Data Scientist, Goodway Group
Marketing automation powered by AI and machine learning would rule the roost in B2B. To better understand how AI-based marketing automation technologies would fare in 2018, we spoke to Scotty Pate, Data Scientist, Goodway Group.
Tell us about your role at Goodway Group and the team and technology you handle.
I am a Data Scientist at Goodway where I work with a small, brilliant team of software engineers and another Data Scientist. We work on problems such as fraud detection, accurately pricing digital media, campaign objective optimization, and user behavior targeting. Our technology stack is a mainstream, open-source, distributed computing platform with lots of tools such as Python, R, and various data stores.
Define the state of ‘AI-based marketing automation’ technology in 2018?
As the scope of problems that we solve with machines increases, we have a tendency to reduce the definition of what qualifies as Artificial Intelligence. I think we have to be careful to not have a standard in which the boundaries of what constitutes AI becomes only the things that we haven’t yet accomplished. At the same time, there is a potential to misuse the term and overstate its application. I like to think of our model as complementary intelligence. I would be skeptical of anyone laying claim to having implemented an AI that solves all marketing technology problems. There are areas of marketing technology that are great use cases for machines, and there are areas where machines just aren’t needed. AI-based marketing tools are within reach of more organizations than ever. I suspect there is an inverse relationship between the amount of praise a company gives their AI tools and the sophistication of their implementation. The companies that have high functioning AI systems will tell you as much about the shortcomings as they do the benefits.
As a Data Scientist, how do you enable marketing and sales teams to better optimize their campaigns?
One of the main goals of optimizing a campaign is paying the right price for the right inventory. Digital media exchanges are relatively immature markets. There is a massive amount of inefficiency around price discovery in digital media markets when compared to something like a financial market. Pricing mechanisms tend to be opaque and one of our primary goals is valuing the inventory that we are purchasing accurately. In order to value inventory accurately, we need to have a clear understanding of campaign objectives. If we know what the campaign objectives are then we can let the machines tell us what price we should be paying for inventory. We try to leverage technology to help our marketers know what inventory and which users are contributing to their campaign goals.
Would you consider Data Science and AI/ML turning ubiquitous to every marketing and advertising platform?
The techniques of machine learning are ubiquitous in the sense that platforms which don’t embrace them will go away. We aren’t yet at a place as an industry where these techniques are widely accepted and applied. I still have many conversations with people about the basics of why last touch attribution isn’t appropriate. Much of that resistance to change may just be industry inertia. Explaining to marketers why the abundance of clicks they receive may not be correlated with their campaign objectives is a tough conversation. Another practical reason is the infrastructure and people needed to implement machine learning tools are expensive and time-consuming. An organization has to be committed to competing on the basis of these tools from the top down. As the amount of data generated increases, machine learning tools are going to become mandatory out of sheer necessity.
Tell us about your recent launch- RealValue?
Impression pricing is one of the areas of real-time digital markets that is ripe for machines to take over. RealValue is our suite of tools that we have in place to leverage computational power to value inventory. Our clients care about their campaign objectives, not what price to pay for a single impression. We have to communicate their business value down to the impression level or else we can’t participate effectively in real-time media markets. RealValue is a translation layer between our marketers’ business objectives and impression level pricing that supports those objectives. We want to precisely state our value for any given impression so that we know where to aggressively go after inventory and where to stand down. In order to do this, we have to attribute credit, analyze the data, and take follow-up action based on our analysis. Letting machines solve for pricing algorithms gives our human traders more time for strategic engagement with client objectives.
How do you see emerging technologies like video, AR/VR, and Artificial Intelligence further impacting the B2B buying journeys?
All of these technologies are working toward shortening the sales cycle. We have significant interaction with other businesses in real-time through exchanges without having much human interaction at all. Our algorithms interact with their products and the sales cycle is measured in milliseconds. We have had situations where our algorithms favor other business products so much that it pushes us to reach out and have conversations about what they are doing. It reduces transaction cost and risk for businesses to interact with external product offerings. Businesses can test products without commitments and on their own schedule, and let the product sell itself. It is a “show-me first” mindset.
Thanks for chatting with us, Scotty.
Stay tuned for more insights on marketing technologies. To participate in our Tech Bytes program, email us at firstname.lastname@example.org