Two weeks ago I ran my first Bootstrapping Machine Learning workshop. I thought it would be a great way to get some feedback on the content of the book I am writing on that topic, and on what people are most interested in learning about. I did this workshop for free because it was my first, but I do intend to use workshops as an additional source of revenue in the future.

I had nothing prepared when I sent the first email announcing the workshop. It was a bit scary, but because I work on my own I sometimes need ways to kick myself in the butt. Setting a deadline and involving people from outside implies accountability. I decided to try GoToTraining for running this first workshop. It was quite easy to set up and I started having people sign up very quickly. I could have GTT automatically send reminders and a follow-up email to participants. It can also handle payments, but I didn't explore this feature. I limited the workshop to 10 participants; one had to cancel, another one got mixed up with time zones, so in the end we were 8. There were people from all around the world, from California to India, so the participants were in all sorts of timezones, but starting at 8am PST seemed to work for everyone.

There were four main parts: a presentation, questions on the presentation, a demo, and questions on the demo. I also did a recap before the demo and we had two 15-minute breaks. The presentation was a summary of the book, I had planned 45' for it but it lasted 1 hour 15'. All questions combined took about 45', and the demo was 45' long as well. All in all, the workshop lasted 3h13min. In the end, it felt quite short to cover all that I wanted. I was probably too ambitious with this, and as a consequence I didn't plan enough breaks and repetition. Friends I've talked to afterwards told me it's best to have breaks every 20 minutes so that people's attention is always at its highest, and I know from experience that it's best to repeat the most important points 3 times. For this type of workshop, this could be once in the presentation, once in the demo, and once in a final recap for example.

What I should have done differently

The first thing that struck me while doing the workshop was this weird feeling of "talking into a vacuum", as soon as I'd switched off everyone's microphones to remove background noise. I learnt shortly after the workshop that Brennan Dunn had switched from GoToWebinar to Google Hangouts because of this. Having everyone in the Hangout be able to turn on their webcam and to see each other prompts for having participants introduce themselves at the beginning, which prevents the workshop from becoming impersonal and which I think is just much nicer. Also, having people talk about themselves is a great way to figure out what they know already, what they want to learn, and how they can get value from the workshop. Besides, seeing everyone enables you to check that they are still following and to monitor their attention.

Another thing I quickly realized while going through the presentation was how inconvenient it is to share your screen and to run the presentation full-screen. I wanted to try using Evernote in presentation mode, where you can just scroll through the content. But then, as it was full-screen, I couldn't see the GTT interface anymore, which meant I couldn’t see if people were writing in the group chat or were raising their hands (one of GoToTraining's features), and I couldn't see the timer, so I had to check periodically and switch from Evernote to GTT and back. For the next workshop I need to figure out another way to show the presentation material (back to regular slides?) or I need to set up two screens (but that's impractical if I'm not doing the workshop from home).

One day after the workshop, I sent a follow-up email to participants to ask them for feedback. GTT can send follow-up emails automatically, but if the recipients use Gmail's Priority Inbox the email won't go to their Primary Inbox, so I find it better to write personally. It turns out from people's feedback that the demo is what they liked most about the workshop, but it was also the only area where they suggested improvements: more details on the code, more hands-on work... I guess I should have spent more time on the demo than on the presentation. There's probably things I told in the presentation which I could have introduced and illustrated in the demo instead, so that it would have felt less like theory and more like applied stuff.

In my follow-up email I was also supposed to give a link to download the video recording of the workshop, which adds value to the offering. However, GoToTraining is just not practical for recording as it happens on your computer whereas with Google Hangouts it happens in the cloud. I've had to leave my computer "convert" the video at night (whatever that conversion was), and now I still need to upload it (1.1GB), but I have limited internet connectivity these days...

Ideas for next time

The pricing I currently have in mind for next workshops is relatively high ($400 for half a day), which is why it is super important that they deliver a lot of value to the participants. Workshops go one step further than the book in the way that people can ask me questions and I can show a live demo of what I am talking about in the book. For even more personalized training, I am also considering in-person training over 3-4 days, and then there's consulting. Obviously, the more personalized it gets, the more it costs, and so I expect that these options will only be of interest to a very small portion of my audience.

The demo / hands-on section is what requires the most work when preparing a workshop, but this work really pays off as the demo is what delivers the most value to the participants. There are always ways to improve a demo by making it more interactive (based on people's own questions or use cases) and by practicing to make it smoother, so most of my efforts when preparing a new workshop will go towards this.

As you may have guessed, I am tempted to use Google Hangouts next time. I will also try with a group of 15 people and see if things are still manageable (I want the workshop to be interactive). Regarding the content of the workshop, I need to either restrict the scope or run a 1-day workshop (but that would be priced higher). I am already planning a half-day workshop dedicated to recommender systems, in mid-April — let me know if you're interested. One thing that's also got me thinking is this huge change that Brennan Dunn did to the structure of his workshops:

"I've ditched the slides in favor of discussion (which is still guided by all the material that was on the slides.) And instead of dedicating time for Q&A (which used to be at the end of the day), we regularly break into discussions and figure out how each attendee can apply what we’re covering to their business starting next week. The focus is now on the application of content, and not the instruction of content."