Seamus oversees infrastructure, software, and data security policies and practices. Seamus has an extensive background in IT security, cloud computing, database technology, and machine learning. He graduated from Princeton University. Over the last 13 years, he’s developed and implemented IT security policies and strategies for platforms handling some of the world’s most sensitive consumer data.
Faraday optimizes every stage of your B2C revenue journey, from acquisition to retention. Finally — machine learning and big data that actually work for your business. We believe data is capital. It’s the purest, most powerful fuel your business has ever had. If you’re not putting it in a rocket ship, you’re burning it in a bonfire, every day.
Tell us about yourself and the journey to co-founding Faraday.
Faraday grew out of a desire to apply Machine Learning to data that was otherwise just lying around. We had some early clients with customer lists and mailing lists, and they basically didn’t know what to do with them – they played the traditional acquisition / upsell / retention game.That involves a lot of guesswork – lucky guesses make you look great, but in the end it was just luck, not strategy.
We started to formalize how we looked at the data, and put it in terms that advanced algorithms could process. I will never forget the day a client ran a secret test against us and then showed us the internal report on it: we had beaten their models in the real world by a significant amount.
In your opinion, how have AI technologies been a game-changer for marketers and how they are able to do their job?
AI is helping marketers do their jobs more efficiently by automating and optimizing processes like ad bidding, web design, copywriting, and campaign targeting.
The most important aspect of marketing is truly understanding your ideal customers: who they are, their interests and hobbies, and ultimately, why they choose to engage with your brand. The answers to these questions guide marketing strategies. So, while optimizing ad bidding is undoubtedly important, ensuring your ads are targeted at the right audience with the right message is critical. That’s essentially how we help marketers leverage AI.
Tell us about the FIG database and how businesses can leverage this knowledge resource.
The Faraday Identity Graph (FIG) is Faraday’s nationwide consumer database containing more than 400 demographic, psychographic, and property attributes on approximately 235 Million US consumers. This data is used to enrich our client’s existing customer and prospect data with hundreds of additional attributes. That rich, cross-referenced data is then used to train our machine learning algorithms, which build models capable of predicting — with a high degree of accuracy — our clients’ desired outcomes. For example, a client may want to predict whether a customer is likely to churn. We’ll use the client’s enriched customer data to train the machine learning engine to build a model that can identify churn-prone customers, enabling our client to take preventative actions with those customers.
Can you take us through how Faraday’s technology is used to calculate lead scores?
We build a model that differentiates between positive outcomes and everything else. Sometimes we have data about really negative outcomes; sometimes we don’t (this is a classic problem in machine learning). To get a lead score, we give this model new data that it has never seen before and it comes back with a prediction and a confidence. We combine that confidence with the expected rate of positive outcomes and arrive at a score between 0 and 1.
What kinds of insights are drawn to facilitate predictive targeting for clients? How customizable is this process?
Data-driven insights are especially helpful in optimizing creative processes like content creation and ad creative. Because predictive models can identify individuals likely (or unlikely) to convert on the desired outcome, they’re great for building propensity-based audiences used for targeting purposes. When you combine the two, you can personalize your campaigns for audiences that are likely to purchase your products. We use our clients’ enriched data as the basis for analysis and modeling, and every insight and prediction is innately customized for each client.
AI is often seen as an expensive offering. How does Faraday cater to SMEs?
Traditionally, operationalizing AI requires large datasets, a machine learning engine, data scientists to build and validate predictive models, and software engineers to develop systems to feed predictions to the necessary destinations. Faraday includes everything needed to operationalize AI, and is streamlined for consumer-facing organizations. Thanks to our pure focus on the consumer journey, we can offer actionable AI faster and more cost-efficient than in-house or blank-slate AI solutions.
What role does an outreach partner play in the Faraday ecosystem?
Faraday offers an app-like store for outreach partners: once you have made deliverables and predictions on the Faraday app, you can push them out to any of the supported partners. Our clients require this — they don’t just want Facebook ads, for example — but we also customize each output so that it maximizes the effectiveness on the particular platform. For example, if you are marketing to families, we may generate multiple names for every address that you want to reach, to ensure that the ad network can find somebody.
How does Faraday integrate with other enterprise software that a company may already be using?
As with our outreach partners, we have something similar for existing enterprise systems (business intelligence, data warehousing, etc.). Chances are a system is supported out of the box and then goes through an authorization process to let Faraday make predictions using the data stored on that system. Clients keep their existing systems most of the time and Faraday becomes the AI pipeline that’s hooked up to it.
How does Faraday help businesses prevent churn?
It’s all about the machine learning process. When a client tells us they want to prevent churn, we’ll instruct our machine learning engine to build a model that can identify churn-prone customers. Once the model is deployed, our clients can build audiences of churn-prone customers, push those audiences directly to key outreach channels, and intervene with a special offer or promotion before it’s too late.
Alternatively, you can proactively prevent churn by using the model to identify churn-prone individuals before they actually become customers. This is especially useful when you want to avoid using your marketing budget to target individuals who are likely to churn regardless.
Congratulations on the recent funding! What’s next for Faraday?
We are growing our FIG dataset by an order of magnitude. We are adding bleeding-edge modeling capabilities such as deepnets and hybrid models, where multiple machine learning techniques (logistic regression, random forest, deepnet) vote on a single prediction. We are expanding the “menu” of AI options that every Faraday customer gets out of the box, and figure out how to automatically prune it down to what best serves each customer.
Thank you, Seamus! That was fun and hope to see you back on AiThority soon.