Abacus Weekly Experiment

At Abacus we see ourselves as mad scientists always experimenting to drive the best campaign results for our customers. This week we decided to conduct an experiment to help advertisers drive a more efficient return on ad spend and measure their advertising the same way they measure their business.
Recently, Facebook introduced a new optimization method called Value Optimization. New optimization method relies on deep learnings and an adaptive bidding algorithm which helps increase your Return on Ad Spend (ROAS) per user, not just the volume of purchases.
We decided to test the new optimization method in a series of split tests to determine if:

  • Value optimization drives higher ROAS
  • Value optimization significantly lowers the purchase volume while being more “selective” of the total targeted user pool (not every buyer’s purchase value is the same)
  • Different targeting options (such as interest, lookalike audiences, retargeting audiences) will impact the optimization


Hypothesis #1

Our first hypothesis was that the Value Optimization method could drive a higher value of purchases compared to Event Optimization (standard “Purchase” event), but could significantly impact the volume of purchases at the same time. To verify it, we would compare both the total count of purchases as well as total revenue generated.

Hypothesis #2

Our second hypothesis was that the Value Optimization could provide inconsistent results depending on the source of the targeted audience (Lookalike vs. Remarketing list based on pixel or CRM file).
Before we dig deeper, can you guess if our hypothesis was correct?

Split Test #1:

In our split test #1, we targeted the broadest audience possible, unrestricted by interests, behavior or additional layers of custom/lookalike audiences.
Here’s what our setup looked like:

  • Geo: GTA Area (11 cities with 10+ mile radius)
  • Gender: Male and Female
  • Age: 25+
  • Exclusions: previous purchasers and web visitors in the past 30 days
  • Total Audience size: 3.3 Million People
  • Budget: $3,000 total budget split evenly across 2 Ad Sets
  • Placements: All Placements, excluding Instagram
  • Time: January 23-30, 2018
  • Creative: A single link ad creative was used in both Ad Sets
    • Ad Set A- optimized for Conversions
    • Ad Set B- optimized for Value



After running this single ad split test for a full week, we were able to confirm that Value Optimization outperformed Purchase Event Optimization showing:

  • Higher volume of purchases (23 vs. 9)
  • Lower Cost per purchase ($55 vs. $138)
  • Double the Return on Ad Spend (2.22 vs. 1.08)

What’s even more surprising, the value optimized ad set delivered these impressive returns while reaching 40% less people. Talk about media spend efficiencies!

a laboratory

Split Test #2:

In phase two of the experiment, we launched a similar Value vs. Purchase Event split test, but this time we tweaked our targeting audience to include a couple of highly valuable re-targeting lists based on FB conversion pixel.
Here is what our setup looked like:

  • Geo: GTA Area (11 cities with 10+ mile radius)
  • Gender: Male and Female
  • Age: 18+
  • Inclusions: Users who reached delivery stage on the website -or- Top 10% of users by time spent on the website in the last 180 days
  • Exclusions: previous purchasers in the past 30 days
  • Total Audience size: 6,300 people
  • Budget: $800 total budget split evenly across 2 ad sets
  • Placements: all FB placements only
  • Creative: A single link ad creative was run in both ad sets.



Once again, the Value Optimized ad sets produced stronger results and came as a winner by delivering:

  • Higher volume of purchases (43 vs. 39)
  • Lower Cost per purchase ($9.25 vs. $10.25)
  • 36% Higher Return on Ad Spend (16.61 vs. 12.16)
an image of a facebook interface


Our team understands that there is a lot of value in bidding for Value versus standard Purchase Event. For example, if we are working on a campaign for a large e-commerce store with hundreds (or maybe thousands) of weekly orders, then we can help to leverage Facebook’s new bidding option to drive high ROAS purchases.
Our practice is to focus your early stage acquisition strategy on driving the sheer volume (i.e.: bid for Purchase Events) to feed the Facebook algorithm with learnings about your typical buyers, including their average Order Values.
The Facebook algorithm needs to first understand who your buyers are before it can take those learnings and really produce the best results.
Once we start reaching a minimum threshold of 50+ purchases per week (think of it as a bare minimum), then we recommend switching to Value bidding to increase your ROAS. As our experiments showed, this could be achieved without hurting the sales volume while improving media spend efficiencies.
Have questions about our testing methodology? Reach out to Abacus via info @ abacus .agency

About Abacus

Founded in 2016 and based in Toronto, Canada, Abacus is a performance marketing company with a focus on Facebook and Instagram advertising. We’re decentralized, which means our clients are all around the world, we help everyone from startups to Fortune 500 companies.
Our team includes expert data scientists that build growth engines, not just campaigns. By utilizing the latest technology and unearthing patterns in data, we are able to help companies achieve their advertising goals while driving down costs.

Abacus Agency
Abacus Agency

Table of Contents

Share this.
Share on twitter
Share on linkedin
Share on pocket