TechBytes with Sean Zinsmeister, VP of Product Marketing, Infer

TechBytes with Sean Zinsmeister, VP of Product Marketing, Infer

Sean Zinsmeister
Sean Zinsmeister

Sean Zinsmeister
VP of Product Marketing, Infer

Modern marketers rely heavily on predictive intelligence and AI in building a true, single-source definition of customer experience based on data and analytics. The biggest question in a competitive AI landscape is how human factors would influence the industry. To get a clear picture on everything-AI and intelligent sales platforms, we spoke to Sean Zinsmeister, VP of Product Marketing at Infer.

MTS: Tell us about your role at Infer and how you got here?
Sean Zinsmeister
: Thanks so much for thinking of me for this series. I’m the VP of Product Marketing at Infer. I help craft the messaging and go-to-market strategy delivering AI products to some of the greatest businesses in the world.

Before I joined Infer, I was actually one of their first customers when I directed Global Marketing Operations at Nitro, a document productivity company known to many as the #1 alternative to Adobe Acrobat. Our team was in charge of architecting our marketing automation solution to grow revenue from our online buyers (smaller license deals, e-commerce) and demand for our direct sales team (volume license deals).

Nitro’s software is incredibly popular. So much so, we were dealing with over 100,000 inquiries every month of users who were either downloading content, activating a free trial and more. For a high-growth mid-market company, that’s an impressive amount of data. The challenges were how to effectively pipe all that through the system while not losing our shirts on data storage costs (a growing budget challenge for C-levels).

Infer was purchased by Nitro’s head of revenue at the time. The technology was dropped on my desk to “figure it out.”

A vague and sinuous directive I’ve become accustomed to in my career. Before I got into Marketing Operations and Product Marketing, I studied Music Composition & Theory, which I whimsically call my first introduction to artful algorithms.

From there, I worked professionally as a sound designer and audio engineer for many years – this stage marked my first introduction to systems thinking and design or as I call it “The Art of Signal Flow.”

It was ironic really. Sound Design has very similar goals to marketing. “Make it sound good” vs “Make a message that influences, resonates, or calls to action.” To me, they are one and the same. Audio engineering presupposes many different modules, a sequencer to edit events, and a focus on capturing the highest-quality inputs you can. My inputs at the time was audio, and fast-forward in time just before Infer to draw the correlation with Marketing Operations, quality sound and quality data are keys to the same goal – “Make it sound good.”

The second part that fell into place for me was the technological orchestration. A careful dance any marketing technologist must exercise. As I mentioned earlier, both Ops and Sound Designers must integrate a number of different modules, software, and controls to achieve the most optimal or desired outcome.

When Nitro purchased Infer, my eyes were open to an entirely new blueprint for automating the customer journey.

A way for our system to be more efficient, reduce costs and even provide guideposts for customer journey sculpture. It would feel like adding premium oil to your car, it breathes new energy into the machine. At that point so many locks burst open in my mind.

My imagination was in love with both the Infer team, and the incredible work they were doing in AI and predictive analytics. Infer says that you do “300+ million predictions using thousands of models.”

MTS: How are predictive models built for Sales and Marketing Intelligence platforms? What volume of data would give the most accurate results for sales teams to build their Predictive Behavior Models?
Sean: Before we start diving into model building, etc. I want to state from the outset that it really depends on the type of problem you are trying to solve. The onus is on the business – they have the domain expertise. It’s then up to our team to translate that into a machine learning problem. Machine Learning itself is a solution. To be most effective, we need to clearly understand the business problem you are trying or believe you are trying to solve.

In addition to defining the problem, it’s important to also help define success. For example, our problem at Nitro was that we had too many leads. Human capital couldn’t possibly slug through all of them, but we didn’t want to miss any of those that expressed intent and looking like a great fit for our product line. The solution was our predictive scoring model, and success for us was a conversion point defined as inquiry to SQL.

For model instruction, I like to think in “3’s” which is actually a simple method to think about how predictive models are constructed.

You have inputs, an AI engine, and outputs:

Depending on the type of model, you first need to synthesize all of the databases you want to use. This is where a fluid dialogue with the customer becomes vital. They might believe certain data sets are more valuable than others. Many times they are surprised when new signals are bubbled up, or not considered at all predictive in a model. Once you collect all of the data, which for many businesses starts with CRM and Marketing Automation (MAP) we then work to prepare the data. Datasets need to be sanitized after they are centralized (lots of -izes involved in this process certainly). Customer data is never clean. Most have missing values, incorrect values, and more clever issues. This means data needs to be normalized and enriched with 3rd-party sources in order for a trust-worthy model to be built with confidence.

We then divide the data into training and testing sets, and our platform also allows us to test multiple algorithms (something unique to Infer) to allow us to understand which is the best to be used – it could be Naive Bayes, Random Forest, just to name a few. A majority of the algorithms used are classification models, which makes sense given the common use cases for Sales & Marketing teams – filtering, prioritizing, segmenting and qualifying prospects.

There are two primary types of models that we build: Fit and Behavior. Fit Models consist of mostly static sets of data like technographics, firmographics, and demographics. While behavior models are time-based. The data set is not flat. This is why our data science team has separated these into two separate products that we offer. You can read more about this distinction here.

 

What we are monitoring for behavior data is a collection of content engagement (email), and website visitor traffic to conversion. We crack open the Marketing Automation system and web analytics to look at the full spectrum of activity data being collected. Typically we like to see at least 2–3 months of lead data to build a behavior model.

MTS: Account-Based Behavior Scoring remains the pain point in ABM. How does Infer help marketers gain insights into Full-Funnel Behavior Signals across existing Marketing Automation platforms?
Sean: It’s important to remember that behavior is people-based. The challenge is how you fuse the connection between those people and their accounts. We collect activity data on the people-level (leads and contacts) and then use our proprietary lead-account matching algorithm to roll them up into an Account-Based Behavior score.

This is incredibly powerful for companies that don’t generate a vast volume of inquiry data, or have a limited universe of accounts they are engaging. The key then becomes how you are engaging and then triggering action on the key member of your buying center. Being able to monitor and predict complicated buying center behaviors unlocks an incredible amount of potential for ABM programs to execute effectively, not to mention a quick-look into how well you are penetrating your target market of accounts.

MTS: How should CMOs invest in training marketers and sales reps to better utilize the AI-based sales intelligence and predictive platforms? Does Infer provide any resources for such training?
Sean: We certainly make education a core of our product experience. However, marketers, in general, require a higher-IQ in data education. I highly recommend that go to market professionals teach themselves basic SQL and Python to understand some of the syntax and mechanics behind these AI tools their job will continue to rely on. Before entering the world of Marketing Operations, I was focused on technical marketing primarily around web and SEO. Before getting lost in tactics, I made it a point to teach myself basic CSS and HTML. I wanted to have a basic understanding of the language. You don’t want to deploy my code…but it helped incredibly to bond and collaborating with web developers and designers. The same formula holds true in the AI realm. Over the past few months I’ve made it a point to learn basic Python and SQL. It’s challenging but a lot of fun. It’s given me not only a deeper respect and admiration for my technical counterparts, but a new way of thinking about our technology and the next horizons for AI.

Marketing is going to continue to be a more technical field. My advice is to make sure your team is building technical skills in order to develop a new way of problem solving and thinking. It’s invaluable.

MTS: “What’s the deal with Intent Data?” If marketers are yet to fully capitalize accounts within their addressable market, then how would they hunt down prospects that are not connected?
Sean: 3rd-party intent data is certainly an exciting new frontier. It’s encouraging to observe new vendors innovating in this space and business users experimenting with new use cases. However, if Sales and Marketing are doing their job – shouldn’t you already have a good idea of your TAM? Perhaps it’s not the complete picture where all these new tools can help fill in the gap, but these to me feel like coloring in the edges. There’s so much you can learn about your TAM with 1st-party data generated from owned media (whose signals are highly predictive), not to mention 3rd-party market research, customer interviews, analyst reports and more. Sometimes for the best data you have to hit the streets. Marketers are always looking for a “push-button solution” and my friends it doesn’t exist. Marketing is hard, and you need to take a portfolio approach to be successful. Intent data can be a part of your winning equation, but there are so many other areas to invest in before you get there.

MTS: How do you see historical algorithms, predictive analytics, and AI-based sales acceleration technologies converging to deliver better ROI on MarTech investments?
Sean: I like to think of Sales & Marketing as your GTM (go-to-market) team, which is why I’ve leverages that nomenclature throughout our discussion. They really are a part of the same strategy. What’s cleverly evolving is that things like ABM, and outbound have Marketers creeping down the funnel, while sales automation (mini-marketing automation platforms really) is allowing sales professionals to creep up the funnel.

Before you even jump into the stack I encourage C-levels to look at how their organizations are structured. What is the culture? How is success measured? Is there separation by design like in some large enterprises, or are sales and marketing joined at the hip.

Most importantly, how does work get done? How does an idea travel from boardroom imagination to project deliverable? You don’t need a lot of fancy tech to accomplish that part. Project Management is the unsung hero of organizations. The ones that truly “get things done” and execute effectively have what I call “The Marketing Supply Chain” down to a science. When evaluating your organization, start with your process, then figure out if there is a talent-gap or a technology-gap that needs to be filled.

MTS: Would voice-based search and AI-based sales assistants take over the human factor out of equation?
Sean:
Possibly, but it feels like we are quite a ways away from that. It’s more likely that early use cases aid the frontline, allowing customers to shorten the distance between question and answer. AI-sales assistants should be your best friend. You know, the one you call who has all the answers. But this doesn’t necessarily take the human out of the loop. In fact, I used to help design IVR (intelligent voice recognition) systems for companies (if you’ve ever called FedEx you’re familiar with my work) and many times people just mash “0” to hell in order to get to a human.

People in general are just after the best experience. If we can get there with AI or assisted by AI, then customers will rejoice.

If the technology creates more friction – similar to the phone tree forests we’ve navigated for years, than the focus needs to be shifted onto the product design and user experience for AI products.

MTS: Thanks for chatting with us, Sean.
Stay tuned for more insights on marketing technologies. To participate in our Tech Bytes program, email us at news@martechseries.com

Previous ArticleNext Article

Leave a Reply

Your email address will not be published. Required fields are marked *