This month for Inside SVG, we spoke with Johnny Hilgers, Director of Business Analytics and advocate of small meals.
Let’s get right into it with a straightforward question. What’s your favorite part about your position at SVG?
The fact that I am constantly learning every single day. I think this is due in large part to the amount of faith our CEO has in the things that we are doing. It allows us to continually try out new tools and methods without being afraid of failure, which I strongly believe will ensure that our business never becomes stagnant. Being surrounded by smart, outside-the-box thinkers doesn’t hurt either.
You’ve been working a lot over the past several months with bid optimization. Tell me a bit about what this means to us as a company and why it is so important.
Optimizing marketing campaigns for the right bid price is an art (and a science). Price is sensitive – it can’t be too high or too low and the decision of placing a bid and evaluation of the bid price needs to be performed in a few milliseconds.
So how do we, as a progressive organization build a proper model?
On the modeling side, hundreds of questions must be answered before anything is put into production. What algorithm or combination of algorithms is right for this task? What attributes do we have that will be the most helpful (feature selection)? What attributes can we create that will provide the most lift (feature engineering)? What attributes should we leave out because it actually hinders model performance (overfitting)? When feeding our model data to munch on, how far back should we go (training)? Too far back may represent a time when we had a different data collection process (noisy data) or a different mix of agents (newer agents may initially degrade quality lead performance). Lastly, what evaluation metrics are we going to be optimizing for? These are just a handful of the questions we had to ask ourselves during the modeling process
I would be remiss if I didn’t mention the heavy involvement of our IT team in projects like this. They are the ones who take our concepts and put them into code. Without that side of things, we’d have great ideas without any execution.
Looking back over the past twelve months, what’s been the biggest development outside of the aforementioned optimizations and modeling?
I would say it’s the type of analytics we are doing. We’ve gone from using real-time Tableau dashboards for Descriptive Analytics to answer questions like “what has happened” to using some really cool open source statistical software for Predictive Analytics to answer questions like “what could happen?”
With that in mind, and knowing that a lot of momentum is heading towards programmatic buying, what do you think will be the biggest change in our operations over the next twelve months?
I think we’ll be using a lot more optimization and simulation algorithms that fall into the last stage of analytics, Prescriptive Analytics, which answers questions like “what should we do”? This will help us better understand what the future may bring and advise us on possible outcomes, allowing us to maintain our competitive advantage.
Brunch or nah? Small snack, yes. Keep that metabolism humming.
Best sporting event you’ve attended? Royals/A’s Wild Card game in 2014
Favorite Local Band? Pomeroy
Apple or Android? Seriously?
Do you even lift, bro? Just cardio right now, but I’m planning a bulk for this winter