The Predictive Marketer’s Dilemma: Customer Acquisition or Response Optimization?June 28, 2018
Modern marketers know that predictive analytics fuels growth. Research shows that over $100 billion will be spent on AI and predictive globally, across all industries, by 2025, but how can marketers be sure they’re maximizing their campaign success with the right predictive strategy?
What if your next best customer isn’t a ‘net-new’ prospect? It happens more often than you’d think. Knowing when you should acquire new customers instead of getting the most out of existing first-party data and segmenting existing customers into distinct groups can be difficult; unless you’re using predictive analytics.
Acquisition vs Optimization
Often, marketers have a hard time discerning whether they should use look-alike or response modeling to improve future campaigns.
The idea of expanding your existing prospect universe with look-alike modeling is extremely appealing, but what if it doesn’t need expanding? Sometimes, we get so fixated on net-new customer acquisition that we forget about the mountains of data that we already have.
Enter: Response Modeling.
Rather than spending more money to find new targets, marketers can use response modeling to analyze and score long-lapsed customers to identify who would be most- and least-likely to engage with a new campaign. With a predictive-scored list you can get a detailed view of your past customers ranked into ten tiles (deciles) by propensity to respond to your campaign. Response modeling gives you two strategic options: you can 1) export a list of your most-likely responders, or 2) select your least-likely responders and suppress them for improved targeting at scale for your future campaigns.
If a look-alike model proves that your customers aren’t as unique as you thought, your should use response modeling to dig through your existing leads.
Alternatively, if your customers are extremely unique, look-alike modeling can help you identify net-new people who are most-likely to become your next customer based on their shared attributes with your existing customers. Ultimately, the decision to use look-alike or response modeling will depend on your priorities:
- Reactivating old customers
- Expanding your prospect universe
- Growing your business
Knowing When to Scale Your Campaigns
Knowing when your campaigns are ready to scale can be difficult. To effectively scale your campaigns with predictive marketing you’ll want to start with a campaign that is just large enough to prove scalability. Starting small is recommended so that you can easily measure the profitability of your campaign and then continue to scale up or down from there.
Scaling campaigns isn’t luck, it’s a strategy. If you, for example, run a profitable predictive campaign with 50 thousand prospects, should at least double the number of prospects in your next iteration. Continue scaling your campaigns as long as they remain profitable, it’s when your campaigns are not profitable that you should reassess how to continue optimizing the campaign for the best results.
Refining Existing Programs with Segmentation
Segmentation is a powerful marketing tool, especially when used in tandem with response modeling. To make the most out of your existing database, you would want to create segments using your first-party data.
A non-profit might look into how long ago someone donated and how much they donated. Reactivating people who only donated five dollars ten years ago probably won’t maximize campaign ROI, but what about people who donated five hundred dollars seven years ago? Or who donated two hundred dollars three years ago? These are the people the non-profit wants to target.
Sticking with the example above, you wouldn’t want to upload your entire donor database into a model. You certainly could, but you run the risk of targeting those one-time five dollar donors. This strategy, of uploading your entire database will point you in the right direction, but it isn’t optimal. When you load in too many customers, you end up watering down their uniqueness. Predictive modeling is most effective when you look at really unique groups of people, that’s why it’s important to try and identify sub-segments of your database who are going to be different naturally.
Analyzing Your Results
For small businesses who don’t have a huge customer base, or aren’t generating many prospects on their own, sometimes finding net-new people with look-alike modeling is the right idea, but how do you know which type of modeling – response or look-alike – will help you reach your goals?
Setting specific campaign goals is the biggest step that you can take toward achieving successful campaign ROI with predictive marketing because it makes it easier to determine how much you benefited from the campaign.
Whether you’re using response or look-alike modeling, defining which actions you want consumers to take will make it much easier to determine whether your campaign was a success or if it needs to be reworked.
Which Strategy Is Right For You?
With so many options available to marketers, it’s easy to get wrapped up in wanting to find net-new prospects, but sometimes your best prospects are people you’ve already reached out to. Now more than ever, it is important that marketers understand when they should be optimizing their existing customer lists instead of focusing solely on expanding their prospect universe.
The best way to keep your company a step ahead is to make sure that your marketing team has a winning predictive strategy. The Reach predictive cloud offers brand and agency marketers the ability to create predictive profiles of customers and prospects without needing to wrangle data or involve large teams of data scientists.