Introduction

Our Data Science colleagues had been hosting an internal Data Science Days event for a few years. For our 2,000+ Engineers, we had been missing a similar community event. For several years we wanted to organize one, but got distracted by other priorities and external factors. Finally, in 2022 we decided to commit to hosting an internal Engineering Conference every year and included this commitment in our Engineering Strategy.

Last year, in August 2023, we hosted our first internal Engineering Conference. In this post, we are summarizing how we organized this event and provide tips for those who want to organize a similar event in their company. If you never hosted an event like this before, fear not - when we embarked on the journey we also had no experience in doing so. The event turned out to be a success nonetheless.

Conference format

As this was our first event, we had no reference on the level of interest from potential speakers nor attendees. Without a reference point from prior years, it was a big ask to request that Engineering Managers allow their teams dedicated time to attend, especially given the summer holiday timing (which could work for or against attendance). On top, conference talks are expected to be of higher quality than typical internal presentations, so we needed a format that would ensure quality of talks.

Given these circumstances and following our value think big, act fast, we defined the conference format as follows:

• 1 day event, all online (we're 2,000+ Engineers with sites in Berlin, Dublin, Dortmund, Helsinki, Stockholm, Zürich)
• call for papers to collect submissions across 8 tracks
• track host per track who would moderate the track and act as subject-matter expert during the preparation of the talks
• program committee to review submissions and select talks

Initially, we were thinking that 8 tracks would be too many, but we wanted to encourage submissions across a variety of topics and see where this takes us, adjusting the track as needed. Our tracks covered Building Platforms, Cloud Native, Developer Experience, Data Engineering, and App/Web Development. We also had a dedicated track for Engineering Leadership and (of course) for the hot topic of the year: AI.

The call for papers was open for 3 weeks. Up until the very end, we were not sure if we would get enough submissions to fill all tracks. Only the last two days before the deadline, we received a significant number of submissions. We ended up with enough submissions to fill all defined tracks and struck gold. Now the organizing team had a challenge - to deliver an event with 8 tracks happening in parallel and 54 talks in total.

When we reached out to our broadcasting team who typically assist in hosting internal events, we learned that they never hosted an event that big, with 3 tracks being their technical limit. So we ended up hosting the event on our own, using Google Meet streaming, a slide-based presentation catalogue with talks and descriptions, and 54 calendar events to make it easy to build up one's own schedule.

Conference content

2023 was the year of Large Language Models (LLMs), thus it could not be missing from our event. As LLMs were new for many of our Engineers, we invited our Data Scientists to share their know-how on this topic. We had a talk about the fundamentals of LLMs, followed by a summary on the challenges using LLMs based on two use cases: code generation and building our Zalando Assistant. As expected, these presentations attracted a lot of interest from our community.

Our Engineering Leadership track was focused on talks related to managing teams in challenging times, building trust with the team and sustaining empathy when the team or oneself is affected by the current situation. Other talks focused on driving innovation, continuing to learn as leaders.

The Cloud Native and Developer Experience tracks turned out to be great platforms for sharing new developments in our infrastructure services and promoting their use. Colleagues learned both about proven features that they may be missing out on as well got a peek on improvements in our Kubernetes platform. Our SRE-minded speakers, shared tips about building easy to understand Grafana dashboards using data visualization techniques and demonstrating reference dashboards for applications.

The Data Engineering track was focused on sharing best practices in data processing and data quality. Speakers shared how they monitor data quality in their pipelines, how to simplify data aggregation queries, or how architectural decisions around data design affect data quality and technical debt.

Two teams particulary stood out with multiple presentations across the tracks. The team behind our Web platform shared their journey on evolving their platform into a standalone framework that's now also powers parts of the Zalando Lounge experience, covered the journey to concurrent React, and how we continuously measure and report on the web performance. Our Size & Fit team on the other hand explained how the Size Recommendations based on Body Measurements features work behind the scenes, starting with the on-device computation and ending with the compliance requirements for processing sensitive data. The team also shared how the data acquisition pipelines for the Virtual Fitting Room.

Lessons learned

Conference format

With the 8 tracks in a single day, we triggered massive FOMO (fear of missing out) across Zalandos, as it was difficult to decide which talk to attend. We knew from the get go that this would be a challenge, but decided that the trade-off was worth it. Now that we gained credibility for running the event, in future we will reduce the number of tracks and spread the event out over at least two full conference days. When hosting yearly events, the amount of net new project content is expected to stabilize when compared to the first event.

For first-time speakers, the online format was a great opportunity to practice as stage anxiety is smaller than when speaking to a full room. It's challenging for an online-only event to deliver a full conference feeling, though. While on the following day we had an on-site event with two keynotes and a get together, participants were missing the buzz and networking opportunities known from on-site conferences. Nothing replaces the chatter in the hallway and missing talks due to engaging in conversations with colleagues in a prolonged coffee break ;-)

We had two conference talk formats: full talks (with Q&A) and short lightning talks (without Q&A). The feedback we received from speakers for the lightning talks is that they missed out on the Q&A part and the resulting feedback loop telling them whether the audience was interested in the talk (or not).

CFP

We ran the CFP (call for papers) using Google Forms and scored the submissions in Google Sheets. Each Program Committee member reviewed and scored the submissions based on the topic relevance for the target audience, the abstract quality, and the expected takeaways. We provided a scoring guidance document and removed speaker information to ensure an unbiased selection process focused solely on content. To balance the workload, we assigned each committee member up to 50% of the tracks to score. We then normalized the ranking results and selected the top submissions for the conference. In some cases, we reclassified talks across tracks to ensure balanced content distribution.

Track Hosts

Assigning a track host per track worked well (and is well known from other conferences). The track hosts helped speakers prepare and were an early sounding board for the presentation content. They had freedom to select the order of the talks to ensure a good flow of topics and help in their storytelling when introducing the speakers throughout the day. Hosts also prepared backup questions to use in the Q&A part in case while the audience was busy typing their questions into the Q&A tool.

Summary

The event turned out to be a success and we received a lot of positive feedback from our colleagues who after the closing event were asking when we will host the next one. The event was a great opportunity to learn about projects across the organization and to promote platform solutions to a wide and focused audience. The recordings from the talks serve as onboarding material for colleagues willing to learn about specific projects or just joining the team of the speakers. The on-site event on the following day was a great opportunity to meet colleagues in person and to get their first hand feedback on what they liked from the conference and what they would like to see improved.

Tips for organizing similar events

• Sponsorship: get a sponsor from the leadership team to provide budget and high-level guidance for the event.
• Organizing team: form a small team to organize the event (at Zalando we have a Tech Academy team experienced in organizing events for the Engineering Community).
• Program Committee and Track Hosts are great mechanisms to give visibility to role models and to promote diversity across the organization.
• Program Committee: use a principled-based approach for program committee composition.
• CFP scoring: provide guidance for the program committee on how to score submissions; ensure that the selection is based solely on the content of the submission (via conference software or just plain old spreadsheets).
• CFP scoring: submissions that made it to the shortlist, but did not make it to the conference, should be considered for other internal talks formats or blog posts.
• Track Hosts: consider assigning a track hosts, if only to moderate the track and introduce speakers during the day; they can also help speakers prepare the talks, though you can also assign a group subject matter experts to review the talk early on.
• Communication: meet the target audience where they are; use all possible communication channels to reach them: chats, email, intranet, posters in office, ask leads to promote the event in their team meetings and townhalls.
• Presentations: provide a slide template to ensure a consistent look and feel across all presentations (at least for the first slide). Provide guidance on the font sizes and how to pick accessible color combinations with high contrast.

What changes are we making this year?

This year, we're running the conference already in June and host it as an on-site event with the aim to create a real conference feeling. We spread the conference over two days with three tracks per day, merging some tracks from last year and adding new ones. The event will be streamed to all sites, this time with support of our broadcasting team. The streams will also make it possible for our colleagues to join the event from home.

We ran the CFP for 4 weeks to give potential speakers more time, but the impact on the number of submissions over time was neglible. The due date for submissions is what matters and as in 2023 we received most submissions in the last two days before the end of the CFP. We invited past speakers and track hosts to become part of the Program Committee.

We're excited to host the event again and look forward to learning how the conference format for this year will be received by Zalandos. More on that another time!

We're hiring! Do you like working in an ever evolving organization such as Zalando? Consider joining our teams as a Software Engineer!