At the end of week #470, Iceland had a huge wind storm. It was the first time the Reykjavik area was under a red weather alert. They emphasised that if you didn’t have to travel, you shouldn’t. Everything was alright in the end, but it meant that last Friday’s weeknotes were postponed until now.
Week #470
We contacted lots more people about Material Conference. No matter how much you do, there are still more opportunities to create awareness. We also confirmed another short talk speaker and we’re very excited about it. There are several more we’ve spoken too, but haven’t been able to confirm.
We wrapped-up one data analysis for O’Reilly Media about the state of A.I. research and started in on the next one regarding Infrastructure Operations. Along the way, we are looking at improving the workflow to make it easier for everyone to take-part and look closer at the results in a more real-time.
We also tested out our first Robocall to a large list of survey respondents. The idea is that you have agreed to take part in a survey for one of our customers. First, you get an email with all the information you need and how to login. If you don’t complete that, a few days later we try our second method of contact, SMS. This is useful since people don’t always check their email, give the wrong email address, or messages are routed to spam. So a different method of contact usually gets another bump in respondents. Then, our third method of contact for those who still haven’t responded is another email. Our surveys typically last one month, and by the 3rd reminder we are a little over half-way. Those customers with below acceptable response rates are at risk of not getting and results. This is typically where we get into lots of messy manual work of calling people one-by-one to ask them to respond to bump up the response rate.
This time, before we resorted to manual, human calls, we setup a robocall. We used Twilio with an automated decision tree and hooks back to our server. This allowed us to call all the remaining non-responders with a robocall prompting them to please finish. They could choose to listen in Icelandic or English to the message and request their login code be sent to their email again. We weren’t spamming random people, these are people who agreed to participate and have been contacted three times before. This is hopefully the final push before we needed to get to manual calling.
So far, with a limited test, the results are promising and we’ll continue to explore the possibilities and make improvements over the next few weeks.
Week #471
More Material Conference prepping. It is now a month away, so we contacted the current ticket holders to let them know some basic event details. We continue to round-up some details about the day and had a few video calls with speakers to discuss their topics a bit more.
More survey prepping. March is a very big month for us, so lots todo and prepare to make sure things run as smoothly as possible. We also built a small prototype of a really interesting new survey tool. With several production survey tools under our belt now, this one was very quick to prototype. By re-using a lot of existing pieces, we managed to create something totally new, but it will have to wait. In a few months, once things calm down with the team, we’ll pitch the idea, demo it, and see if it is worth polishing as a potential product.
Wednesdays, we usually work onsite with a customer doing some data mining and analytics work. This week, we took a deep dive, into a tiny slice of users. Based on some work on user retention, we segmented all new sign-ups and how many minutes they used the app on their first day. Then looked at their retention on their second day. Having all the data already structured in a sensible and organised manner made this task pretty easy to execute and the resulting charts were extremely useful. Everyone was happy and we all felt all the hard work in the previous weeks is now paying off in the agility to answer these new types of questions quickly.