Our Summit Speaker has over 15 years of progressive experience in data driven analytics, and is currently managing data science and marketing analytics at Uber.
What gets you up and going in the morning?
Some days are easier than others. But for the difficult ones, I think of one good thing I can do to help someone (mostly at work) – that action is good against anxiety.
What do you like doing in your spare time?
I play the Clarinet and it gives me lots of joy doing it.
You have over 15 years of progressive experience in data driven analytics. How do you think things have changed over the years?
The core math behind analytics has not changed substantially in the past four decades. However, there are meaningful breakthroughs in technology and algorithms that have made previously intractable analysis possible.
For example, in Machine Learning, we have very efficient approximations that get to classification accuracies in the 99% range or beyond (think gradient descent). In AI, modern neural architectures, like CNN/RNN and adversarial nets, are producing results at par with human intelligence in many areas (image recognition) and improvements in forecasting and prediction. In big data, we have extremely affordable parallel systems that can process big volumes of data. There are optimization and simulation packages that allow for sophisticated analysis (bayesian A|B test, multi arm experiments) with far less effort than a few years ago. In summary, what I’ve seen is that we can have more sophisticated analysis now and run complex experiments.
Are you able to share some insights on how the use of Machine Learning and Artificial Intelligence has boosted the customer experience at Uber?
At Uber, we communicate with millions of users around the world. We send hundreds of millions of emails per year, among other communication channels, and improving opening rates for example, translate into a better customer experience and retention. We train predictive models that use ~100 or more variables to predict open rates, and we also use natural language (NLP) techniques to classify content using AI techniques.
In experimentation, we rely on ML models for segmentation by ‘propensity’ to act and isolate the true signal from the test. Additionally, in marketing allocation, we are using neural nets in forecasting as inputs to allocation models. The results of these efforts – the fine tuning of communication frequency, segmentation and therefore communication content differentiation, better the allocation of incentives. As for the product side of things, ML and AI are used to improve the pick experience, better routing and better occupancy of cars among many others.
As an industry player, what do you think is key to driving innovation in the industry? We need to be aware of industry and social trends and be able to experiment rapidly and adapt. In marketing, we constantly run tests to better communicate with our users and partner closely with product to educate and promote usage of new features gathering feedback very quickly and iterating.
Grab has recently announced its venture into financial services. Do you see Uber heading down a similar path?
The only thing I can comment on is the recent launch of the Uber Visa card, which is a card with large cash back rewards for riders and faster payment times for drivers.
You will be speaking at ConnecTechAsia Summit 2018. What do you think will be the most important takeaway for your audience?
My presentation includes a subsection on how we can interpret ML models without relying on heavy math.I will also show how we have used these models creatively; the audience will learn these practices and gain a deeper understanding of these exciting technologies.
Keen to know more about the use of machine learning and artificial intelligence for marketing purposes at Uber? Join Mario A. Vinasco (Data Science & Marketing Analytics Manager) at ConnecTechAsia Summit 2018’s EmergingTech Track. 28 June 2018, Level 4, Marina Bay Sands.
Delegates may register for the Summit here.