Book a demo

Book a meeting with one of our marketing consultants and see how we can help you improve your ad spend and increase your LTV.

Oops! Something went wrong while submitting the form.
arrow to left
January 19, 2022

This major beauty subscription brand is using predictive UA to amplify their FB campaigns

The Voyantis team

New year, new case study—and a particularly exciting one too!

I’m sure you're familiar with BoxyCharm (owned by Ipsy). It is the largest full size monthly beauty subscription box with over 1M+ subscribers in North America. The beauty items featured in their boxes vary each month, and includes an assortment of makeup, skincare, hair care items and beauty tools.

The company is just as big a beast behind-the-scenes in terms of their growth marketing efforts. They are one of the more sophisticated advertisers that we’ve come across, thanks to their robust data structure and strong growth team.

During the course of July 17-July 31, 2021, Voyantis’ predictive UA platform was able to help increase their ROAS by 30% (based on the 12 months view) of Facebook ads by optimizing acquisition campaigns. And it was all based on Voyantis’ prediction model.

Here are some more details on that!

How this campaign panned out

Let’s start with some background info. Although BoxyCharm is based on a subscription business model, there is a lot of variance in retention across subscribers. With that in mind, the BoxyCharm team wanted to conduct an LTV-based campaign, instead of only focusing on CPA. 

You may have seen our previous post about how the Pareto Principle can be used as a starting point for growth campaigns. That’s pretty much what we did in this case.

It was found that BoxyCharm’s highest 30% of LTV users generate 70% of the company revenues. Naturally, they wanted to bring more users just like that top 30%, at the expense of users with low LTV that are not profitable considering marketing acquisition cost. 

It’s a no-brainer. I mean, why pay the same CAC for low-LTV users, when you can just as well target users that demonstrate high LTV to further drive up revenue?!

So here’s what we did here at Voyantis.

We built a powerful model to predict LTV at the time of subscription, using customer data signals collected in the quiz acquisition funnel and BoxyCharm’s historical retention data. We went on to create a high LTV signal sent via Facebook CAPI and shifted investment from low LTV users to high LTV users on Facebook. 

For this video ad campaign, we decided to target females in the US, between the ages of 18-65, that demonstrate high LTV. 

Now here’s a quick rundown of the results, based on BoxyCharm’s internal reporting system:

  • CPA for new users was similar between the test and control cells
  • Retention (actual) of 4 months: 23%+ higher for the “high LTV” optimization cell
  • Retention (predicted) of 6 months: 35%+ higher for the “high LTV” optimization cell
  • Retention (predicted) of 12 months: 81%+ higher for the “high LTV” optimization cell
  • ROAS (12 months view): 30%+ higher for the “high LTV” optimization cell


If you would like to dive deep into the specifics of the campaign, have a look at the official case study posted on the Facebook website.

And the credit goes to: Predictive UA

It was amazing for us to have been able to help improve BoxyCharm’s ROAS by 30 percent! This win further established the importance of incorporating predictive UA into growth campaigns, especially for brands that bank on the loyalty of their customers.

Of course, there are a few other perks that go hand-in-hand with using predictive UA for growth campaigns. These include a significant reduction in churn, greater cohort retention, and a major reduction in UA-related costs over time.

Competition is fierce across all the ad networks, and now is the time for the growth leaders behind DTC brands to explore new ways to stay ahead of the UA curve, while getting the most bang for their buck PLUS significant long term returns. And predictive UA is the answer.

stay updated

join our newsletter

Request Sent. We'll be in touch!
Something went wrong. Please try again.

Continue Reading


Stop Retrospecting. Start Futurespecting.

Request Sent. We'll be in touch!
Something went wrong. Please try again.