By: Winnie Shen | Director of Data Operations, Zeta Global

Moving Beyond First-Party Data

Every marketer understands the value of first-party data. It’s a given that we all understand how important our data is that we house in our own databases. The challenge comes when we need to grow beyond our existing data, enrich those audiences and expand sales. That means gaining a truly 360-degree view of our customers. Many marketers today operate between a “probabilistic” and “deterministic” landscape, lacking a complete view of who their customers actually are. Our goal at Zeta is to utilize your existing customer data while leveraging our deterministic behavioral signals to identify and activate high value customers and reactivate dormant users, creating new growth opportunities for our clients.



Natural language processing (NLP) is the key to helping us understand the content consumers are engaging with. When we leverage it properly with artificial intelligence technology to look at behavioral signals across the open Web, we can identify clusters of individuals, most relevant to our client brands. Here’s an example of how we’ve done just that for one of our apparel clients.

 

Using NLP to Activate, Reactivate and Build Brand Awareness

Our client is a national clothing retailer with multiple subsidiaries. For this program, we were working with the following client segments: basics for all; full-figured apparel; fashion forward for younger consumers; and women’s apparel.

This client’s challenges were not unlike many other retailers or, for that matter, any B2B or B2C clients we have today:

• To activate and find new users (across all brands)
• To reactivate dormant users (for two brands)
• To build brand awareness

Our approach to helping our client with these challenges included utilizing their first-party data and segmenting customers based upon the partner’s use cases. We divided our client’s customers into active vs dormant users to segment out those consumers we needed to re-engage.  Within the active segment, we further differentiated between high and low-spenders so that we could identify unique behavioral signals to the high-spender group and find similar new customers across our network who we could engage on behalf of our client.



Using our Data Cloud methodology, we developed deep insights around our client’s customers and provided actionable solutions catered to the specific challenges our retailer was facing.

Zeta’s Data Cloud is powerful due to 3 key pillars:

• Identity: The heart of our database is identity and the permission-based users we have the rights to market to. We resolve identities definitively back to email which is a persistent identifier and more effective than cookies which have a short shelf life and can exist across multiple devices.
• Behavioral signals: We “listen” to behavioral signals across the open Web and understand those signals to derive intent and build audiences around those signals
• Connectivity: We connect the dots by mapping signals to real individuals and activating audiences across digital channels

At Zeta, we monitor over 2.5 million unique monthly visitors which translates into 18 billion content consumption signals on a monthly basis. We use that information and leverage NLP to translate those signals into audience clusters.

 

What’s in a Customer Signal?

Let’s spend just a minute talking about the importance of signals and what exactly they are. Signals depend upon content, subject matter, frequency, and recency. We use NLP to help us understand the content itself. For example, we may review topics from a blogging web site that reports on how to create chic maternity looks without using maternity clothing. Some of the keywords we might include might be things like accessorizing, cute clothes, affordable apparel. Once we identify the keywords and concepts from lifestyle blogs, we translate them into audience taxonomies.

Each cluster is derived from multiple keywords at different confidence levels.  For example, silk blouses is related to women’s clothing at a higher confidence level than say dry clean which is a more general term and relates to women’s clothing at a lower confidence level.  We then score consumers based on their content consumption, frequency, and recency of that consumption against these audiences.

There are two things to note about our audience taxonomies.  First, they are custom and growing where we’re continuously building new audiences based on our client needs.  Secondly, they are fluid where consumers move in and out of audiences in a real-time basis.  As an anecdotal story to bring these points to life, we created a Black Friday taxonomy based on customer demand.  Once developed, we saw a 12% day-over-day growth of consumers moving into the audience the week leading up to Black Friday, illustrating the value of real-time signals across our network.

We use environmental, transactional and most importantly, behavioral data to help truly understand the unique individual.  Each data point alone provides us some context around a customer, but it is the combination of these signals that provide us insights around individuals that we can activate in real time.



 

In the case of our apparel client, we analyzed 27.4 million customers across four of their brands. We found that 37% of those customers were not digitally connected, meaning our client didn’t have a way to communicate with these consumers through digital channels. We were able to match 14% of the digitally disconnected users to our permission-based universe where we have the rights to market to those customers.

We then matched their database against our network and found 47% of the client’s customers were active in our network. We performed a deep dive of the active vs. inactive and high vs. low spender segments to develop insights that were less intuitive than others. For example, we saw that moving and phone and internet services indexed higher for the basics for all brand which meant these customers were likely new movers. If consumers recently moved, they may have purged a lot of their apparel and are looking for a wardrobe refresh. The second interesting find was around frequent travelers. We saw customers interested in cruises and vacations indexed higher as well.  When thinking about vacations, consumers often look for new apparel to bring on their trips, especially when traveling from cold to warm weather locations and vice versa.

For each brand we found some interesting insights:

• Full figured women were very active in their lifestyle and aligned perfectly with the athleisure wear explicitly designed for them by our client.
• Our fashion-forward women’s wear group were frequent shoppers, new moms with most likely changing body types.
• The women’s apparel group were HOH shoppers who were price conscious, looking for a good deal and high quality.

 

Acquisition, Reactivation and ‘Real-Time’ Customer Experience Journeys

At Zeta, we have behavioral signals for customers in market for specific products and services and can overlay the behavioral information with demographics of users active within our network.  We can identify consumers who are looking for women’s apparel deals, are interested in fashion-forward merchandise, or are social media influencers. All of this rich data provides us acquisition opportunities for our clients.

But, what about reactivation? The second client challenge was around reactivation. We looked at dormant customers as defined by the client – 5.8 million customers. Of those inactive users, 30% were active within our network in the last 30 days, meaning the consumers were still active across the open web, but just not with our client. We then took that segment of users and performed the same behavioral analysis to gain insights that would enable us to reactivate those former customers.

We bring to life all the rich insights we developed through the customer experience journey, based on the specific client use case: acquisition and reactivation. We monitor behavioral signals on a real-time basis, identify audiences to activate against based on our insights and engage them through an omni-channel approach with 1:1 messaging. Finally, we set up a continuous feedback loop so that we can optimize and refine our solution for our clients.

If you are looking to grow beyond your first-party data like this client, let’s talk! Are you interested in how you can identify audience clusters to activate on a real-time basis and monitor behavioral signals using NLP? Visit us at rhyzetamigrdev.wpengine.com to learn more.