Cardinal Path’s response to COVID-19 Cardinal Path is sharing all we know to help marketers during COVID-19.  Learn more.

Businesses in all industries and of all sizes have had to scramble to deal with the effects of COVID-19. And as our 2020 Online Behavior Dashboard shows more specifically, many businesses are facing the double-whammy of less online traffic and lower engagement. Based on our data, these businesses are particularly likely to be in travel, hospitality, and food & beverage.

Access the Free 2020 Online Behavior Dashboard

Faced with this two-pronged challenge, businesses will be looking to scale up their traffic and their pool of potential customers but will want to find a way to do this efficiently. After all, it’s rare that businesses of any kind are interested in simply ramping up digital traffic for its own sake. While this is a serious undertaking, the good news is that there are a variety of tools & techniques that can help businesses successfully meet the challenge. In this post, we lay out a spectrum of possibilities for businesses to innovate their way out of a COVID slump by ramping up traffic and efficiently converting visitors into leads and customers.

Reduce Customer Churn with Predictive Modeling

It’s hard to fill a leaky bucket. For organizations looking to scale their customer base, start by taking proactive steps to prevent additional customer churn. To do this, it’s helpful to know which of your customers are most likely to churn, so that you can draw up customized plans to retain their business.

Build a predictive model for customer churn starting with an inventory of available data. You’ll want historical records (more is better) of customers who have or have not churned. In addition, you’ll want data about those customers: website behavioral data, purchasing history, in-store records (if applicable), and so on.

Fundamentally, this modeling exercise is about using your historical data to predict which of your current customers are most at risk of abandoning you. A common output of a model like this is a classification of your customers into cohorts — think deciles, quintiles, etc. When you’ve grouped your customers from most to least likely to churn, you can then design customized “interventions” to take with at-risk customers.

For example, at-risk customers might receive more personalized outreach. This could involve customized email campaigns, personalized website landing pages, and much more. Discounts and other incentives may work to build loyalty with the customer and reduce the risk of churn — and those incentives can be scaled to fit the urgency of the need.

How does this kind of churn modeling work? This is an example of using machine learning for classification rather than regression, because the outcome we’re trying to predict is binary (will they churn or not?) vs. numerical (e.g.: how many purchases will they make in the future?). 

Here are the puzzle pieces you’ll need to fit together in order to be able to bring this to life:

  • Relevant data sources
  • Move and merge relevant data in a cloud environment (e.g. Google Cloud Platform, Microsoft Azure, Amazon Redshift)
  • Develop and run a churn prediction model on top of the unified dataset
  • Ability to take action on the results

That last point bears further discussion. All organizations that successfully develop and run a churn prediction model will have some capacity to take action on the results, but the specifics of what this looks like may vary widely from organization to organization. For example, assuming you have email addresses associated with your customer records, you’ll be able to send customized emails to groups of customers deemed to be at risk. Similarly, if you can tie your customer records to Google Analytics or Adobe Analytics IDs, you’ll be able to deliver a personalized experience on your website, via tools like Google Optimize or Adobe Target.

Regardless of exactly how your organization will be able to action the results of a churn model, you’ll want a firm foundation to build on. Investing in growing your customer base will generate a much stronger return if you can also slow the rate at which existing customers churn out. So, having plugged the leak in our bucket, let’s look at some ways to fill it!

Lookalike Modeling

One way to scale your customer base is to go for the “spray and pray” approach. Put a lot of dollars into the top of the funnel and hope that a lot of leads/customers come out the bottom. Luckily, there are smarter, more cost-effective ways to invest in growth. One such way is to use “lookalike modeling” to expand your reach to new people who resemble your existing customers, high-value leads, engaged audiences, and so on.

Although various technology platforms refer to the concept of “Lookalike Modeling” by a variety of names, the underlying concept is quite simple. Who are your most valuable customers? What traits do they have in common? Using this information, we then attempt to find new people who share those traits. Here’s an example of how to execute this tactic using the Google marketing “stack.”

If you’re using Google Analytics to measure your website, create a segment of people that you believe are your best customers. This could be based on their purchase history, lifetime value, content consumption, visit frequency, or whatever other criteria you think are most relevant. Regardless, creating a segment like this is a point-and-click exercise that should only take minutes.

Once you’ve got a segment created in Google Analytics, you can share that segment — or “audience” — with other Google tools, such as Google Ads. Note that you’ll need to link your Google Analytics account to your Google Ads account, but that’s also a very easy process.

Once you’ve shared an audience from Google Analytics to Google Ads, log in to your Google Ads account. You’ll now be able to create a campaign that uses “Similar Audiences” targeting (Google’s name for Lookalike Modeling) to reach new people via Google search ads and the Google Display Network. Keep in mind that because you’ll be reaching a fairly specific type of person, you may want to customize your messaging/creative/offers.

Google isn’t the only platform that offers this kind of targeting; many other ad platforms do as well. Regardless of which paid media platforms your business uses, keep in mind that targeting “lookalike” audiences isn’t the best way to simply maximize your reach. As a reminder, that’s not the goal of this tactic. Given that, measure the success of lookalike targeting not so much by the sheer volume of impressions, clicks, likes, site visits, etc. that you receive, but by the efficiency with which you can convert these lookalike audiences into actual customers. Efficient, profitable growth is the name of the game.

Testing & Personalization

So you’ve taken steps to reduce customer churn and started to target growth by reaching new people who are similar to your existing customers. There’s still more that you can do to ensure that you’re driving growth efficiently, though. If you don’t have a robust program in place for testing and personalization, it’s more important than ever to get started. Done well, a testing and personalization program drives a “virtuous cycle” in which conversion rates are pushed ever higher as you learn more and more about how to optimize the experiences you’re giving your potential customers.

Testing websites and mobile apps — whether via simple A/B tests, or more complex multivariate experiments — is nothing new. The technology has been around for well over a decade, but it’s easier than ever to use, meaning that it’s no longer only a tool for deep-pocketed enterprises, or digitally-savvy organizations to use. Literally any company can use testing, and its cousin, personalization, to improve conversion rates and make your marketing dollars work harder for you.

So what does it take to get started? First, of course, you’ll need the tech. There are many testing platforms to choose from, running the gamut from enterprise-ready solutions to free options that even small businesses can get started with right away. Google’s Optimize product is free (there’s also a paid version, Optimize 360), while Adobe’s Target product is popular with larger enterprises. Regardless of the tool you’re using, though, the technology by itself won’t get you anywhere.

You’ll also need to be able to bring a few different skill sets to bear. Sometimes you can find one person with all the necessary skills, but more likely you’ll want to convene a team with expertise in:

  • Strategy: what should we be prioritizing for our testing efforts, and why? 
  • Data analysis: what does our data tell us about where we can optimize our site or app?
  • Development: what would it take, from a technical perspective, to make the changes that we want to test?
  • Creative: what should our test experiences look like?

For each of these categories, there’s a spectrum in terms of the depth of expertise that’s truly required. For example, can you get started with testing without having a clear strategy laid out beforehand? Of course. Sooner or later, though, the “quick wins” dry up and having a coherent strategy becomes more important.

Similarly, if you’re just starting out, it’s likely you’ll be deploying simpler tests — think “red button vs. blue button” type tests. In those cases, you might not need any development resources at all. Once you’re ready to start deploying more complex tests, though, having development resources becomes a requirement.

Finally, high-performing testing programs typically share a few key attributes:

  • Commitment from executives: while some organizations organically drive testing and optimization “from the bottom up,” most need a level of sponsorship from the executive level to ensure momentum is built and maintained.
  • Clear processes: there’s no need to get bogged down with detailed processes for organizations who are just getting started. However, as a testing program matures, clear processes — meaning documented steps from conception to launch, each with defined owners — help ensure that the return on the time invested in the program remains strong.
  • Roles & responsibilities: as we noted above, a few different skill sets need to be brought into concert to make a testing program successful. As such, clear definitions of roles & responsibilities help your team avoid duplication of effort, confusion about ownership, and the frustrations that naturally stem from those kinds of issues.


If your business is struggling with downturns in traffic and engagement, there are a variety of steps you can take to turn your fortunes around.

First, consider taking proactive steps to stem any customer loss you may be experiencing. “Churn modeling” is a method by which you can predict which of your customers are most prone to abandoning you, so that you can target them with customized outreach and offers in an effort to retain them. Second, using “Lookalike Modeling” can help you cast a wider net in search of new customers, without resorting to “spray and pray” tactics. Finally, regardless of how you’re driving traffic to your sites, building out an optimization program will help you make consistent, incremental progress in converting more of your traffic into paying customers. Combined, these tactics should help you drive efficient, profitable growth during this critical time.


State of Digital Marketing Analytics

The 2020 State of Digital Marketing Analytics examines the marketing technology that supports the world's most successful enterprises and highlights the challenges and strategies for navigating the new normal..