Approaching a methodology to select speakers for conferences

Planted April 12, 2018
Pruned June 9, 2025

After a great first edition, this year I organised the second edition of the Droidcon Vietnam with some local folks. Before I organised a conference like this, my experience was limited to local Meetups in Munich (I am currently the organiser of the Kotlin User Group Munich, and the Firebase User Group Munich). The latter has a different nature in terms of resources, logistics and efforts required. They are community-based events, local and — without requiring an easy trajectory — they are certainly less complex than the former.

The Droidcon Vietnam has a different nature. This year we welcome a total of 269 attendees, 17 speakers from 7 different countries, sponsors, etc. Without acquiring the dimensions of other events, it was challenging enough. I was lucky enough to collaborate with some bright and enthusiastic local organisers and collaborators that made it possible. Without them, it would have never happened, and they become a substantial part of the equation.

This year we have improved and increased our awareness in many aspects. The conference has gained in traction, we have gathered statistics and metrics that are enormously helping to attract more speakers, sponsors and attendees. We are growing in almost all the aspects: venue size, attendees, budget, speakers, media interest… and predictably, the complexity and matters that require attention are growing together.

One of those aspects is the selection of the sessions for the conference. Last year we struggled to cover the spots and, contrastingly, this year we have received almost 90 proposals. This requires a proper selection method to ensure that the conference possesses some desired traits:

• A high-quality mark
• Diversity
• Fair opportunities for all the speakers

I started researching on optimal methods for paper selection, since at this stage a fully cherry-picking approach was not feasible, neither appropriate. I posted the following on my Twitter account:

Some of the suggestions were rather controversial, so I decided to post a poll trying to gather some answers:

Thereafter, and in contrast with some articles regarding other topics, I encountered that the ecosystem was rather lacking any direction or information regarding this issue — which is central to any conference. Some individuals speak of prioritising speaker’s reputation, some others of giving importance to content, and is in in general easy to find opinions fitting your own beliefs everywhere. With this article, I am trying to give an answer to this labyrinth and propose a model that focuses on the aforementioned items:

• A high-quality mark
• Diversity
• Fair opportunities for all the speakers

We haven’t fully followed this model for the 2018 edition, but we intend to do it for the upcoming Droidcon Vietnam 2019 — with any modification that can improve it.

High-quality mark

You are attending a conference. Your company sends you there, or you have decided to do it by yourself. You are taking the time to possibly travel to eventually another region, pay ticket and accommodation, and you are entitled to receive a premium experience. You do not travel to just have a “ok” experience, you need to make the best out of it. You are expecting high quality.

Regarding a conference, for me this translates into having sessions of topics that are interesting and relevant enough, that provide you value, and eventually from speakers that have some reputation (yes, in each technology constellation there are certain individuals that have a track of conferences and some fame preceding them). You might be expecting events to network and connect with other peers, but this is out of the scope of this article.

Diversity

An image is worth a thousand words. Take a look at the following Tweet, if you haven’t done so yet:

That is. A hand dryer, that was likely designed by white-male engineers. It was likely tested with white folks. And this lack of diversity in a team leads to a malfunctioning product.

You want to have a diverse team, whether it comes to product design or selections of papers for conferences. A single profile team will ultimately fall into biases. There are many articles already written about this and it is not the purpose of this article to unravel the reasons for diversity, but fundamentally: diversity will respect your audience, will provide a safer environment and a wider range of views.

Fair opportunities for all the speakers

Conferences frequently fall into the ultimate trap of becoming a “family reunion” full of endogamy. The set of speakers attending has met previously in a few other conferences, and everybody seems to be very familiar with each other. This might be common up to a certain point, but if you cross the border into“conference inbreeding” new speakers will never get the opportunity they deserve to get started. And it might lead you as well to the black hole of diversity — a confluent party of people tends to converge in thoughts and ideas. It is fine to do some cherry-picking and invite speakers that are key in their areas, as long as you find a trade-off and avoid exclusivity.

The method proposal

After having introduced a few of the characteristics we want to attain at our selection, I will introduce a particular proposal for paper selection. Please, keep in mind this — and eventually every other methodology — should not be taken as an immutable Holy Book. Read it through, take your notes, have a thing about whether it fits in your particular domain or not, and maybe do some cherry-picking. Take the ideas that result interesting, adapt them, and discard the ones that aren’t.

Forming the committee

Depending on the conference nature, you might want to aim for a different size of the committee. It should not be too small to avoid bias, but neither too big. Is there something such as a magical number?

Probably not, but research shows that agreements are easy to achieve when you are working in teams of up to 5 people. Surpass this amount, and the number of social interactions start to explode, making conflict negotiation hard.

If you work in a SCRUM team and you are asking yourself this question: yes, this is intended for SCRUM as well. Teams of around five people are functional.

Ensure as well that you are fulfilling the requirements of diversity we have previously discussed: ensure that you compose a team with a broad set of ideas, religion, gender, ethnicity and life experience to avoid biases.

How to compose this team? In my experience at the Droidcon Vietnam, after each year you earn traction and build knowledge from previous experiences. If this is the first time you are organising the committee, you might need to adopt the role of “benevolent dictator” and give it a try by yourself. Maybe you can do a sanity-check with individuals you trust, to ensure that you are not falling within your own bias.

Cherry-picking

A conference might want to cherry-pick (i.e., to pick directly a speaker outside of the selection process) for different reasons. I can directly think of speakers that are strong in their area — and will boost the interest in the conference — and speakers due to sponsored talks. This is a common practice at conferences: a sponsor gets offered one slot at the conference to hold a session related to the sponsoring company. Whether this is not intrinsically bad, I would advice a hypothetic conference organiser to ensure that the talk aligns with the interests of the audience. You need to have sponsors to run the event, but it cannot be at the cost of detracting from the general quality of the event. Radical transparency and communication will help you to solve this issue, since it is obviously in the interest of the sponsor to keep the quality of the conference high. At the Droidcon Vietnam we did not have this problem — in fact, the so-called “sponsored sessions” where actually some of the sessions fitting better within the conference scope!

Let’s go back to cherry-picking. There is no absolute number, and it will depend on the size of the conference. Some conferences have more sessions than others, some conferences have a different amount of keynote sessions and many features can vary in nature. I have felt comfortable with a small number of cherry-picks — maybe 2–3 for the sponsored sessions, and maybe 2 for the keynote speakers. You will have to decide on the appropriate number for your event based on the characteristics of your own universe.

Session selection

After doing the cherry-picking, you want to proceed with the actual sessions selections. Here the committee comes into play — they will have to emit a vote based on the decided parameters, and you will take an average from there. Let’s see some parameters:

• Speakers’ reputations
• Session usefulness/interest for the conference
• Traveling support — yes, this is unfortunately a relevant parameter for many conferences

At the Tweet I linked to earlier, most of the folks voting opted for giving more value to the session rather than the speaker. I think this makes sense, since you have already done the cherry-pick and had already the chance to incorporate key speakers.

Price it is unfortunately an issue to several conferences, specially if they are non-profit — like we are. We got fantastic requests from speakers overseas that I would have loved to bring, but their travel expenses would have used a significant chunk of the conference budget. I am working on bringing further sponsors to the conference that can potentially sponsor this expenses, but until it happens, cost must be taken into consideration.

To give an example: each member of the committee could proceed to extend their vote based on the following values:

• Speaker reputation: from 0 to 4
• Session usefulness: from 0 to 7
• Traveling support: from 0 to -3 (the more expensive the traveling support becomes, the more negative points it gets)

You would obviously need to adjust the scoring system to your own conference requirements, but you get the point. I would recommend to use a system with blind votes. This further helps to avoid unconscious bias, this time by committee members seeing what how others have voted before providing their own votes.

Clustering

It can happen that two sessions belonging to the same topic obtain a high result, but you generally want to avoid overlapping topics. How can you solve this?

I think an easy solution is clustering. As an example, at the Droidcon Vietnam we received many topics about Kotlin, and even more specifically about Coroutines. But having a session on coroutines can probably be enough, so how do you do it?

Cluster all the topics. After the selection has been done, group them by topic. And take the selection from there. It will happen that one fantastic speaker gets left out because the topic he/she proposed had many submissions, and another one received a better score. This can be solved by allowing (and encouraging) a speaker to submit several proposals.

Summary

There are many other aspects that can be a subject of discussion and can be controversial (Should the sessions be anonymised? Should there exist the role of a “benevolent dictator” that can interfere in the process at any time?). That would convert this article into a more extended series, which is out of the scope.

I hope I have outlined a few points that could contribute to your conference having a fair and efficient selection of speakers. Although this truth works for me, and I do not believe in any absolute truth and I aim to be a critic of my own thoughts and ideas. If you have any contribution, I would be more than delighted to hear about.

I had discussions on this topic through the glass of a beer or a coffee with several people. I would like to thank Kai Koenig — who has been as well a speaker at the Droidcon Vietnam — as well as Victoria Quirante and Nacho Martín, who are doing a great job at organising the ReactAlicante conference, and have experienced the same dilemma.

I write my thoughts about Software Engineering and life in general in my Twitter account. If you have liked this article or it did help you, feel free to share it, ♥ it and/or leave a comment. This is the currency that fuels amateur writers.