Mathematics has always attracted those brilliant few who see more than ones and zeroes. Some mathematicians — like Alon Amit co-founder and VP of product at the marketing analytics platform Origami Logic — look beyond figures and see stories.
These stories are the bread and butter of a data scientists, individuals who thrive in the sea of data that has swallowed many marketers by bringing context to numbers. “A lot of the work of data science, machine learning, algorithms is not in the making of algorithms themselves, but in the handling and transforming of data [and] getting data to the right place and making sure it’s clean and consistent,” Amit says.
This is the supreme challenge of the modern marketer, taking disparate datasets from multiple sources and channels, and making that data make sense. Using data science to achieve this is more than simply filling a role for Amit. For him, data and the math behind it are a way of life.
Amit has been immersed in numbers since his childhood. His post-secondary education largely mirrors that of a standard mathematician. “I studied mostly math. I have a PhD in mathematics, and an undergrad in computer science and physics,” he says.
Amit has a long and storied career as a practitioner of data science, though that’s not always been his official role. Indeed, the title of data scientists is itself quite new. “I was always interested in algorithms, and in doing what it now called data science. Back then it wasn’t called that,” Amit says.
Amit witnessed the evolution of the algorithms first hand, from its early uses in fields of science through to its more recent implementation in marketing and advertising. Amit worked as a project manager on Google’s machine learning technology for nearly three years, before moving to Facebook, where he was project manager of the company’s decision making system, the backend of its advertising.
He then went on to co-found Origami Logic, a company singularly focused on contextualizing marketing data, where he’s worked as its VP of product for almost five years.
“All the data that feeds into [Google’s] algorithms is Google data. Google has complete control over this data, and even then the challenge is not simple, Amit says. “Contrast this with the typical marketer today. They need to take into account information that is flowing in from Adwords, Doubleclick, YouTube, Facebook, Twitter, their email marketing system, their mobile marketing system, and a bunch of others. So, the challenge of extracting useful insight from marketing data starts first with the data itself.”
Through Origami, Amit aims to address this common pain point. The organization helps marketers bring data in, put it together, and essentially, finally, get a good look at it.
A significant portion of this type of work is canvassing. Amit believes that data scientists have to get intimately familiar with data before they can put it to any effective use. After this scouting period, data scientists can begin to iterate on algorithms that contextualize that data, and ultimately tell a story from it. “This characterizes a lot of data scientist work elsewhere,” Amit says.
There is nothing simple about this journey. But once marketing data has been harvested and harmonized, the algorithmic process begins, and it’s here with the raw mathematics that Amit truly excels.
“The process [of developing algorithms] is by necessity and definition, very iterative. One of the things you learn very quickly is [how] important it is to have consensus on what you’re trying to optimize for or achieve,” he says.
Amit illustrates a scenario where a team works for two years producing better algorithms, and everyone knows that that’s been the focus of the team. But it’s difficult to truly know whether these new algorithms are actually better than the initial ones due to the potentially divergent goals of different people within the organization. “Think about an advertising platform, for example. You could say, ‘We have higher click-through rates,’ and that’s great. But you have lower costs-per-click, [so] maybe not as great,” he says. “What do you care about more? Making money, or keeping users or advertisers happy, a combination of the three? What combination? So, if you don’t have consensus on those ad serves you can end up spinning around.”
Amit concedes getting this level of consensus is not easy to do, but when everything is aligned, the workflow breaks down into various production and experimentation cycles. Production generates dataset A, while the various experiments to incrementally improve the algorithm generate dataset B. The organization continues iterating on this path until launch, if launching is plausible. “That’s the way these teams operate. They keep tweaking, they keep optimizing. The question is what they’re optimizing for.”
Data science is obviously highly complicated material, not unlike the advanced math that powers the discipline itself. Indeed, many businesses outsource their data science needs to companies like Origami Logic. But even in that case, it’s important for marketers to understand what is happening with these advanced algorithms and systems.
“In a B2B context where you’re selling a product or algorithm that involves some aspect of machine learning, it’s important that your sales team understands how this new thing is better, and why,” Amit says. To this end, Amit avoids jargon and trendy buzzwords in favor of directly focusing on the goals of the algorithm, and the expertise of the team that created it. “That’s what the sales force needs to focus on. What is the benefit to [the client or customer] and how are you gauging it,” Amit says. “It’s different for consumer facing products that only need to improve metrics for the consumer.”
In the end, this all comes back to the story in numbers, a narrative which is spun by the algorithms people like Amit help construct. These stories explain both the purpose and viability of an algorithm, which in turn helps marketers better understand what their data means. It’s a point of passion for Amit. So much so that he remains a force in the Bay Area mathematics education community by teaching kids.
“I’m immersed in the world of mathematics, but in my career, I chose fields that seem to rely on statistics, data analysis, data mining, machine learning. All of these different buzzwords that are essentially saying the same thing: you have a lot of data that you need to parse and do something useful with.” This idea manifests in Origami Logic, which Amit says started with a goal of showing that good algorithms can help glean better insights from data.
Amit’s is a non-traditional, but increasingly conventional path for the modern mathematician. “For decades mathematicians could only do math that wasn’t much applicable to anything. Today that’s just not true,” he says.
