Machine Learning: The New Frontier for Marketing
NextRoll's Tech Stack Series #1
There isn't much that doesn't run without machine learning in today's marketing technology industry. Machine learning is the backbone of MarTech and has been for over a decade, giving companies the insights they need to make smart decisions and achieve success. At NextRoll, our core technology fuels two high-growth, separate business units, which means we digest a massive amount of data. And our longstanding, well-trained machine learning system allows us to create over 150 billion daily buyer predictions (roughly 20x the average daily volume of the New York Stock Exchange) for our customers in both arenas. So, where are we on the machine learning journey? We started early, adapt often, and continue innovating to provide the most value for our customers and stakeholders.
Machine learning has been an early foundation of NextRoll’s technology stack. We started investing in and testing machine learning in 2014 because we realized our early systems collected terabytes of user data, but it was not being used to its full potential. This led to the formation of our Data Science Engineering (DSE) team, which started to create the algorithms that power both our AdRoll and RollWorks platforms.
Our DSE team is unique. While most companies separate data scientists from machine learning engineers, we combine those responsibilities into one role. The benefit is that our teams understand the mathematical consequences of altering code while giving them a faster feedback loop. This permits them to test changes to systems quickly and helps them focus on techniques that will improve our services.
All of this means that our machine learning tools are constantly being upgraded and enhanced to provide our clients with the most significant insights and predictions. Both the AdRoll andRollWorks platforms leverage these powerful tools for everything from the placement of advertisements, to selecting the most effective creative asset, to targeting the best possible audience in a marketing campaign.
It works like this: our system processes data from the massive aggregated and anonymized user base of the internet, not from known individuals. To further help protect privacy, we remove any specific user identifiers before our algorithm combs through the data to detect patterns and make predictions. The software can use these patterns to serve the best ad with dynamic messaging or automate an email to the most relevant viewer or target in an economically viable way.
Online ads use an automated bidding system for placement on web pages and in search results. They all utilize a cost per mille (CPM) system, which is the rate an advertiser pays for 1,000 impressions the ad receives. However, each advertising platform uses a different system optimized for either CPM, cost per action (CPA), or cost per click (CPC), which is the rate per interaction, like a click-through of the ad.
Our AdRoll platform uses an automated bidding system that supports all businesses and brands, regardless of size, to compete for these ad spaces in a strategy that gives them the best bang for their buck. Its machine learning platform leverages massive and unique data sets optimized for small and mid-sized businesses to make intelligent, contextualized decisions on what to bid on for any specific opportunity.
The system goes even further by looking at other dynamics like time of day and what website the user is on, then it weighs the economic benefit for every ad placement. The massive amount of data allows the system to accurately evaluate the likelihood of the user becoming a new customer, permitting the system to make an appropriate bid. If, for example, the user is not a great match, it might bid a low value but will determine an appropriate value at which it’s worth the chance of gaining what might be an unlikely customer. However, if it is a high match, it will bid more to win a more likely customer.
AdRoll helps marketers and e-commerce sites do more than just ad placements. Its best-in-class retargeting algorithms use trillions of data points on shopper behavior to identify and reach site visitors with targeted information whenever they go online. When visitors abandon a cart, AdRoll can send automated emails to get the user to return and improve the chances of a purchase.
Our RollWorks platform leverages the machine learning algorithms to do similar tasks but focuses on B2B customers. These powerful tools can help businesses in several ways. For example, we have a massive database of companies across the country and around the world. A customer could give us a list of companies they successfully sell to, from which RollWorks can create a list of other net new prospects that would likely lead to further success. Moreover, our platform scans tens of thousands of news sites daily and interprets them for opportune moments in the company’s lifecycle. For example, the RollWorks system may notify sales teams about prospects that may have recently received new funding, or that have made a recent change in leadership allowing them to reach out with the proverbial “right message, at the right time.”
Furthermore, both AdRoll and RollWorks can help marketing teams in managing campaigns. They can identify the best platform to reach the targeted audience (social media, ads, emails) and set budgets to maximize exposure. Both services include robust metrics that illustrate the effectiveness of the campaign.
At NextRoll, our philosophy is to democratize digital marketing by building an accessible platform for businesses of all sizes. The digital marketing world is highly competitive, with thousands or hundreds of thousands of companies competing for a finite amount of space or ways to reach potential customers. Everyone is looking to target the right customer in the right place for the right price. This fragmented market space requires small and mid-sized businesses to spend their marketing dollars wisely. They need to be flexible, intelligent, and agile, plus possess the ability to reach their target audience for the least money possible. And, with our machine learning-powered technology, we give them that opportunity every single day.