Data science is exploding — which is only natural, given our ever-changing culture and ability to learn anything within the grasp of our fast-moving fingers.
Consider, for instance, that within the next year or so, the internet will have reached its 5 billionth user. Meanwhile, people will be performing about 1.2 trillion Google searches a year to access the internet’s 1 billion+ websites.
With each Google search (and other such internet activity), a data footprint is created, replete with information about the user’s interests, behaviors and demographics. That means the world is swimming in data. Therefore, many martech firms, Strike Social included, must rely on the brainpower of data scientists to sort out what information is useful and what should be discarded.
Rethinking the impossible with data science
Data scientists are often the ones to ask the big questions that may seem impossible to answer. Their inquisitiveness also enables them to develop new models at the forefront of technology. They achieve this through structured experimentation, such as restructuring parameters or combining diverse data sets.
It’s not surprising then that many data scientists come from academic backgrounds, with advanced degrees in fields like biology or physics.
Strike’s data scientists Bing Bu, Dmitry Bandurin and Jeongku Lim work at the company’s headquarters in Chicago.
The latter is precisely the case for Strike Social’s trio of data scientists — Dmitry Bandurin, Bing Bu and Jeongku Lim — who all hold Ph.D.s in either experimental or elementary physics (that is, the study of the basic building blocks of matter and their interactions). Before joining Strike, Bandurin, Bu and Lim were research scientists who focused on smashing elementary particles inside big colliders to make sense of the universe.
Now, they are taking on the world of paid social.
As Strike’s Senior Data Scientist, Bandurin credits his research background with the ability to rethink what is possible with big data. And he’s accustomed to working with real, experimental data to produce new results.
“It is never the same, and it is always changing,” he says.
Bandurin, Bu and Lim all agree that data scientists need strong mathematical and analytical skills as well as programming chops to be successful in harnessing the power of AI for paid social.
So what exactly does an average day look like for them? Bandurin estimates that meetings take up about 25–30% of his time, with another 10% spent working with developers or discussing new findings with the data science team. The rest of the time, these data scientists are focused on testing and developing models, then implementing them in prototype codes.
Bu says he also likes to summarize each day’s work and plan his schedule for the next day, week or even a month in case a long-term project comes his way.
How data scientists can improve paid social
The clusters created through complex modeling help data scientists test their results in managed advertising campaigns. Strike’s data scientists and media teams work together to develop micro-campaigns that enable discrete testing of data combinations. When a combination performs or meets key performance indicators, ad spend is re-allocated from under-performing ad sets to those more on target.
Campaign results are then fed back into the data mix, where Strike’s scientists continue to refine statistical models for improved performance over and over again.
The continuous process of inquiry, modeling and testing never stops in the world of data analytics — nor can it stop, for the data is ever-changing. When people grow up, they gain interests and drop old habits. Culture also evolves, as evidenced by improved communications methods.
With technology, techniques perfected over years are now eclipsed in a matter of months, such as the accelerated pace of learning available with AI.
“Given the exponential rise of usage of the smartphone, smart TV and other advanced electronics, collecting individualized information is achievable, which will enable dedicated delivery of individualized ads,” Bu says. “Going forward, the network will be cheaper, and coverage of usage will get broader, so the audience for video ads will grow quickly.“
Without the inquisitive minds of Strike’s data scientists — and their fearless commitment to data experimentation — AI’s technological advancementswould not be possible.
Here’s how their work is changing social media advertising.
Strike’s data scientists Jeongku Lim, Dmitry Bandurin and Bing Bu all hold Ph.D.s in either experimental or elementary physics.
Data science for better audience management
With the massive amounts of data being created every second, marketers must control against faulty results from inherent biases, incomplete data sets or too small of samples.
A data scientist understands that audiences aren’t built solely on demographics but are composed of people with different behaviors, pain points and interests.
Quality data analysis incorporates behavioral clues from cookies, web analytics, user-generated content and other big data sources. To build out detailed and useful audiences, data scientists fuse large data sets to allow big data to form segments that offer real insight into the behaviors of their customers.
Quality audiences are verified by testing in advertising campaigns and are dependent on the recency, frequency and depth of data.
Remember, audience creation starts with a hypothesis based on known variables and goals. For example, an insurance carrier’s initial assumption might be: individuals seeking online auto insurance, between ages 18–50, who own at least one car. A well-formulated hypothesis sufficiently narrows your analysis while yielding enough results to discover behavioral and motivational insights.
Data science for proper attribution modeling
Proper marketing attribution, or the science of determining what message drove a purchase, relies on data from both converters and non-converters alike. As this data can be very big, advanced modeling is needed to correctly identify and credit the event that led to user conversion.
Thanks to improved technology, such as AI, brands now better understand the consumer’s pathway to purchase. With enough data, scientists can look across marketing channels and devices to improve touch points and enhance messaging.
Data science for better real-time bidding
Advancements in audience segmentation and a deeper understanding of conversion events have led to the practice of RTB, a method to buy and sell ads. RTB allows an individual ad impression to be purchased simultaneously with a user’s visit to a website.
If you’ve ever looked at a product on a website, then transitioned over to check out your social media feed, only to see an ad for the same product, you’ve probably experienced RTB through a targeted ad.
Or, say you bought your first home, and you’re tired of eating off plastic plates. You decide to visit Macy’s online to look for new flatware. Not ready to buy yet, you decide to visit Facebook to see what’s happening with your family and friends. While scrolling through your feed, you spot an ad with the exact image of the dinner plate you had just been viewing.
RTB scales up the buying process and enables direct targeting of individual users. To participate in the process, data scientists must have access to vast amounts of data and possess the right expertise to sort through and retrieve useful information for actionable insights.
Where data science is headed
At times, the field of data science appears to be expanding as quickly as the universe that Bandurin, Bu and Lim spent so many years exploring.
“Data science will continue helping different businesses solve problems, making things more automatized,” Bandurin says. “Development of self-driving cars is one example — but also automation of other vehicles, including aircraft, playing chess, helping people with disabilities and true androids in all spheres of human life.”
The result will be an entirely new world as we know it.
