Issue 3: The Hidden Reason Your Facebook Ads Are Broken

It’s easy to get dizzy from the amount of AI news coming out every day. Jaw-dropping images. Copy that inspires. Weird and wild new use cases. The progress feels exponential.

All that said, the outputs are a reflection of the inputs. Not the prompts, but the underlying data that feeds large language models. There’s ongoing work to make the models better by filling in the gaps in knowledge for tools like ChatGPT and its latest iterations.

But there’s another model missing data that’s critical for your e-commerce business: Facebook’s bid algorithm. Privacy updates to browsers and mobile devices are unfortunately depriving the algorithm of information about your users that can help find your future customers at the best CPA.

So we decided to get to the bottom of this. In this newsletter you’ll read the first chapter of the story and we hope you’ll click through to finish it with us to see Black Crow AI’s response to how to solve this challenge for e-commerce brands.

The Hidden Reason Your Facebook Ads Are Broken (And How to Fix Them)

Every e-commerce brand knows that Facebook advertising is much less effective than it was a few years ago, but the reason why isn’t quite as well understood. Surprisingly, the answer often involves user data loss.

This data loss is the direct outcome of updates made to Apple’s Safari web browser, which accounts for 34.5% of the overall browser market and 52.2% on mobile devices. In service of increased privacy, these updates shortened the lifespan of Safari browser cookies to a single day from ad channels and seven days for direct or organic traffic. But since the buyer journey often extends beyond a day, these updates suddenly made it much more difficult to understand the complete buying cycle.

But don’t worry, the story doesn’t end there! This article will dive into the specific ways this user signal loss is impacting the effectiveness of Facebook advertising. It will also explain why server side user tracking has become the go-to solution for fixing Facebook ads and improving your e-commerce brand’s return on ad spend (ROAS).

Anatomy of a 30 day buy-cycle

To illustrate the user data loss created by shortened user tracking timelines, let’s join Jo as she shops for an e-bike. She’s looking for a bike she can use to get groceries a couple times a week without having to use her car, and take her three-year-old along for the ride. She heard about a good e-bike brand from a friend, and pulls up her phone after her child’s bedtime to do a little research.

Day 1

Jo notices an ad on Instagram from the same bike brand her friend recommended. She taps on it, and browses through different models and feature options on the brand’s website. Each of these actions are “events” that include important signals about Jo’s preferences and intentions as a buyer.

Day 15 

The weather’s been cold and rainy, so Jo hasn’t felt very motivated to continue her research. But there’s one e-bike model she’s been thinking about, so she returns to the website to compare color options, and decides on a cream colored bike with navy accents. These actions are additional events that further flesh out her journey as a potential buyer.

Day 30

The nice weather has returned, and Jo is feeling ready to finally purchase the e-bike and spend some more time outside. She returns to the company website, adds the model and color she’s been dreaming about to the cart, and makes a purchase. Yep, that means a few more events are added to her customer journey.

User tracking limitations spoil Facebook’s attribution 

Plot twist: Jo is actually three different people! Just kidding. But because the cookies in Jo’s browser expire so quickly, Facebook actually does see Day 1, Day 15, and Day 30 as three separate people.

Even though each of Jo’s interactions with the e-bike brand triggered several different events, these disappeared before the next interaction. Each time Jo returned to the site, it was like starting from scratch. Rather than building on previous events and corresponding data, it was like a blank slate.

This resulted in “Day 30 Jo” getting credit for the purchase, while “Day 1 Jo” and “Day 15 Jo” were completely disconnected from the equation altogether. They might as well have been different people.

 

Signal loss is training Facebook’s algorithm to become worse  

Not being able to understand Jo’s complete customer journey isn’t ideal. But the long term implications for the e-bike’s company’s marketing efforts are actually much more concerning. Over the same 30 day window during which Jo made her purchase, they sold more than 2,500 e-bikes to other customers, each one with their own unique (and misattributed) journeys.

Every customer further compounds the problems that result from user signal loss. It’s impossible to correctly attribute purchases to the ads that generated them, which impacts future ad bidding. And since Facebook only sees “Day 30 Jo” as the one who made the purchase, it avoids targeting future ads to someone similar to “Day 1 Jo” and “Day 15 Jo”. This results in smaller targeting audiences and/or less effective marketing efforts.

What’s more, since the whole Facebook advertising platform relies on algorithms, all of the negative outcomes above are essentially training the algorithm to be less accurate and more expensive. It’s an unfortunate cycle, with ever-decreasing ROAS and higher CPMs at the center.

 

Continue Reading to Learn How to Fix Your User Data Loss


What we're Reading

Pure play DTC is increasingly difficult in this market environment, but that doesn’t mean it shouldn’t be a core part of your channel mix. In fact, some brands are thriving with DTC. 

 

Read more about the future of DTC brands from Retail Brew. 

 

Rising customer acquisition costs and challenges brought on by Apple’s iOS 14 privacy changes have opened the door to innovation and pushed investors to look for “category-defining” companies. 

 

Read the latest on e-commerce deals and valuations from Retail Dive.

 

Knowledge databases are as important to AI progress as foundational models. And people who organize, store, and catalog their own thinking and reading will have a leg up in an AI-driven world.

 

Read more about how GPT-4 is a reasoning engine from Every.


What Does Black Crow AI Do Again?

Black Crow AI is used by over 150 E-commerce brands, like Sakara, Bearaby and ShortyLove. We're the first to bring enterprise quality machine learning to e-commerce challenger brands to boost their profitability. We help you optimize and accelerate your entire DTC marketing operation with real-time predictions that reveal and activate hidden value in the data you already own.

Request your 30 day free trial

Funnel Smarter.
Black Crow AI is a leader in Predictive Analytics on G2 Black Crow AI is a leader in Predictive Analytics on G2 Black Crow AI is a leader in Machine Learning on G2 Black Crow AI is a leader in Americas Predictive Analytics on G2 Black Crow AI is a leader in Predictive Analytics on G2 Black Crow AI is a leader in Machine Learning on G2