How to move a major conference online in five days

The Center for Effective Global Action
CEGA
Published in
14 min readApr 9, 2020

We moved a 300-person conference online in five days because of COVID-19. Here’s what we learned.

CEGA staff facilitate PacDev 2020 from our offices on campus, days before strict social distancing measures were put in place.

The Pacific Conference for Development Economics (PacDev) is the largest annual convening for development economics on the West Coast. As this year’s host, CEGA was getting ready to convene over 200 attendees from the US and beyond in mid-March at the UC Berkeley campus. But as COVID-19 began to spread across the U.S., it became clear that we couldn’t have an in-person gathering without endangering attendees and the Berkeley community.

We considered canceling PacDev altogether, but it felt tragic to throw away months of preparation and disappoint the 70+ speakers who had planned to take part. At the same time, we were daunted by the thought of hosting the conference virtually. This wasn’t a small event — in a single day, we’d have to orchestrate four sets of five concurrent panels, each with three or four speakers, as well as a keynote talk and a poster session. And we had never organized any kind of online event before. Would people even be interested in attending? Would speakers drop out en masse? And if everyone did turn up, how long before the technology failed?

Fast forward to a few days later, as we nervously watched the first speakers log on. We waited for sessions to fall apart, for attendees to lose interest, for the Zoom server to crash. But — nothing. 70 of the 79 original speakers presented, without a memorable technical glitch. Over 130 people watched Gabriel Zucman’s keynote, sharing messages of congratulations and gratitude at being able to participate. And at final count, we had 340 unique attendees over the course of the day — a huge increase over the 200-odd people who had registered earlier. It turns out, when registration is free and you only need to roll out of bed to take part, attending conferences becomes a whole lot easier.

And the kicker: the entire event cost practically nothing: coffee and lunch for the five session moderators, and month-long licenses for Zoom’s webinar add-on (which we were also able to use for another 300+ person conference this week). But our savings pale in comparison to those of the attendees, most of whom would have had to buy flights (some international), take cabs, and rent hotel rooms. While of course we would have preferred to connect with them in person — and not all events are suited to an online format — we thank we managed to make the very best of a tough situation.

Since we learned a lot of this from scratch, we’d like to share our experience so others can build on what we did. Here are some key takeaways, numbers, and a detailed walkthrough:

1. Hosting a virtual event is scary and a lot of work, but ultimately rewarding and fun!

  • More people will attend than you think, and things will (hopefully) go more smoothly than you expect.
  • You’ll lose some benefits of being in the same place (like networking), but you can simulate much of the experience of an in-person event by getting creative, simplifying the event flow, and preparing everyone taking part.

2. Before the conference:

  • Vet, troubleshoot and stress-test your tech, and consider password protecting the event so you don’t get “Zoombombed” (yes, this is a thing now).
  • Write down strict protocols and best practices for moderators, panelists, and attendees — this is new for most people, and the rules of engagement are unfamiliar.
  • Practice to make sure that session moderators know the platform, the protocols, best practices, and fail-safes.
  • Onboard panelists beforehand to make sure they have the tech installed and functioning, and are familiar with the platform, the logistics, and the format of the session.
  • Circulate instructions and best practices to all attendees, and make sure the online agenda has visible and tested links to each session so people can sign in without hunting through their email.

3. On the day of:

  • Use “green rooms” (time buffers or parallel “sign-in” sessions) to ensure panelists are signed in with their tech working well before their panel begins.
  • Read out a structured introduction script that lays out the session format and rules of engagement at the beginning of every session — remember that people will sign in and out throughout the day, including in the middle of sessions.
  • Go all out to keep the audience engaged — remind attendees to introduce themselves when they enter, highlight attendee numbers, give callouts to participants you know, encourage and collect questions throughout (even if q&a is at the end of the session, having questions in hand will save time!
  • Strictly follow the online schedule and be draconian about time-keeping so that people who tune in for a particular session aren’t thrown off.
  • If you’re planning to record sessions (to post videos later, or just for posterity), make sure to let participants know before they sign on.

4. After the conference:

  • Request feedback from panelists and attendees so you can improve the tech and logistics next time around (and share with others who may want to learn from your experience).
  • (Optional) Post slides / videos / summaries of talks so people who couldn’t make a session (or had tech failure) can catch up later.
  • Start preparing for your next virtual event!

At the end of the day, we are incredibly proud of and humbled by participation at PacDev 2020. While there are some obvious drawbacks to virtual conferences, we found that the benefits easily outweighed the costs in this case.

Attendee Brian Daza reflects on the benefits of transitioning PacDev online. (Credit: Brian Daza)

We look forward to strengthening CEGA’s virtual event expertise even after “sheltering in place” is a distant memory. In fact, CEGA will consider hosting virtual events in addition to (or as an alternative to) in-person events in the future, as they allow us to engage with more and different people in new ways. We believe our community wants this, too — in our post-event survey, we learned that nearly 60% of respondents are more likely to participate in virtual conferences, while 24% are equally likely, and just 16% are less likely to participate.

Below, we report the numbers and layout details of our process and some lessons learned, including improvements that were made when we co-hosted another 300-person conference (Measuring Development VI: Data Integration and Data Fusion) two weeks later. We hope that this more detailed “how-to” walkthrough will serve as a useful resource to others planning virtual conferences in the time of coronavirus (and beyond).

BY THE NUMBERS

  • PacDev had 206 registered attendees prior to moving online, but ended up attracting 340 attendees over the course of the day. 138 people attended the keynote, with an average of 32 attendees per session.
  • MeasureDev had 111 registered attendees prior to moving online, but ended up attracting 358 attendees over the course of the day, with as many as 171 attending the keynote and an average of 157 per presentation.
  • For PacDev, 70 speakers presented over the course of the day. 9 speakers cancelled (12% attrition).
  • Zoom measures “attentiveness” by the percent of time attendees have the Zoom window open and active. This was 83% for PacDev and 75% for MeasureDev, showing people were actively engaged.
  • We had expected to spend $5,500 on PacDev, but spent only $1,540 on the online event — most of which was the cost of 5 webinar licenses. We were able to use the same licenses for MeasureDev 2 days later, so the cost was effectively zero. (Note that we had expected to recover costs through registration fees, but since this was a new format we decided to refund all registrants).

WALKTHROUGH: BEFORE THE CONFERENCE

Choosing a platform

We looked at both Zoom and Google Hangouts Meet, and decided to use Zoom because it is becoming the industry standard and has explicit “webinar” functionality, which mimics many features of large conferences including explicit roles (with varying levels of permission) for moderators, panelists, and attendees (the audience).

Notably, the webinar severely restricts audience participation, in ways we didn’t always like — for example, audience members cannot transmit their own videos, and can only speak when unmuted by the host, which makes it harder to emulate the active back-and-forth and feeling of collective participation that comes from a conference. In balance, we felt safer doing this rather than hosting a regular Zoom meeting because we thought it would (i) reduce bandwidth requirements, as we weren’t sure how the technology would hold up to a large number of video participants (ii) make it easier to manage a large audience — for example, attendees need moderator permission to unmute, making audio disruptions less likely, and (iii) make us less vulnerable to attackers, since attendees have far less ability to affect the session. That said, we think a regular Zoom meeting would work if session sizes are small.

UC Berkeley provided us free Zoom Pro accounts, and we bought the Webinar Add-on for five accounts, which allowed us to host up to 100 people for concurrent panels and 500 people for the keynote.

Takeaways:

  • The Zoom webinar is a lower risk choice than a regular Zoom meeting, but it loses in terms of audience participation, needing the chair/moderators to put in extra effort keeping the audience engaged. For smaller events a regular meeting might work fine, but we haven’t tried it out.
  • The webinar add-on costs extra. We suggest checking with your institution to see if you have (or can get) discounted access, or to other video conferencing software that provides similar functionality.

Creating sessions

For PacDev, we needed to broadcast five concurrent panels (A — E), each with three to four speakers, in four consecutive blocks over the day. For each panel, we built in buffer time to give panelists time to sign on, familiarize themselves with the session format and Zoom interface, and make sure their A/V was working. To avoid having to remove the audience during the panel switch, we created a separate webinar (with a different corresponding Zoom link) for each session. We put all webinar links up on the PacDev agenda (this was in the pre-Zoombomber era — we advise password protecting either your agenda page or the links).

MeasureDev was one long plenary with no parallel sessions, so we decided to go another way: we used a single webinar for all the talks (the “conference hall”), and ran a second webinar in parallel (the “green room”), with its own moderator, which we used to onboard participants 20 minutes before their talks. The single sign-in link made it easier to communicate with participants, and also meant we didn’t need buffer time between talks — at the end of a session, the moderator would simply demote the previous panelists to attendees, and promote the new panelists once they signed back in from the practice room.

Takeaways:

  • If you’re running concurrent sessions, consider using a separate seminar per session, and build in buffer time between talks.
  • If you have a single plenary format, you don’t need buffers, but make sure you have a practice room to onboard panelists.

Training moderators

Moderators spent a few hours the week before each conference collectively familiarizing themselves with Zoom and learning by doing. We spent some time exploring its many presets — for example, green rooms for panelists, enabling participants to sign in using the web client, password protection, etc. We also set protocol and scripts for getting consent from presenters, recording, and sharing sessions; opening/closing each session, keeping time, and moderating chats and questions; and technical preferences like splitting screens between presenter and slide mode.

For PacDev, we had five moderators, one per parallel stream. Moderators were meant to be responsible for everything to do with their session, including: (i) setting all the webinar presets, (ii) communicating with panelists, including sending them each a unique link to allow them to enter their own session with panelist privileges (iii) onboarding the panelists in the green rooms (iv) troubleshooting the tech (v) chairing the session (vi) managing q&a and (vii) keeping time.

For MeasureDev, we split up these responsibilities across a moderator (i — iv), a chair (v — vi) who had co-host privileges, allowing them to mute/unmute participants or remove them from the room, and a timekeeper (vii), which dramatically reduced everyone’s stress levels during the conference.

Takeaways:

  • Make sure moderators know the technology and logistics inside out, and have scripts and protocols clearly laid out in front of them so they’re ready for anything.
  • Appoint a separate chair to manage the panelists and q&a, and make sure the panel doesn’t overrun time.

Preparing panelists

Each moderator was responsible for sending instructions to presenters in their sessions two days before the conference. Instructions included details on how to install Zoom and guidance around the format of each session (length, Q&A, etc). We also asked for a final confirmation from presenters that they will be able to attend & present remotely.

Speaker Miranda Lambert prepares for a day of virtual presentations. (Credit: Miranda Lambert)

We also hosted optional Zoom training sessions for panelists the day before the conference, and more than one-third of panelists opted in for each conference (27 of PacDev’s 70 panelists, and 7 of MeasureDev’s 18 panelists, respectively). People installed Zoom, created profiles, tested their A/V, learned how to share slides and so on. This saved us a lot of time and hassle when preparing panelists on the day of!

Takeaways:

  • Invest time in training panelists before the session to minimize tech-related problems

WALKTHROUGH: THE DAY OF THE CONFERENCE

AV check and troubleshooting

For PacDev, we asked panelists to login to their session’s webinar 30 minutes early. There, each moderator checked each panelist’s audio and video, familiarized themselves with the chat function and the session format (if they hadn’t attended the training sessions). They also practiced sharing their slides on the screen by “taking over” the screen sharing.

For MeasureDev, all this happened in the “green room” webinar. After all panelists were ready, they signed into the “conference hall”. At the end of each panel, moderators demoted the previous panelists to attendee privileges, and promoted the new panelists.

During dead time between sessions, we put up a slide that gave details of the composition and start time of the next panel.

Takeaways:

  • Have all panelists keyed up and ready to speak before the panel begins.
  • Put up a “next panel” slide during buffer time for the benefit of new attendees.

Moderation

Moderators read out a prepared script that laid out the session format and the ways the audience could participate (texting in questions throughout and after the session, “raising hands”, protocols for muting and unmuting). encouraged attendees to introduce themselves via chat at the beginning of each session to make the sessions more personable and transparent, which mostly worked well. However, attendees who joined the session late couldn’t see the chat history.

Q&A was better when moderators strongly and frequently encouraged attendees to ask questions, and when there were attendees in the audience who knew the speakers fairly well. We also encouraged fellow panelists to lead the questions since they could do so over audio. This also gave the moderators/chair the opportunity to collect questions from the audience over chat.

Takeaways:

  • Lay out everything the participants need to know at the beginning of the session, and continuously encourage attendees to introduce themselves and ask questions.
  • Regularly read out attendee numbers and mention any “big names” signed on, to keep attendees interested and involved.
  • Stick to timekeeping strictly to prevent session overruns.
CEGA staff moderating panels throughout the day.

A Simple Q&A Format

After every individual speaker’s presentation, the moderator would read audience questions from Zoom’s Q&A feature as well as the chat. Worried about audience participation, early on we invited panelists to ask their questions first, since they could do so over audio. However, this alienated the audience members even further, since panelist questions took up most of the time allotted to Q&A.

We then tried reversing the order, by asking the audience for questions and reverting to the panelists if no audience members had questions (and if neither had questions, the moderator asked a question), which worked better. As moderators grew more comfortable, some chose to un-mute audience members so they could ask their questions out loud, making for a more personalized experience. This required messaging audience members beforehand to make sure this was okay, and worked well most of the time.

Q&A worked best when moderators strongly and frequently encouraged attendees to ask questions, and when there were attendees in the audience who knew the speakers fairly well. In sessions where speakers couldn’t answer all the questions they received, we shared the chat transcript with speakers and encouraged them to follow up with the attendees.

At MeasureDev, we had an issue with an attendee that came to disrupt the Q&A chat, but we caught it quickly and removed him. We were prepared to disable the chat if it happened again but thankfully, it didn’t.

Takeaways:

  • Agree on a simple Q&A format and stick to it to avoid confusion. Panelists who cannot answer all their questions may want to write their responses into the chat box (using the “reply to ‘all panelists and attendees’” option) during the next talk, so everyone gets their questions answered without disrupting the schedule.

How we made a virtual “poster session”

Originally, 5–10 PacDev attendees had been invited to present posters during lunch hour. When we moved online, we added an extra session before the keynote where each poster presenter gave 5-minute lightning talks and shared their screen to project their poster. This didn’t work too well since posters are too large to display well on regular screens.

For MeasureDev, we had 7-minute lightning talks where everyone had slides, followed by a 3 minute Q&A, which worked much better.

Takeaways:

  • Virtual poster sessions don’t work, consider lightning talks instead.

The Keynote

Research Director Bilal Siddiqi moderates the keynote.

In PacDev, lightning talks led directly into the keynote, which caused some confusion for the previous presenters, some of whom rejoined the webinar as a panelist during the keynote by accident. We fixed this for MeasureDev by having speakers enter with the universal attendee link and be manually promoted to panelist and demoted back to attendee by the host — this turned out to be a more stable system.

For PacDev, we reframed the keynote as “Gabriel Zucman in conversation with Ted Miguel” to make the presentation and Q&A feel dynamic and engaging, given the limited ability of audience members to engage directly.

Takeaways:

  • Consider a discussant or a conversation format for the keynote.

Feedback and networking

One of the most important drawbacks to a virtual conference — according to post-conference feedback — was that attendees didn’t get a chance to network and receive informal feedback on their work in-between sessions.

While creating a substitute for in-person networking is virtually impossible (no pun intended!), there are a few approaches we will consider for next time to facilitate relationship-building: (a) set up a Slack chat room (or equivalent Zoom breakout rooms) for attendees to spend time in and engage with each other during or between sessions; (b) hold “speaker office hours,” using an app like Slack, where they will be online after the session to answer questions and online chat with attendees; (c) consider pre-event “match-making” using an app (like Grip or Whova) or surveys to facilitate conversation, networking and feedback among participants with similar interests and/or between panelists and attendees.

Takeaways:

  • Try out other apps and approaches to help participants network and provide each other feedback.

--

--

The Center for Effective Global Action
CEGA
Editor for

CEGA is a hub for research on global development, innovating for positive social change.