csv,conf is a community conference that brings diverse groups together to discuss data topics, and features stories about data sharing and data analysis from science, journalism, government, and open source. Over the years we have had over a hundred different talks from a huge range of speakers, most of which you can still watch back on our YouTube Channel.
csv,conf,v1 took place in Berlin in 2014 and we were there again v2 in 2016 before we moved across the Atlantic for v3 and v4 which were held in Portland, Oregon in the United States in 2017 and 2019. For csv,conf,v5 we were looking forward to our first conference in Washington DC, but unfortunately, like many other in-person events, this was not going to be possible in 2020.
People have asked us about our experience moving from a planned in person event to one online, in a very short space of time, so we are sharing our story with the hope that it will be helpful to others, as we move into a world where online events and conferences are going to be more prevalent than ever.
The decision to take the conference online was not an easy one. Until quite late on, the question csvconf organizers kept asking each other was not “how will we run the conference virtually?” but ‘will we need to cancel?’. As the pandemic intensified, this decision was taken out of our hands and it became quickly clear that cancelling our event in Washington DC was not only the responsible thing to do, but the only thing we could do.
Once it was clear that we would not hold an in-person event, we deliberated on whether we would hold an online event, postpone, or cancel.
One of our main concerns was whether we would be able to encapsulate everything good about csv,conf in a virtual setting – the warmth you feel when you walk into the room, the interesting side conversations, and the feeling of being reunited with old friends, and naturally meeting new ones were things that we didn’t know whether we could pull off. And if we couldn’t, did we want to do this at all?
We were worried about keeping a commitment to speakers who had made a commitment themselves. But at the same time we were worried speakers may not be interested in delivering something virtually, or that it would not have the same appeal. It was important to us that there was value to the speakers, and at the start of this process we were committed to making this happen.
Many of us have experience running events both in person and online, but this was bigger. We had some great advice and drew heavily on the experience of others in similar positions to us. But it still felt like this was different. We were starting from scratch and for all of our preparation, right up to the moment we pressed ‘go live’ inside Crowdcast, we did not know whether it was going to work.
But what we found was that hard work, lots of planning and support of the community made it work. There were so many great things about the format that surprised and delighted us. We now find ourselves asking whether an online format is in fact a better fit for our community, and exploring what a hybrid conference might look like in the future.
There were a great many reasons to embrace a virtual conference. Once we made the decision and started to plan, this became ever clearer. Not least was the fact that an online conference would give many more people the opportunity to attend. We work hard every year to reduce the barriers to attendance where possible, and we’re grateful to our supporters here, but our ability to support conference speakers is limited, and it is also probably the biggest cost from year to year. We are conscious that barriers to entry still apply to a virtual conference, but they are different, and it is clear that for csv,conf,v5 more people who wanted to join could be part of it. Csv,conf is normally attended by around 250 people. The in-person conferences usually fill up, with just a few attendees under capacity. It feels the right size for our community. But this year we had over 1,000 registrations. More new people could attend, and there were also more returning faces.
Attendees from around the world during csv,conf,v5's opening session
Despite the obvious differences, much about organising a conference remains the same whether virtual or not. Indeed, by the time we made the shift to an online conference, much of this work had been done.
From about September the organising team met up regularly every few weeks on a virtual call. We reviewed our list of things and assigned actions. We used a private channel on Slack for core organizers to keep updated during the week.
We had a good mix of skills and interests on the organizing team, from community wranglers, to writers and social media aces.
We would like to give a shout out to the team of local volunteers we had on board to help with DC-specific things. In the end this knowledge just wasn't needed for the virtual conf.
We recruited a group of people from the organising team to act as the programme committee. This group would be responsible for running the call for proposals (CFP) and selecting the talks.
We relied on our committed team of organisers for the conference and we found it helpful to have very clear roles/responsibilities to help manage the different aspects of the ‘live’ conference. We had a host who introduced speakers, a Q&A/chat monitor, a technical helper, and a Safety Officer/Code of Conduct enforcer at all times. It was also helpful to have “floaters” who were unassigned to a specific task, but could help with urgent needs.
We were keen on making it easy for people to complete the call for proposals. We set up a Google form and asked just a few simple questions.
All talks were independently reviewed and scored by members of the committee and we had a final meeting to review our scores and come up with a final list. We were true to the scoring system, but there were other things to consider. Some speakers had submitted several talks and we had decided that even if several talks by the same person scored highly, only one could go into the final schedule. We value diversity of speakers, and reached out to diverse communities to advertise the call for proposals and also considered diversity when selecting talks. Also, where talks were scoring equally, we wanted to ensure we we’re giving priority to speakers who were new to the conference.
We asked all speakers to post their slides onto the csvconf Zenodo repository. This was really nice to have because attendees asked multiple times for links to slides, so we could send them to the Zenodo collection.
Though it proved to not be relevant for 2020s virtual event, but it’s worth mentioning that the process of granting travel or accommodation support to speakers was entirely separate from the selection criteria. This was an entirely separate process, and although we asked people to flag a request for support, this did not factor into the decision making process.
Before we could decide on a schedule, we needed to decide on the hours and timezones we would hold the conference. Csvconf is usually a 2-day event with 3 concurrently run sessions, and we eventually decided to have the virtual event remain 2 days, but have one main talk session with limited concurrent talks. Since the in-person conference was supposed to occur in Washington, DC, many of our speakers were people in US timezones, so we focused on timezones that would work best for those speakers. We also wanted to ensure that our conference organizers would be awake during the conference. We started at 10am Eastern, which was very early for West Coast (7am) and late afternoon for non-US attendees (3pm UK; 5pm Eastern Europe; ). We decided on 7 hours of programming, meaning the conference ended in late afternoon for US attendees and late evening for Europe. Unfortunately, these timezones did not work for everyone (notably the Asia-Pacific region), and we recommend that you pick timezones that work for your speakers and your conference organizers whilst stretching things as far as possible if equal accessibility is important to you. We also found it was important to clearly list the conference times in multiple timezones on our schedule so that it was easier for attendees to know what time the talks were happening.
screengrab above shows the start of our deliberation on what time to start and end the conference each day. See our options across timezones
Although most of what makes csv,conf successful is human passion and attention (and time!), we also found that the costs involved in running a virtual conference are minimal. Except for some extra costs for upgrading our communication platforms, and making funds available to support speakers in getting online, running the conference remotely saved us several thousand dollars.
We have always used an honor system for ticket pricing. We ask people pay what they can afford, with some suggested amounts depending on the attendees situation. But we needed to make some subtle changes for the online event, as it was a different proposition. We first made it clear that tickets were free, and refunded those who had already purchased tickets.
Eventbrite is the platform we have always used for registering attendees for the conference, and it does the job. It’s easy to use and straightforward. We kept it running this year for consistency and to ensure we’re keeping our data organised, even though it involved importing the data into another platform.
We were able to make the conference donation based thanks to the support of the Sloan Foundation, Moore Foundation and individual contributors and donations. Perhaps because the overall registrations also went up, we found that the donations also went up. In future, and with more planning and promotion, it would be feasible to consider a virtual event of the scale of csv,conf funded entirely by contributions from the community it serves.
We spent significant time enhancing our Code of Conduct for the virtual conference. We took in feedback from last year’s conference and reviewed other organizations’ Code of Conducts. The main changes were to consider how a Code of Conduct needed to relate to the specifics of something happening online. We also wanted to create more transparency in the enforcement and decision-making processes.
One new aspect was the ability to report incidents via Slack. We designated two event organizers as “Safety Officers”, and they were responsible for responding to any incident reports and were available for direct messaging via Slack (see the Code of Conduct for full details). We also provided a neutral party to receive incident reports if there were any conflicts of interest.
We used Slack for communications during the conference, and received positive feedback about this choice. We added everyone that registered to the Slack channel to ensure that everyone would receive important messages.
We had a Slack session-bot that would announce the beginning of each session with the link to the session and we received a lot of positive feedback about the session-bot. For people not on Slack, we also had the schedule in a Google spreadsheet and on the website, and everyone that registered with an email received the talk links via email too. For the session-bot, we used the Google Calendar for Team Events app on Slack.
Another popular Slack channel that was created for this conference was a dedicated Q&A channel allowing speakers to interact with session attendees, providing more context around their talks, linking to resources, and chatting about possible collaborations. At the end of each talk, one organizer would copy all of the questions and post them into this Q&A channel so that the conversations could continue. We received a lot of positive feedback about this and it was pleasing to see the conversations continue.
We also had a dedicated speakers channel, where speakers could ask questions and offer mutual support and encouragement both before and during the event.
Another important channel was a backchannel for organisers, which we used mainly to coordinate and cheer each other on during the conf. We also used this to ask for technical help behind the scenes to ensure everything ran as smoothly as possible.
After talks, one organizer would use Slack Private Messaging to collate and send positive feedback for speakers, as articulated by attendees during the session. This was absolutely worth it and we were really pleased to see the effort was appreciated.
Slack is of course free, but its premium service does offer upgrades for charities and we were lucky enough to make use of this. The application process is very easy, and takes less that 10 mins so this is worth considering.
We made good use of Twitter throughout the conference and there were active #commallama and #csvconf hashtags going throughout the event. The organisers had joint responsibility for this and this seemed to work. We announced the hashtags at the beginning of the day and people picked them up easily. We had a philosophy of 'over-communicating' - offering updates as soon as we had them, and candidly. We used it to to share updates, calls-to-action, and to amplify people's thoughts, questions and feedback
One of the biggest decisions we had to make was picking a video conferencing platform for the conference. We originally considered using Zoom, but were concerned about a few things. The first was reports of rampant “zoombombing”, where trolls join Zoom meetings with the intent to disrupt the meeting. The second concern was that we are a small team of organisers and there would be great overhead in moderating a Zoom room with hundreds of attendees - muting, unmuting, etc. We also worried that a giant Zoom room would feel very impersonal. Many of us now spend what is probably an unnecessary amount of our daily lives on Zoom and we also felt that stepping away from this would help mark the occasion as something special, so we made the decision to move away from Zoom and we looked to options that we’re more of a broadcast tool than meeting tool.
We saw another virtual conference that used Crowdcast (https://neuromatch.io/) and were impressed with how it felt to participate, so we started to investigate it as a platform before enthusiastically committing to it, with some reservations.
The best parts of Crowdcast to us were the friendly user interface, which includes a speaker video screen, a dedicated chat section with a prompt bar reading “say something nice”, and a separate box for questions. It felt really intuitive and the features were considered, useful, and we incorporated most of them.
From the speaker, participant, and host side, the experience felt good and appropriate. The consideration on the different user types was clear in the design and appreciated. One great function was that of a green room, which is akin to a speakers' couch at the backstage of an in-person conference, helping to calm speakers' nerves, check their audio and visual settings, discuss cues, etc. before stepping out onto the stage.
Another benefit of Crowdcast is that the talks are immediately available for re-viewing, complete with chat messages for people to revisit after the conference. This was great as it allowed people to catch up in almost real time and so catch up quickly if they missed something on the day and feel part of the conference discussions as the developed. We also released all talk videos on YouTube and Tweeted the links to each talk.
But Crowdcast was not without its limitations. Everything went very well, and the following issues were not deal breakers, but acknowledging them can help future organisers plan and manage expectations.
Top of the list of concerns was our complete inexperienced with it, and the likely inexperience of our speakers. To ensure that our speakers were comfortable using Crowdcast, we held many practice sessions with speakers before the conference, and also had an attendee AMA before the conference to get attendees acquainted with the platform. These sessions were vital for us to practice all together and this time and effort absolutely paid off! If there is one piece of advice you should take away from reading this guide it is this: practice practice practice, and give others the opportunity and space to practice as well.
One challenge we faced was hosting - only one account has host privileges, but we learned that many people can log into that account at the same time to share host privileges. Hosts can allow other people to share their screen and unmute, and they can also elevate questions from the chat to the questions box. They can also kick people out if they are being disruptive (which didn’t happen for us, but we wanted to be prepared). This felt a bit weird, honestly, and we had to be careful to be aware of the power we had when in the hosts position. Weird, but also incredibly useful and a key control feature which was essential for an event run by a group rather than an individual.
In Crowdcast you can only share 4 screens at a time (so that would be 2 people sharing 2 screens). Our usual set up was a host, with one speaker sharing their screen at a time. We could add a Speaker for the talks that only had a single other speaker but any more that this we would have had problems.
It was easy enough for the host to chop and change who is on screen at any time, and there’s no limit on the total number of speakers in a session. So there is some flexibility, and ultimately, we were OK. But this should be a big consideration if you are running an event with different forms of presentation.
Crowdcast was also not without its technical hiccups and frustrations. Speakers sometimes fell off the call, or had mysterious problems sharing their screens. We received multiple comments/questions on the day about the video lagging/buffering. We often had to resort to the ol’ refresh refresh refresh approach which, to be fair, mostly worked. And on the few occasions we were stumped, there’s quite a lot of support available online and directly from Crowdcast. But honestly, there were very few technical issues for a 2-day online conference.
Some attendees wanted info on the speakers (ex: name, twitter handle) during the presentation and we agree it would have been a nice touch to have a button or link in Crowdcast. There is the “call to action” feature, but we were using that to link to the code of conduct.
Crowdcast was new to us, and new to many people in the conference community. As well as these practices we found it helpful to set up an FAQ page with content about how to use Crowdcast and what to expect from an online conference in general. Overall, it was a good decision and a platform we would recommend for consideration.
Finally, it would not be csv,conf if it had not been for the #commallama. The comma llama first joined us for csv,conf,v3 in Portland and joined us again for csv,conf,v4. The experience of being around a llama is both relaxing and energising at the same time, and a good way to get people mixing. Taking the llama online was something we had to do and we were very pleased with how it worked. It was amazing to see how much joy people go out of the experience and also interesting to notice how well people naturally adapted to the online environment. People naturally organised into a virtual queue and took turns coming on to the screen to screengrab a selfie. Thanks to our friends at Mtn Peaks Therapy Llamas & Alpacas for being so accommodating and helping us to make this possible.
As we reflect on the experience this year, one things is very clear to us: The conference was only possible because of the community to speak, attend and supported us. It was a success because the community showed up, was kind, welcoming and extremely generous with their knowledge, ideas and time. The local people in DC who stepped up to offer knowledge and support on the ground in DC was a great example of this and we are incredibly grateful or the support, though this turned out not to be needed.
We were lucky to have a community of developers, journalists, scientists and civic activists who intrinsically know how to interact and support one another online, and who adapted to the realities of an online conference well. From the moment speakers attended our practice sessions on the platform and started to support one another, we knew that things we’re going to work out. We knew things would not all run to plan, but we trusted that the community would be understanding and actively support us in solving problems. It’s something we are grateful for.
We were also thankful to Alfred P. SLOAN Foundation, Moore Foundation and our 100 + individual supporters for making the decision to support us financially. It is worth noting that none of this would have been possible without our planned venue, hotel and catering contracts being very understanding in letting us void our contracts without any penalties.
Many people have been asking us about the future of csv,conf. Firstly it’s clear that the csv,conf,v5 has given us renewed love for the conference and made it abundantly clear to us of the need for a conference like this in the world. It’s also probably the case that the momentum generated by running the conference this year will secure enthusiasm amongst organisers for putting something together next year.
So the questions will be ‘what should a future csvconf look like?’. We will certainly be considering our experience of running this years event online. It was such a success that there is an argument for keeping it online going forward, or putting together something of a hybrid. Time will tell.
We hope that this has been useful for others. If you are organising an event and have suggestions or further questions that could improve this resource, please let us know. Our Slack remains open and is the best place to get in touch with us.