Beyond the Search Secret Sauce: More Cooks in the Kitchen

In my recent blog series, we took a deep dive into the secret sauce of Search. We ventured from the basics of offline data ingestion to a fictional activation user story. We even teased out five different avenues from which paid search teams can regain a competitive edge within a crowded auction environment. The common thread centered around data, and the novel ways we can put it to work towards improving ROI and business-side margins.

Today, we’re going to revisit that series, but in a broader context that isn’t as deeply rooted with the Search Ads 360 (SA360) endpoint in mind. We’ll explore some of the other endpoints for a first-party data model, as well as the projects we can undertake with that data. As we explore, we’ll hear from a couple of my colleagues who also help our clients implement durable first-party data strategies using the Google Marketing Platform (GMP). Let’s dive in!

Predictive Lifetime Value (pLTV) Modeling

"We refer to Predictive Lifetime Value modeling as ‘pLTV’, but it’s commonly referred to as ‘CLV’ (Customer Lifetime Value)."

Pat Grady | Adswerve Solutions Engineer

The premise around pLTV is that brands obtain data on prospective and current customers through their interactions both on- and off-site which, with proper analysis, could help identify and predict how valuable those customers will be over time. A customer who makes a large purchase every year or so is valuable, but those that purchase more routinely over the same period, even with smaller basket sizes, demonstrate a different degree of engagement and brand affinity. With these conclusions we can then market to each group differently manifesting in different ad copy, calls to action and/or budgets. This practice is known as Value-Based Bidding (VBB).

“We refer to Predictive Lifetime Value modeling as ‘pLTV’, but it’s commonly referred to as ‘CLV’ (Customer Lifetime Value). This is a customer-specific score that’s assigned using a model trained on historical data.” explains Pat Grady, my esteemed colleague and solutions engineer at Adswerve. “These scores help brands better align marketing investment with expected outcomes.”

Creating a pLTV model using tools like Google Cloud Platform and Big Query and activating it through bid automation in SA360 can yield margin lift while optimizing the budget efficiency. Pat further explains that “models can make decisions at the microsecond level and are able to optimize decisions at a scale that marketers cannot achieve on their own. The transition from columnar-level aggregates to row-level assessments means that we transition from analyzing ‘populations’ to ‘people’. There is power in that evolution.”

Adswerve’s experience in this area has shown that undertaking these types of projects not only leads to margin lift, but also more efficient media execution in general, which leads to incredible business and customer outcomes.

Propensity Modelling

Pat describes propensity modeling as “when we train a model to predict the likelihood that someone will accomplish a certain task. These tasks are typically defined as a ‘conversion’, however, they could include the likelihood to ‘churn’ or other actions.”

Understanding a customer’s likelihood to churn out-of-market or convert, or something that’s less differentiated such as visiting an in-store location or buying online, are additional factors that marketers can use to help inform SA360’s bidder on how to bid to a particular user’s query.

During my informal interview with Pat over Slack, he explained further that “propensity models give us the ability to do user-level scoring and analysis that isn't otherwise possible. The machine learning aspects of these techniques give us incredible insight into the problems, and help guide us to prescriptive solutions and increasingly optimized media budgets and performance.”

The end result means less waste (from bidding on customers who we expect to take undesirable actions) and higher margins as we hone in on customers who demonstrate intent, as well as the ability to activate these signals in near real-time using SA360 and Auction Time Bidding.

Conversion Segmentation Redux

"If your ideal conversion segment is nestled in the Floodlight as a custom floodlight variable, then you can create a Custom Bidding algorithm in DV360..."

Mary Kotara | Adswerve Sr. Programmatic Consultant

What I discussed in part one and part two of our recent blog series is equally applicable to Display and Video 360 (DV360), which is the Programmatic buying arm of the GMP. DV360 can utilize the Floodlight and CFV data in a manner that’s objectively similar to SA360. We can use them to segment conversion data from a larger aggregate in order to hone in on a specific CFV value, and bid to those values differently.

“If your ideal conversion segment is nestled in the Floodlight as a custom floodlight variable, then you can create a Custom Bidding algorithm in DV360 to have the system automatically optimize to impressions based on the likelihood they will result in that specific variable being obtained,” describes Mary Kotara, awesome human and senior programmatic consultant at Adswerve. “The outcome to such an exercise is greater optimization capability and stronger business outcomes through DV360’s machine learning.”

Underscoring this capability is how DV360 and SA360 can share the same Floodlight pixels, so long as each of those platforms are connected to the same CM360 advertiser. This means that any conversion segmentation exercise employed in one platform has an activation equivalent in the other, doubling the capability and ROI potential across search and display.

DV360 has an ace up its sleeve and can take this exercise a step further, allowing the creation of audiences based on specific CFV values. “Narrow your targeting to only serve ads to users in certain custom variables for better targeting,” describes Mary, who also adds “pro tip: align your creatives with this audience.” CFVs then not only allow for conversion segmentation in DV360 but can also be used to devise an audience and creative strategy as well.

As a practical example, perhaps we have a campaign in DV360 and we want to target our most valued customer segments with a special customer appreciation offer code. We can specify our audience targeting to include users who’ve been placed in our “high value” customer segment based on our pLTV model, and show them a unique offer code within the creative or send them to a unique landing page.

The ability to pivot audiences off of CFV values within DV360 is yet another activation by-product of a 1P data and modeling project and is truly an impressive capability to wield.

Seeing the 1P Data Through the Trees

I hope this blog has helped paint a Bob Ross-ian image in your mind of the incredible potential that lies within 1P data from both an analysis and activation perspective. Its usage is becoming normalized as the primary means for adapting to declining usage and consumer sentiments around privacy.

The crux of the 1P data story is how this data can be made available to media activation teams. For brands, this requires an internal analysis of customer segmentation and profitability backing into KPIs and an ROI strategy delivered by paid media. There’s an added relationship layer in the agency context, as these organizations need to deliver thought leadership to their advertiser clients and to help strategize and activate brand-owned first-party data.

Adswerve can and has acted as sherpa in both brand and agency contexts. We believe the key to customer delight is when media and data can be used together to deliver incredible outcomes. If your organization is interested in how to deliver or deploy 1P data for media activation, please contact us to learn how we can partner together to achieve your goals.