A few weeks ago we ran our flagship US conference Haystack, the first real in-person event since April 2019 (I’m sure I don’t need to tell you why!). It’s still an uncertain world for events like this, being hard to tell who will or can travel, who will accept, reject or require particular precautions and whether an online option is now a necessity. We approached the event with a certain amount of nervousness but took the approach to copy our previous 2019 Haystack conference as closely as possible – we’d had excellent feedback from this and had more or less sold it out.
You can see pictures and comment from the events as it happened, tagged #haystackconf on Twitter. We are working on editing the videos of all the talks, you can find slides and links to the videos as they appear here.
First, some MICES
The day before Haystack US 2022, OSC sponsored the first Mix Camp E-Commerce Search (MICES) event in the USA – created by a team including René Kriegler our Director of E-commerce Search, this event has run successfully in Berlin, Germany for several years now. It’s a free day of talks, workshops and ‘unconference’ style discussions decided and run by the participants, and attracts not just developers but product owners and others concerned with creating excellent e-commerce search.
The speakers at MICES US included René himself on a new model for e-commerce search, OSC alumnus Doug Turnbull and his colleague Chen Karako, both now at Shopify who talked about the balancing act that is data and engineering team collaboration. Daniel Wrigley of OSC described how the Chorus open source e-commerce search stack now works with Elasticsearch and independent consultant Johannes Peter talked about how relevance is more than just algorithms and needs to be organisationally scalable. The formal talks finished with Lucian Precup and Radu Pop from Adelean on approaches to autocompletion. After a buffet lunch (which was particularly pleasant eaten on the top floor balcony of our venue the Graduate Hotel, with views over Charlottesville to the mountains) we had a series of self-organised breakout discussion sessions covering topics as diverse as scaling & growing search teams, and vector search.
The first MICES US was definitely a success, although like the first MICES in Berlin it attracted a relatively small crowd of 30-40 people. Videos of the talks will appear on the MICES website soon and we hope to repeat the event next year.
Haystack day 1 – Flying objects
It was so nice to be back in the Violet Crown movie theatre, watching attendees arrive for the conference, catching up after so long and meeting new people! I had a surprise for the crowd in my opening Keynote talk where I revealed my interest in juggling and magic and even managed to persuade an OSC colleague to help with a two-person club passing pattern. My message that like jugglers, those in the search relevance community should focus on openness, sharing and generosity was well received (and at least this time I had actual props rather than the vegetables I had to resort to some years ago).
We really didn’t know what to expect in terms of numbers, so it was gratifying to have 90 people in the conference venue (and all happily complied with our mask mandate and request for proof of vaccination, thankyou all for making this as painless as possible). However what we didn’t expect was another 90 people to join us online in two Zoom rooms where we streamed all the talks – making a total of 180 attendees for this year’s Haystack US, the most ever! People were still registering online at 2pm on the first day. Using Zoom chat they were even able to ask questions in real time of our speakers, via our OSC hosts.
Our speakers came from companies that are old friends of Haystack (LexisNexis and Elsevier), from sectors including online learning (Course Hero) and featured a range of topics including machine learning, image search, Elasticsearch, Vespa and legal information search. Doug Turnbull who was instrumental in creating the Haystack conference while at OSC returned with colleagues from Shopify and spoke about Bayesian optimization of relevance. After lunch we returned to the Search Radar analysis started at last year’s mainly virtual Haystack and Eric Pugh and Jeff Zemerick hosted a fantastic brainstorming session to identify trends in search – watching up to 180 people collaborate in real time on a Miro online whiteboard was awe inspiring, with sticky notes flying everywhere!
After some great Lightning Talks (the only challenge being how to rapidly swap laptops around for projection) we stopped the formal part of the day. Later on, we met up at Kardinal Hall for drinks and a buffer dinner. During the evening I’m pretty sure I walked over a supine attendee while juggling various objects, although my memory is a little hazy – kudos to them for bravery considering how late it was!
Haystack day 2 – Radars & Panels
The day started a little slowly for some but we were soon into the swing of things with presentations from the Adelean team (all the way from Paris), on personalised search, AI-driven search and on whether the wisdom of the crown is indeed as wise as we might hope. Working through the night, Jeff Zemerick had pulled the Search Radar into some kind of order and we broke into groups to try and gauge how critical each topic should be:
Eric and Jeff will be working on the Radar and we hope to publish the final result soon. All in all it was a very interesting exercise and it was great to see the live collaboration happen.
After lunch we had the chance to listen to a panel put together via the #Women-of-Search channel in Relevance Slack – hearing about the experiences of women working in search, both good and bad. This felt necessary and timely and I’m very grateful to host Audrey Loberfeld and all the panellists – this is a group who will go far and we expect it to grow! If you want to join just ask in the Slack.
After a few more talks on test collections, OpenSearch and e-commerce and how to improve data quality for indexing we closed this year’s conference. Judging from the feedback we’ve received our in-person attendees enjoyed the chance to get together again in Charlottesville and our online attendees were happy to be able to participate in some way. Not everyone is willing or able to travel and it feels like Haystack should now be a hybrid event wherever possible – we’ll do our best to make that happen.
Thankyou again to everyone who helped make Haystack US 2022 possible – we’ll be back in Europe for Haystack EU, probably in Berlin in September or October – keep an eye on the Haystack website, follow us on Twitter and join our events mailing list for updates.