By continuing to use this site you agree to our Cookies Policy.
Interviews

3 Predictions about Facebook’s Thanos-Level Ad Algo

June 18, 2021
Mae Rice

Now that iOS 14.5 has launched, Facebook's ad targeting algorithm has less data to train on. How will that impact its performance? Surge AI founder and data scientist Ferris Jumah gave us his predictions.

Table of Contents

This is an excerpt from MarketerHire's weekly newsletter, Raisin Bread. To get a tasty marketing snack in your inbox every week, subscribe here.

Media buyers used to love Facebook’s algorithm. 

Before iOS14.5 rolled out in April, they often told us that buying pure reach on Facebook was just as efficient as manual targeting.

Now, Facebook’s ad targeting algo has a LOT less real-time data to learn from.

Most users with iOS14.5 — about 95%, in the US — haven’t opted into tracking, which means Facebook can only log their behavior on Facebook apps.

What will that do to the algorithm’s efficiency? 

We asked Ferris Jumah, a former LinkedIn data scientist and co-founder of real-time market research startup Surge AI, for his take as an informed observer. 

1. It will still be able to target based on interests — for a price. 

Facebook’s ad placement algorithm is “an interest-based machine,” Jumah said. 

Facebook has always collected first-party data on the pages and posts people like. (The like button is almost a second Facebook logo.)

Even in the privacy-first web, Facebook’s algorithm will give marketers a (robotic) hand with psychographic targeting, learning which interests correlate with an interest in your product. 

The downside: “It will be way more costly” (and time-consuming) to use the algorithm, Jumah predicted. 

2. It won’t be able to optimize for conversions the way it used to. 

In the old days — so, March — Facebook got insight into what drove sales and what didn’t with its tracking pixel.

But now, “the pixel’s going to be a lot less effective,” Jumah said.

Facebook will still have some sales data from people who opt into tracking or buy through Facebook Shops, but not as much as it used to. 

Its algorithm can likely still learn to target people who CLICK on your ads, but it may struggle to target people who actually BUY, as opposed to people who bounce. 

3. Facebook’s backlog of historical data won’t help much.

“The internet moves super fast,” Jumah said — and that’s especially true for digital marketing

“Information on how to do good Facebook and Google marketing from two years ago is now basically irrelevant,” he said. 

Even if the algorithm deeply understands why everyone bought what they bought in 2020 — that’s not as predictive as what they’re buying now.

Our takeaway? 

All digital ad platforms’ targeting algorithms will take a hit as the privacy-first web unfolds.

But with first-party data from nearly 4 billion monthly active users across Facebook and Instagram, Facebook’s algorithm will likely stay best-in-class — especially for interest-based targeting. 

"It's like Thanos lost a couple infinity stones,” Jumah joked. “Still powerful, but less omnipotent."

Mae Rice
about the author

Mae Rice is senior editor at MarketerHire. A long-time content marketer, she enjoys interviewing marketing experts and learning about the weird and wonderful feedback loops that connect marketing and culture. She also loves any ad that makes her say "Huh?!" — especially Tim and Eric's ad for Totino's pizza rolls, which she watches way more often than you would think.

Hire a Marketer
Join MarketerHire Today
We'll match you with a perfect expert.