Using Lookalike Models to Acquire High-Value Customers, Part 2

If you missed our previous post, go back to learn about how using your first-party data, smart segmenting, and predictive analytics will vastly improve your marketing campaigns when paired with the power of lookalike modeling in channels like Facebook, Google, LiveRamp, and others.

In this post, we’ll continue to guide you through best practices of using your own data to inform the best possible — i.e., most valuable and most likely to convert — lookalike audiences across the web.

 

Step 3: It’s Not Alchemy. It’s Lookalike Models

Generally speaking, the underlying mechanics of lookalike prospecting are quite similar regardless of which modeling tool—or combination of tools—you choose.

In broad strokes, you’d give Facebook, Google, or LiveRamp a seed audience derived from analyses of your first-party data. Then the channel partner examines the past behavior of the seed audience on its own platform and gives you a new, broader audience that has exhibited similar behaviors to the seed group.

In addition to investing in the infrastructure necessary to produce more accurate, more complete first-party data, you can differentiate your lookalike prospecting efforts by improving the strategic distribution of your seed audiences.

Imagine, for example, that you sell two distinct product lines—one catering to younger customers and the other to middle-aged customers.

For the first line, your highest-CLV customers are almost all Millennials. For the second, your highest-CLV customers are almost all Gen X-ers.

Knowing that Facebook is now a platform not-so-much used by the younger set, you’d be ill-advised to funnel a seed audience optimized for the first product line into Custom Audiences. Your highest-CLV customers for that line are just not on Facebook in large enough numbers to justify the investment.

Instead, you’d be better off sending a seed audience optimized for that Gen X-oriented second product line to Facebook, while ferrying your Millennial seed audience along to a more youth-oriented platform—Instagram, perhaps.

When executed properly, such strategic lookalike prospecting can exponentially improve your return on ad spend (ROAS).

 

Step 4: Use Predictive Analytics to Get There Faster

Time is arguably the greatest roadblock to continue executing this process at scale.

Even if you approach CLV in one-year cycles, you’ll need to make adjustments to your marketing far sooner than a year from now, i.e., sooner than you’ll have access to concrete data on the lifetime spending of the new customers you acquire through lookalike prospecting.

As advanced as it’s become, lookalike prospecting is still a probabilistic science. There’s never any guarantee that a new customer who looks like a proven high-CLV customer on paper will, in fact, develop into a high-spending repeat buyer.

This is where leaning on predictive analytics shortens the gap between now and a lifetime from now. By evaluating a new customer’s first few purchases, the acquisition channels that led them to those purchases, and any relevant demographic information, a solid predictive model can forecast a given customer’s likeliest trajectory with the brand with tremendous accuracy.

In other words, predictive analytics gives you a way to conduct initial performance checks on your lookalike prospecting. Instead of having to wait a month or six months or even a year before you can pass judgment on a given acquisition campaign, predictive analytics can reveal which keywords, channels, and seed groups are actually generating new high-CLV customers in near-real time.

Faster customer insight equals faster decision-making, which lets you tweak ad spend on a rolling basis instead of waiting for a campaign to play out in full.

Step 5: Nurture High Potential

While the details of this nurturing may vary, predictive analytics-driven identification is just the first step in developing high-CLV customers.

After you home in on a set of new(er) customers who are particularly likely to evolve into high-value repeat buyers, you still have to engage these customers in a relevant, meaningful way.

This might involve extending to your potential future VIPs the perks that you only offer to your current long-term VIPs. For instance, many retailers bump their top customers to the front of their customer service queues (whether digital, telephonic, or physical) and ensure a senior member of the customer service team is assigned to handle these VIPs’ requests. Others periodically send customized, handwritten thank you notes to their highest-value customers, and back up this show of appreciation by offering VIPs early access to new collections or free shipping and returns on all online orders.

Regardless of the approach you take, your brand should always be striving to make its most important customers feel special. This holds as true for customers who have proven their importance time and again as for customers whom predictive analytics algorithms have identified as the next wave of your most important buyers.

At the end of the day, high potential CLV means nothing if retailers don’t put in the requisite legwork to nurture potential into purchases.

Step 6: Use Predictive Analytics to Reach a Larger Lookalike Audience

Predictive analytics aren’t only valuable during the later stages of the high-CLV customer acquisition process. They can also help retailers maximize the value of their first-party data during seed group identification

Imagine a retailer that’s on the brink of launching a new line of winter wear, featuring a high-priced down coat as the collection’s flagship product. In the absence of a predictive-analytics tool, the retailer would comb through its first-party data and pick out individuals who had previously purchased a winter coat and gone on to become (or continued to be) a high-CLV customer. These criteria might produce a seed audience of, say, 10,000 customers.

But with the help of a predictive solution’s pattern-matching capabilities and algorithmic analyses, the retailer might be able to locate 70,000 additional people within its customer base who have not purchased a coat but share many behavioral and demographic traits with high-CLV customers who have.

Thus, instead of sending Facebook Custom Audiences or Google Customer Match a seed audience of 10,000, the retailer can send an audience of up to 80,000, exponentially increasing its potential reach.

 

In the meantime, you can learn more about the absolute most important customer-centric KPI (spoiler alert) in our latest book, The Chance of a Lifetime: How to Use Customer Lifetime Value Reporting to Grow Your Retail Business.

 

Like this? You might also enjoy these.

How Supergoop Topped Custora’s Retail Benchmarks, part 1

This past May, Custora announced a new benchmarking feature within our...
Read

Your Revenue Forecasting Ignores the Source of Your Revenue

If you’re like most organizations, your finance department focuses on overall...
Read

4 Steps to Solving the One-Time Buyer Problem, Part III

In parts one and two of this series, we laid the groundwork on a methodology to...
Read