Approach Search Problems with Querqy and searchHub
Term mismatch, typos, compound words, semantic terms, terms in context: which Querqy rewriter fixes which search problem, where rules stop scaling and what searchHub takes over.
Business
Why searchHub?
Over 90% of ecommerce queries are long tail, and no team can optimize them by hand. How searchHub clusters variants, picks a master query, and lifts search revenue from day one.
Business
hmmh and searchHub Partnership Announcement
searchHub partners with hmmh, Germany’s leading connected-commerce agency — a bolt-on for every onsite search across their brands, without vendor changes.
Education
Why Getting Your Query Preprocessing Technique Right Makes Onsite Search Better
Query preprocessing decides what your engine actually sees. Why expansion, stemming and stop-words belong in a separate step outside the search engine.
Education
Wie Produkte gefunden werden
Internetworld Business nennt searchHub als Beispiel für intelligente Erweiterungen der Shopsuche — ein kurzer Kommentar (Deutsch).
Education
Query Understanding: How to Really Understand What Customers Want — Part 1
How searchHub’s Search Interpretation Service tags product types and attributes in real time — and grows its own ontology from anonymous shopping behavior.
Business
valantic and searchHub Partnership Announcement
searchHub partners with valantic, the digital transformation consultancy with over 2,000 in-house experts across SAP, Spryker, Magento, Shopware and Scayle projects.
Education
The Importance of Synonyms in eCommerce Search
Synonyms close the vocabulary gap — and quietly break search through tokenization, transitive compounds and missing context. How searchHub concepts fix all three.
Education
How to DIY Site Search Analytics Using Athena — Part 3
Random sample data across days, CTR and CR queries in jOOQ, two Athena JDBC gotchas, and a line chart to close the series.
Education
How smartQuery Boosts Onsite Search Query Parsing with Querqy
Rules aren’t error-tolerant: one typo, compound or word-order change and the synonym never fires. Let smartQuery normalize first, then let Querqy add knowledge.
Business
Optimise Business Process for Communication
Mycelium networks communicate efficiently at any size. Human teams don’t — efficiency collapses right after one-to-one. What a mushroom teaches about building software.
Education
Benchmark Open Commerce Search Stack with Rally
From a production searches.log to a Rally track in one script. How OCSS benchmarks search, search-while-index and indexing — and what Kibana showed.
Education
From Search Analytics to Search Insights — Part 1
Why most search analytics never turns into action: zero-results without context, and attribution that credits search for things it never did.
Use Cases
How to easily add AI to your Auto-Complete (Suggest) functionality
Four things shoppers expect from suggest, why most implementations miss them, and how STEG Electronics put smartSuggest in front of its own Elasticsearch UI.
Education
Find Your Application Development Bottleneck
Three favourite bottlenecks from a decade of running systems — the kernel’s dirty-page ratio, an Apache still on its default worker limit, and the 23 seconds a JVM lost while its host looked away.
Business
How to Cook Soup in a Team – And Not Spoil It
13% of startups fail over team conflict. A year into searchHub: how a “Daily” hides what people really mean, why the same words taste different to everyone, and what a small team decided to change.
Business
Why Your Source Code Is Less Important Than You Think
Publish the code behind your product and not much happens. Without the team’s mental model, passion and roadmap, source code is just bytes in oblivion — which is why 90% of Open Commerce Search Stack is public.
Education
How to DIY Site Search Analytics Using Athena — Part 2
From human-readable CSV to compressed Parquet, partitioned by day on S3 — an 88% smaller file, and one SQL statement to make Athena notice.
Education
How-To Solr E-Commerce Search
Solr was built for text, not products. Set tie to 0.0, turn similarity scoring off, and run a two-step facet request so Solr only computes the facets your result set actually has.
Education
Simple, Simpler, Perfect – Finding Complexity in Simplicity
A drum teacher’s lesson applied to software — master the complex stuff, then polish the simple core experience. Simplicity is what scales, and it is harder to build than complexity.
Education
How-To Setup Elasticsearch Benchmarking with Rally
Part one of the Rally series — why no guide can tell you the “perfect” cluster, how tracks, challenges and operations fit together, and a first track you can race yourself.
Business
Three Pillars of Search Quality in Ecommerce — Part 3: Sellability
A product is only relevant if it can actually sell right now. How price and availability bias every click metric — and your nDCG — without the query changing at all.
Business
E-Commerce Site Search Overhaul – Super “Selection” Year 2021?
Make or buy a new site search? Before any vendor pitch, settle the corporate-strategy, IT-architecture and team questions — the honest answer may be to keep your platform and fix its weaknesses.
Business
Sustainable Development for Ecommerce Site Search
Cache, not cash. Recomputing the same fuzzy match for “ihpone” billions of times burns CO2 — remember the decision once and reuse it.
Education
Artificial Stupidity – How To Avoid It Before It’s Too Late
Algorithms automate, approximate and interpolate — extrapolation is the dangerous part. Fix #1 for artificial stupidity, after Vincent Warmerdam: if we don’t know, don’t predict.
Education
My Journey Building Elasticsearch for Retail
There is no silver-bullet query. Analyze your data, track regressions, layer analyzers — and relax queries in stages instead of drowning in OR-matches.
Education
Quick-Start with OCSS – Creating a Silver Bullet
There is no silver bullet for Elasticsearch product search — so OCSS ships one pre-built. docker-compose up, index a CSV, and a “preset” of query stages (EAN, article number, a shingled default query, prediction and n-gram fallbacks) you reshape without touching code.
Education
How to DIY Site Search Analytics — Made Easy
No analytics database to run: Athena + Spring Boot + jOOQ, and your first type-safe query against S3 in one sitting.
Education
Three Pillars of Search Quality in Ecommerce — Part 2: Discovery & Inspiration
A search for “jeans” returns 260 products in one flat ranked list. Why diversifying by product type is the oldest trick in retail — and the one search forgot.
Use Cases
Search Event Data Collection — Progress So Far
Two years after open-sourcing the search-collector — close to a billion events, no personal data, what worked (SQS, composable collectors and writers) and what still needs fixing.
Education
The Art of Abstraction – Revisiting Webshop Architecture
Most webshops abstract everything except the interface that changes most — shop to search engine. Arne Heinemeyer on what an abstraction layer buys you: engine autonomy, simple A/B tests, painless migrations, one place to add capabilities.
Education
Monitor Elasticsearch in Kubernetes Using Prometheus
Why to watch Elasticsearch from the outside, how to expose its metrics in the Prometheus format with the prometheus-exporter plugin, a ServiceMonitor for the Prometheus-operator — and the Grafana dashboard that goes with it.
Business
Farewell 2020 – What’s Next in 2021?
Mathias Duda closes the Corona year 2020: ARR growth above 50%, searchInsights and smartSuggest launched, new customers in their own words, a new shareholder, a 180 km hike — and what 2021 holds.
Business
Ain’t No Autoscaling People
December 2020. The restaurant next door lost most of its year while our customers grew. You can scale servers — you cannot “kubectl scale” senior developers or ICU staff, and autoscaling only deserves its name while the SLA holds.
Education
Introducing Open Commerce Search Stack — OCSS
Why open source matters for e-commerce search, where it hurts, and how OCSS puts a slim API layer on top of Elasticsearch so you stop reinventing the wheel.
Business
Use Site Search to Optimize Your Customer Journey
64% of buying moments start with a search, yet 42% of shops neglect site search entirely. The business case for taking it seriously.
Education
Three Pillars of Search Relevancy — Part 1: Findability
Relevance judgments and click metrics both mislead. Findability measures clarity, effort and interaction instead — and gives every query a 0–100% score.
Education
How to Achieve Ecommerce Search Relevance
Relevance is a stack — clean, normalized data first, then weighted fields, ranking rules, tracking-based boosts and query preprocessing on top. Skip the foundation and no amount of machine learning saves the building.
Education
Search Orchestration with Dumb Services
The brain tells the muscles what to do — services that know nothing about their environment, one orchestrator that does, and client-facing nodes that stay up while everything behind them scales or fails on its own.
Education
How to Deploy Elasticsearch in Kubernetes Using the cloud-on-k8s Elasticsearch-Operator
A six-node Elasticsearch cluster — three master, three data nodes over three availability zones — deployed with the Elastic Kubernetes Operator and no Helm: node taints and affinities, the S3 snapshot plugin, and Kibana on top.
Business
Business at the Speed of the Atom
Technology keeps buying us comfort and speed. Jonathan Ross on honest digital development: one litmus test before every launch, why value must flow to the customer, and when to let a feature die.
Business
Hire or Higher to Go Further?
Profit is up. Do you hire new developers or pay the team you have higher salaries? Why the answer is a team decision — and why growth is a social responsibility, not an end in itself.
Business
Why the smartSuggest Module Might Matter to You
Search is a process, not a feature. Why we built our own query-suggestion engine — the goal, what 57,000 search sessions revealed, and the jobs a suggest system has to do: build the corpus, match intent, then scope, filter and rank.
Education
Humans — Search for Things Not for Strings
Walmart’s results for “women red dress size s” vs. what shoppers actually judged relevant: a 100-user experiment on intent vs. strings.
Business
searchHub Dramatically Improves Revenue for E-commerce by up to 39.6%
The first public A/B test, 2018 — two weeks, almost 45,000 visitors, searchHub layered on the customer’s existing search: +39.6% revenue, +25.5% purchases at 99% confidence. How the test was set up and what it showed.
Business
searchHub.io Wins Award for Most Promising Start-Up 2017
London, November 2017: the BCS Information Retrieval Specialist Group named searchHub the most promising start-up of the year — for a system that infuses human understanding into existing search applications.
Use Cases
Why AI Driven Site-Search Is the Key to Success for Ecommerce Marketplaces
Marketplaces multiply the long tail — more products, more ways to ask, more zero-result exits. Andreas Wagner’s 2017 case for AI-driven query clustering that tunes the search you already have instead of replacing it.
Education
The Future of Ecommerce Site Search
The second-oldest searchHub post, 2017: why site search was moving from keyword matching to meaning — natural language processing, self-learning engines, and search that forgives misspellings and synonyms instead of punishing them.
Business
How searchHub.io Changed the Way of Working for Site Search Consultants
The first searchHub post, 2017: why “t-shirt” and “tschirts” return different results, why synonym lists cannot fix spelling errors, and the idea that became searchHub — help the engine instead of replacing it.