Intercom has this wonderful tradition of creating an amazing personalized illustration of you for each Interversary. They collect interesting facts about you from your team and colleagues, and then a wonderful designer, Sebastian creates something like this!! Beautiful!!
Today I published Intercom’s first ever data report with my colleague Sara Yin where we explored emoji trends in business messaging. We analyzed more than two million anonymized conversations that took place between our customers (i.e. businesses) and their end users (i.e. consumers) during a 3-month period, from June to August in 2015 and 2016. Below is a just a couple of the interesting things we found. You can read the full report here and a fun blog post here.
#1 Emojis improve message engagement
We discovered that messages started by a business that contained an emoji were four times more likely to elicit a response from a consumer than those that didn’t!!
#2 The Top 20 Emoji
The top 20 emoji found in messages in 2016 were quite similar to ones you might use in your personal life. Category-wise, over a half (51%) of the top 20 emoji fall under a facial category, followed by object-based emoji (18%). As you might expect in a business setting, money-related emoji also featured (11%).
When we compare the top 20 emoji used by businesses vs. consumers we find some real differences! Consumers used facial emoji 30% more than businesses did (83% vs. 51%); businesses stuck to objects (18%) and money symbols (11%), neither of which show up in the consumer list.
Predict is an analytics conference that took place in Dublin’s RDS in October 2016. Its mission was to:
mobilise an international community to solve important human challenges through the power of data and predictive analytics.
The theme of this year’s event was the Journey from Data to Predictive Analytics.
It was my first time to attend and present at the conference and I wasn’t disappointed! I gave a talk on what it means to deliver great analytics and shared the 4 key areas that my team and I in Intercom have been focusing on to build a strong analytics team. Specifically:
- Foster an open feedback culture
- Develop close partnerships
- Use a common language
This year’s RecSys was jam packed and fantastic. The dual track 3-day conference took place in beautiful MIT in Boston and showcased lots of great talks around advances in recommendation system algorithms and approaches. Below are some of my highlights from the conference. Full proceedings are available here.
#1 Awesome Industry Track
I co-chaired the industry track along with Paul Lamere, Director of Developer Platform at Echonest / Spotify and Hrishi Aradhye, Engineering Director at Google. This involved selecting and curating a set of industry track talks / speakers from a diverse range of companies who actively work in the recommender systems space. The industry track resulted in a set of 15 talks across 3 sessions featuring speakers from Mendeley, Meetup, Bloomberg, Foursquare, Spotify, Netflix, Pandora, Stitch Fix, Expedia, Nara Logics, GraphSQL, Retail Rocket, Quora, Google and Pinterest. Here’s just some of the awesome industry track talks that were presented.
- Mendeley: Recommendations for Researchers by Saúl Vargas (Mendeley)
- When Recommendation Systems Go Bad by Evan Estola (Meetup)
- Marsbot: Building a Personal Assistant by Max Sklar (Foursquare)
- Music Personalization at Spotify by Vidhya Murali (Spotify)
- Recommending for the World by Justin Basilico & Yves Raimond (Netflix)
- Feature Selection For Human Recommenders by Katherine Livins (Stitch Fix)
#2 Record Attendance at Women’s Lunch!
I co-organized a women’s lunch with Tao Ye, Principal Scientist at Pandora. We had a record number of women attend (almost 50). The lunch resulted in a range of action items for next year’s conference including (hopefully) the organization of day care and a listing of female speakers so that future organizers can choose from a set of talented female researchers and practitioners who want to speak at conferences.
#3 Combining Machine Learning with Human Curation
A prominent theme at this year’s RecSys is the combination of machine learning techniques with large-scale human curation to improve recommendations. Stitch Fix, a clothing delivery service where customers get their own personalized selected of five clothing items called a “Fix”, gave a great talk on this topic. Stitch fix employ over 3000 stylists across the globe! They have a large algorithms team who build recommendation algorithms to help the stylists choose what to send to customers. Customers keep only what they like and send back the rest so it’s important that the recommendations are good! Katherine Levins, a data scientist from Stitch Fix talked about how they approach understanding, measuring and optimizing the role of human selection in a recommendation system. It turns out that they employ a combination of cognitive research, eyetracking, and machine learning models to tune the behavior of stylists. As data related initiatives like merlin take shape in Intercom, it feels like this combination of algorithm and human curation is something we could learn from and try going forward. Katherine’s talk is available here. Earlier this year Katherine published a great related blog post.
#4 Deep Learning
There were lots of talks around deep learning at this year’s RecSys, with several papers accepted accepted to the main conference track and an entire workshop dedicated to deep learning for recommender systems. Paul Convington presented a really interesting talk on deep neural networks for YouTube recommendations. The problem he’s working on is how to predict what movie a user will want to watch next. He discussed how they’ve implemented an age feature designed to remove the bias towards recommending movies / videos from the past. What was most intriguing is that despite the promise of deep learning to advance the field of recommendation, he said that they still have to do lots of feature engineering in YouTube. And overall they have found that user’s interactions with similar items are the best features for improving recommendation.
#5 The Explainability Spectrum & Signal Decay
Shashi Thakur, a Distinguished Engineer and head of the Google Now team gave an interesting keynote on personalization, recommendation and exploration in Google Now. Shashi talked about the importance of setting the right expectations for users and explaining why recommendations are being made at a given point in time. He talked about the explainability spectrum where on one end a recommendation is so clear that it doesn’t require any explanation, and on the other end the recommendation is higher risk / less clear and so needs a clear, concrete explanation. Depending on the circumstances surrounding the recommendation and the person for whom the recommendation is being made, the spectrum of explainability will shift in one direction or the other. I thought this was interesting to consider in the context of merlin and how we surface merlin’s recommendations.
Another interesting point raised across a few talks (including Google and Netflix) related to signal decay. The general advice is to revisit recommendation features often because signals that were once useful may not longer be as useful / as impactful as when the recommender system was first built.
MobileHCI is a conference near and dear to my heart and one I’ve been involved in for almost 10 years. I’ve been publishing and attending the conference yearly since 2007. I’ve also helped organize different tracks over the years — I gave an invited tutorial in 2012 in San Francisco and co-organized the interactive tutorials track in 2014 in Toronto. This year’s conference took place in beautiful Florence, Italy. And I was one of the conference program chairs alongside Prof Antonio Kruger from the German Research Center for AI (DFKI); Prof Jonna Hakkila from the Industrial Design at Faculty of Art and Design, University of Lapland; and Dr Marcos Serrano, from University of Toulouse.
I also took part in an invited panel on the Future of Mobile Interaction, Computing and Life. My fellow panelists included Daniel Ashbrook, Associate Professor in Rochester Institute of Technology; Anind Dey, Director of the HCI Institute in Carnegie Mellon University (CMU); Kori Inkpen, Principal Research at Microsoft; Lucia Terrenghi, UX Researcher and Designer at Google; and Kaisa Väänänen, Professor in the Human-Centered Technology Group at Tampere University of Technology. It was an amazing conference and while there is just too much to mention, what follow’s are just a few highlights from this year’s conference.
#1 All about Emoji!
There were 3 emoji related research papers which shed light on how and why people use emoji in their communications. Super interesting and fun! We’re looking into exploring similar patterns of emoji usage in business communication at Intercom so watch this space!
- Sender-intended functions of emojis in US messaging (Cramer et al.)
- EmojiZoom: emoji entry via large overview maps (Pohl et al.)
- Smiley Face: Why We Use Emoticon Stickers in Mobile Messaging, (Lee et al.) [link to paper for those interested]
#2 The Future of Communication
Adrian David Cheok gave a super opening keynote entitled Everysense Everywhere Human Communication. He talked about new types of communication environments which use all the senses, including touch, taste, and smell, to increase support for multi-person multi-modal interaction and remote presence. Some of his quirky demo’s included:
- A device that attaches to your mobile phone and enables you to feel (and give) a kiss remotely.
- A device that attaches to your mobile phone and emits a smell/scent instead of audio sounds to act as an alternative alarm clock. He demoed an actual use case — Oscar mayer, the bacon company in the US, have an alarm called called “wake up and smell the bacon”!!
- A device that enables taste signals to be transmitted virtually. This prototype “digital taste machine” was featured on BBC One’s Tomorrow’s Food and enables people to taste certain things like sweetness, sourness, etc.
While much of what Adrian presented is pretty out there, it opens up a bunch of questions about the future of personal and digital communication 🙂
#3 Handling Notifications
Notification management was a key theme in many talks. That is, understanding if/how the growing number of mobile notifications impact on people, how people attend to notifications, the cost of interrupting the user and methods for helping them manage inbound notifications they receive on their mobile phones and smartwatches. Research included:
- Novel ambient displays for better handling of notifications on smartwatches
- Use of deep learning to identify important notifications and to use the resulting model to launch associated applications in a timely manner
- A dashboard for enabling people to reflect on the notifications they receive
- A study that explores characteristics of face-to-face conversations like depth/importance/formality as indicators of receptiveness to receiving notifications. The authors find that while engaging in certain types of conversation, in particular small talk, people are more receptive to notifications compared to during other more focused / goal-oriented discussions.
In fact there was an entire workshop dedicated to the topic of notifications and attention management.
In April 2016 we held our first analytics meetup in Intercom, San Francisco and had a fantastic lineup of speakers from Twitch, Wish, Keen IO, Google and Bayes Impact, along with panelists from Social Capital, Mode, Zenefits, ClearSlide and Luminant Data!
We recorded all lightening talks from the evening and recently published those recording online. If you’re interested in anything related to analytics I bet there’s something in this list for you! Happy watching!