The Changing Face of Audience Segmentation

January 5, 2017 — by MediaMath    

Let’s all pause for a moment and remember a simpler time. It was a time when consumer behavior was largely confined to a single desktop device. It’s back when the customer journey was described as a linear funnel. It’s a time before programmatic approaches had taken hold across the broad landscape of digital media.

Today’s digital ecosystem is vastly more complex, and one of the areas most deeply impacted is audience management. In the old days, it was reasonable to understand consumers by demographic, typically represented by segment attributes. Now, it’s important to also know events or behaviors related to a particular consumer and to be able to move them in or out of audience segments as these factors change. For instance, if you have a consumer who purchased some expensive shoes with an order value over $100, but you lack specificity on the conversion including product line, exact order value, or the product itself, you have no means to effectively present the complementing handbag that she doesn’t yet know she needs. You’ll probably just keep promoting the same shoes she’s already purchased!

Tried-and-true broad customer segments can work, but when competition is one click away, the most sophisticated marketers need to be able to leverage their wealth of customer data to interpret, create, test and adapt new segments in real-time. Technology now exists to make every single customer behavior and interaction available at marketers’ fingertips, to define and re-define customer journeys without waiting around on things as archaic as campaign spend tests.

Your father’s segmentation

Yesterday’s and, sadly, today’s often-used approaches to segmentation pose two challenges. First, many segments “factory-stamp” the segment definitions each user has been evaluated on based on whether they are in any currently defined segment. Once that evaluation takes place, all of the underlying data such as cart value, order quantity, product name or device type is stripped away. As a brand, it’s as if each interaction gets put on a one-way conveyor belt for evaluation. Of course, this is great for pre-identified, “set-and-forget” segments modeled off your prototypical customer journey. With the explosion of customer touchpoints, the customer journey is no longer linear—it’s changing and dynamic, and should be met with individual, relevant communications. Do you already know the optimal recency window and frequency at which customers convert with every specific site category? Do you know how this changes over time? If it changes, can you adapt quickly?

In many cases, you need to learn how to construct the best segments, so the factory stamp approach has to change. If you can’t go back and retroactively evaluate past user behavior, you have to start from scratch and build your segment as you see customers, making the learning process almost impossible to implement, or at best, a time-consuming and expensive trial-and-error process.

Secondly and most importantly, segments are only as valuable as they are actionable. Customer purchase propensity is highest within a couple hours of browsing a product, but if those segment snapshots are being shipped once a day or once every couple of hours to your media-buying platform, the lag translates to missed opportunities around your most valuable remarketing segments. What’s more, marketers must consider the loss in fidelity when syncing their users between a DMP and their DSP platform. This loss in reach—anywhere from 10 to 30 percent in some cases—combined with visibility only within cookie environments, is detrimental to performance and scale. Consider the high-value shoe shopper—how many times do you think you’ll buy media to show her the same shoes she’s already purchased in the following 24 hours when you could have been marketing complementary accessories?

What the world needs now: granular segmentation

Data sources are not consolidated by nature; they loop in from all directions and play on events, necessitating a consistent feedback loop so that marketers can analyze how their audiences engage with the media they send them. The ability to integrate your DSP with your DMP now makes it possible to leverage data sources from both platforms to aggregate, segment, analyze and distribute audiences in one central platform. This all but eliminates data loss and latency and provides a more seamless workflow.

With a full user history of events and behaviors at their fingertips, marketers can have a complete view of audiences as they define them. Gone are the days of having a user fall into a data segment and remain there until the segment is deleted. Marketers can now proactively build sequential audience segments, reliant upon creative sequencing with global reach and frequency capping.

Just as not all segments are created equal, not all site visitors are created equal. A site visitor that might be new to an advertiser site could fall into a “prospect” audience, and be shown a custom branding creative, based on the referring page URL. A past customer who spent more than $20 revisiting that site could be shown a sale promotion creative. Let’s assume that past customer visits that sale page and also sees two sale ads—that customer could then be shown a reminder creative.  Let’s say that same customer turns out to be our newest high-value accessory shopper, but doesn’t convert, leaving a $400 handbag in the cart—we can then show her a dynamic creative featuring that specific handbag model.  The relevance of the messaging to each site visitor, at each stage in the funnel, is precisely what can drive up to a 120 percent increase in ROI and a 50 percent reduction in CPA.

This is a huge boon for marketers. You can now enable immediate action on the most accurate segment definitions. Imagine setting real-time triggers so that, for example, as a prospect visits your site, they are immediately put into a granular remarketing segment. This means no more wasted spend on out-of-date segment associations and irrelevant conversations. You can now have fully scaled, complete audiences available immediately in media.

This is where the market is going—real-time segmentation without losing the scale in audience reach. Are you ready for the opportunity?