While there are many ways that a marketer can leverage Bluecore Predictive Audiences to build various campaigns, here are five examples that have resonated with a number of clients and are a good starting point for using Predictive Audiences.


  • Marketer Quandary: Sending promo codes to all customers will diminish margins on sales.

  • Bluecore Solution: Segment customers based on Discount Preference and only send sale products, coupons, etc. to your customers who are "Discount Buyers".

Discount Preference predicts the customer’s preference to buy products that are full price or discounted, and to what extent.  Typical offer affinity calculations may account for past purchase behavior but completely ignore past browse behavior. This approach is limited because it does not extend to non-buyers, and can misrepresent preference when a customer has limited purchase data. To overcome these limitations, Bluecore computes a preference score using both browse and purchase data as a weighted average between the browse discount preference and the purchase discount preference. The relative weighting in the average is designed so that it assigns more weight to the purchase discount preference with each additional purchase the customer makes. This way, marketers can distinguish between full price and discount buyers, and can use this information to offer a discount to discount buyers, and no discount to full price buyers. This has two major benefits for a marketer:

  • Don’t give away margin when you don’t need to. Giving a discount to a full price buyer is like giving money away.

  • Don’t discount your brand to your full price buyers. The more someone sees discounts from you, the less likely they are to ever buy from you at full price

Keep sending discount messaging to your “Discount Buyers” since these customers tend to buy only when there is a discount, but send standard messaging to your “Full Price Buyers” since these customers don’t care about waiting for deals and will pay top dollar anyway. This maximizes your margins. This type of campaign can be executed on multiple channels: email, Facebook, or Google AdWords.


  • Marketer Quandary: Sending subsequent touches to all customers without clear indication of engagement or sending the same message to all customers on a subsequent touch.

  • Bluecore Solution: Version messaging based on predictive filters or add touches to existing campaigns based on predicted customer behaviors.

Bluecore’s Likelihood to Take Action filters predict the likelihood that a customer will Unsubscribe from Emails, Open an Email, Click an Email, or Convert (i.e., make a purchase in the next 14 days).

For example, if you have an existing Product Abandonment campaign with one touch and you’re considering whether to send a second touch, here’s what you can do:

Re-send Product Abandonment Touch 1 to customers who have received Touch 1, didn’t open Touch 1, but have a medium or high Likelihood to Open emails. This way, you’re essentially sending the customer a second touch but, to them, it’s still a first touch since they never opened the first email.


  • Marketer Quandary: Using rule-based approach to save at risk/lost customers is arbitrary and does not take into account an individual customer’s unique buying behavior. How do you know when a customer is truly “at-risk” or “lost” and to reach out with the appropriate messaging to win them back? 

  • Bluecore Solution: Reactive at-risk and/or lost customers where delivery of subsequent touches is determined by predictive filters or replace an existing win-back/reactivation program with a Bluecore Predictive Audience of At-Risk/Lost customers. Bluecore predictive filters take into account each customer’s unique buying cycle.

To create an audience for all “at-risk” customers for the purpose of a batched send or display ad:

To turn this into a trigger and send a message to customers when they become at-risk, select the Became an At-Risk Buyer checkbox and add the number of days this audience should account for. Consider the following when sending messages / creatives:

  • Winback messaging like “We miss you!” or “It’s been too long”

  • Promotional messaging including coupons, promo codes, and/or free shipping offers


  • Marketer Quandary: Which of my customers will have the potential to be valuable?

  • Bluecore Solution: Filter for the top 10% of customers with the highest predicted lifetime value and send them a loyalty discount or free shipping code.

Identify who your current high value customers and boost ROI on CRM-targeted campaigns by cutting out low value customers and targeting high value customers. Bluecore’s Predicted CLV model isn’t dependent on a prior purchase since Bluecore’s calculations are forward looking rather than relying solely on historical data. You can leverage Bluecore’s ability to predict an online customer’s projected spend with their brand over the next two years and target those top customers through email (to nurture existing top spenders) or through building Facebook lookalikes (to acquire new customers). In the screenshot above, we’re targeting the top 10% of customers by their projected spend.  But we can easily change that to top 5%, bottom 20%, between 15% and 55%, or whatever ranges you’d like. Consider the following when sending messages / creatives:

  • What type of message do you want to convey to your top spenders?

  • Do you have a loyalty or VIP program, or are you thinking of starting one? Your high PCLV audience is a great group to which to promote this, since you will nurture existing loyal customers and can convert projected spenders into actual spenders.

You can also use Predicted CLV filter to acquire new customers by building Facebook lookalikes and messaging to those customer via Facebook ads. Seeding Facebook with a targeted audience of your projected top spenders will allow Facebook to find new customers that are similar to your projected top spenders rather than to your general customer list. Then, when you message these lookalikes with Facebook ads, you can expect to see a higher return. To get the seed audience from Bluecore to Facebook, simply export the high predicted CLV audience (like the one in the screenshot above) to Facebook by clicking on the Platforms tab in the Audience Builder and Enable the sync to Facebook. For more information, see Integrate with Facebook.


  • Marketer Quandary: What products will my customers want to buy because they just bought something else?

  • Bluecore Solution: Category Preference predicts the customer’s preference for specific product categories in your catalog.

Bluecore associates behavioral data– beyond just past purchase and view data– to email addresses instead of anonymous pixels, so we can target customers consistently across marketing channels. Our category affinity models include co-view, co-purchase, and co-cart data to discover relationships between products and categories that may not be immediately obvious. Co-view, co-purchase, and co-cart are collaborative filtering models (i.e., customers who viewed this also tend to view that). For example, if someone buys a TV, we’ll know they have a certain level of affinity to a sound system. Use the Add Category Filter dropdown to select the categories by which customers should be filtered. Then use the slider to select a preference level for the selected category. For instance, the screenshot below shows us customers who have at least a “High Preference” to the category “footwear”. We can adjust both the category we want to find customer preference for AND the strength of their preference.

Category Preference can also be used to build Facebook campaigns. It’s great for campaigns targeting a specific product, brand, category, or breadcrumb. For example: a new product launch, a promotion on a specific brand, or an event around a category. Category Preference is a more dynamic option than just using past purchasers or past viewers to decide how to target. Best practices to consider:

  • The creative of your Facebook campaign should match the attribute value you chose to create the audience (e.g., the specific brand, category, etc.).

  • Depending on the type of category, you should consider whether to include or exclude past purchasers. Past purchasers will be included in the Category Preference audience by default, as they will likely have very high preferences for the attribute you used as an input. However, if you are looking for existing contacts that have a preference for a product or category that is rarely purchased more than once, like a refrigerator or a mattress, you should consider excluding past purchasers from the audience to avoid targeting them with your Facebook campaign. Exclude past purchasers by using the Customer Behaviors filter within the Audience Builder.

  • If you find your campaign is performing very well, you can consider lowering the level of preference to broaden the audience and increase the reach of the campaign. If your campaign is not performing as well as you would like, you should raise the level of preference and/or consider layering in the Predictive CLV filter to remove low CLV contacts from the audience.

Did this answer your question?