Protocol | Policy

Facebook and Apple’s privacy war is mucking up voter turnout efforts

"This is bad. I didn't think it would be this bad."

"I voted" stickers
Photo: Element5 Digital/Unsplash

Tatenda Musapatike is as familiar as anyone with the way Facebook's political ad system works. During the 2016 election and the 2018 midterms, she helped Democrats run Facebook ads as a client solutions manager for the company, before leaving in 2019 and eventually launching her own nonprofit to focus on turning out voters from underrepresented groups.

So this year, when Apple introduced privacy changes that allowed iOS users to opt out of being tracked by apps, including Facebook, Musapatike knew better than most that the switch would be hard on Facebook's political advertisers.

She had no idea it would be this hard.

In a new report released Thursday by Musapatike's organization, Voter Formation Project, the group found that in the run-up to Election Day this year there was a precipitous decline in the percentage of conversion ads Facebook was serving to iOS users. Conversion ads are ads that ask users to go to the advertiser's website and take an action. In politics, that can mean asking them to fill out a form, commit to vote or even to donate. According to the Voter Formation Project report, not only was a smaller slice of the group's conversion ads reaching iOS users than before, but the cost of driving people to the website also shot up to an average of $424 per conversion — more than three times what Musapatike said she would have expected during an off-year election.

"Having worked at Facebook itself for years, it's hard to surprise me," Musapatike said. "This kind of set off a panic. This is bad. I didn't think it would be this bad."

Since the 2016 election, when President Trump's campaign aggressively used conversion ads to fundraise, political advertisers of all stripes have leaned into this type of advertising as a core part of their digital strategies. Musapatike's organization, for one, was using conversion ads to send potential voters in Texas and Virginia to the Voter Formation Project website to make a plan to vote.

Facebook started warning advertisers that Apple's privacy changes would be a blow before Apple even rolled them out. Voter turnout groups like Musapatike's were no exception. "Our system is built to deliver ads to people who will find them most relevant, regardless of the device they are using," a Meta spokesperson said. "As we've said over the past year, Apple's changes make it harder to do that for some iOS users, which is negatively impacting small businesses, organizations and developers."

The idea that Facebook might not actually be showing these ads to iOS users at the same rate set off immediate alarm bells for Musapatike and her team. They began digging into the data and found that there was a substantial discrepancy between the percentage of conversion ads being delivered to Android users versus Apple users.

Before the iOS change, more than 60% of the group's conversion ads on Facebook were delivered to iOS devices. After the privacy update, that figure was just under 37%.

Image: Voter Formation Project

The drop-off in iOS impressions Musapatike's team observed was specific to conversion ads. Ads with other goals like, say, getting the broadest reach, were still being delivered to more iOS devices than Android. And even though companies like Snap are also struggling with Apple's privacy changes, the impact on iOS users wasn't appearing on other platforms. (The report acknowledges that could be because some platforms don't report impressions by device type.)

Image: Voter Formation Project

Still, even if other platforms were having the same issue, it wouldn't matter as much as it did on Facebook, Musapatike said. Facebook is where political advertisers had the most success with conversion ads before the iOS update, so it's also where a drop-off would hurt the most. "Facebook had, when I was there, roughly 90% of the conversion market of ads in politics, period," Musapatike said. "It was the most efficient place you could get donations or sign-ups."

'Let them click'

The question is: Why did this happen? The report accuses Facebook of "suppress[ing] delivery to devices that are more likely to block its ability to track the conversion action." Meta's spokesperson insisted that wasn't a deliberate decision on the company's part and said Facebook's automated ad system simply targets the users it predicts are most ripe for conversion on any given ad.

The problem is, those predictions are based on clicks — clicks that Facebook can't see clearly now on iOS, because it's not getting data back from users' devices. So the system appears to be overlooking them. That's not to say iOS users aren't receiving any ads; other types of ads that don't depend on third-party data from devices don't face the same problem.

Facebook, of course, would like to blame Apple for this issue, and it's true that the type of unbridled third-party tracking political advertisers have traditionally done is precisely the sort of thing Apple is trying to prevent. But the inability to see exactly who clicked through a conversion ad isn't itself an insurmountable challenge for advertisers, Musapatike said. In her case, even if there's no data exchanged between the device and Facebook, it wouldn't be so hard to figure out how many people filled out a form on her website after she ran a Facebook ad campaign asking people to do so.

But she can't do any of that if Facebook isn't showing iOS users her conversion ads to begin with. "If you're going to ask most advertisers, we'd say: 'Put our ads in front of these people and let them click,'" Musapatike said. "Even if you're not getting as much data back from these devices, that doesn't mean putting your ads in front of them or getting them to fill out your form isn't valuable."

Meta's spokesperson pointed to public statements the company has made both on its blogs and in earnings calls related to conversion issues in light of the iOS update. Specifically, the company has said it believes it's underreporting actual conversions by 15%, which may be leading to the higher costs per conversion that Musapatike saw. In other words, advertisers may actually be getting more conversions for their money. But that has little to do with whether Facebook is actually showing these ads to begin with.

The company has also recently introduced a new API for conversions that it said will "help businesses optimize ad targeting, decrease cost per action and measure results."

As for Musapatike, after realizing where conversion ads were falling short, her team began investing in lead generation ads on Facebook, which still enable Facebook users to fill out a form without ever leaving the platform. Those ads don't afford her the flexibility of directing people to her own website, but at least they're getting seen.

It's still unclear exactly how the ad changes impacted actual turnout in crucial states like Texas and Virginia. Those calculations take voter data that won't be available for months. But as the 2022 midterms draw closer, Musapatike's team is already reassessing where and how they'll spend their money. "I do not know the exact mechanism as to what's behind it," Musapatike said, but for advertisers like herself, at least, it means "if you're running a campaign where you want people to go to your website, it is less likely to happen through Facebook."

Protocol | Enterprise

Meta thinks it can now use smaller data sets to flag Facebook content

Despite the constant deluge of content flowing into Facebook and Instagram, Meta has struggled to get enough data to train AI to spot harmful content, so it’s banking on an emerging approach.

Meta plans to announce that few-shot learning shows promise in its constant battle to weed out disinformation or other content that violates its policies on Facebook and Instagram.

Image: Meta

After a terrorist attack on a mosque in Christchurch, New Zealand was livestreamed on Facebook in 2019, its parent company, now called Meta, outfitted London police officers with body cams while they conducted terrorism training. At the time, Meta said there wasn’t enough video data to train its artificial intelligence systems to detect and remove violent content, so it hoped the body cam project would produce more of that scarce AI training data.

A year prior to that horrific incident, the company acknowledged that it failed to keep up with inflammatory posts from extremist groups in Myanmar. Again, it said the problem was a lack of data — there wasn’t enough content in Burmese to train algorithmic moderation systems to spot more of it.

Keep Reading Show less
Kate Kaye
Kate Kaye is an award-winning multimedia reporter digging deep and telling print, digital and audio stories. She covers AI and data for Protocol. Her reporting on AI and tech ethics issues has been published in OneZero, Fast Company, MIT Technology Review, CityLab, Ad Age and Digiday and heard on NPR. Kate is the creator of and is the author of "Campaign ’08: A Turning Point for Digital Media," a book about how the 2008 presidential campaigns used digital media and data.

In a tight labor market, businesses are competing for top talent, even as employees leave in droves. A record 4.4 million Americans resigned in September 2021 — the highest on record for nearly 20 years — ushering in what some call the Great Resignation. That same month, 65% of U.S. workers said they were looking for a new job.

Business leaders have to respond to mitigate the negative impacts of this disruptive churn, with 36% of CFOs saying they're very concerned about turnover remaining high indefinitely and weighing on revenue growth. The answers to this challenge should be informed by the root causes of employee dissatisfaction as well as retention drivers.

Keep Reading Show less
Suneet Dua, PwC
As PwC’s US Products & Technology Chief Revenue and Growth Officer, Suneet Dua is responsible for driving more than $1 billion in product revenue and executing PwC’s product revenue strategy. He’s focused on driving innovation, delivering world-class, forward-thinking products and digitally upskilling the workforce and society at large. With 20+ years of technology, media and entertainment industry experience, he’s positioned as a catalyst for organizational transformation and delivers on the firm’s promise to solve the world’s most important problems. Additionally, he launched Salesforce and client-focused centers of excellence, such as our Cybersecurity centers in Israel, Singapore and India––all to improve the way PwC serves its clients. During his tenure as US Chief Product Leader, Suneet, and his team, played a critical role in designing and implementing digital tools that upskilled more than 55,000 of its US employees, which led to the development of PwC’s digital learning platform, ProEdge, that addresses the digital skills gap crisis facing today’s workforce. He also serves as a board member of PwC’s Trifecta Consulting (US, China, Japan and Mexico). Previously, Suneet served on PwC’s US leadership team and was Global Client Market Leader for PwC’s Global Network.
Protocol | Workplace

What will work look like in 2022? Glassdoor makes four predictions.

Tech companies will continue to have trouble hiring workers.

According to a report from Glassdoor, local companies will also have to pay more to compete with companies that are offering San Francisco or New York rates to remote workers.

Photo: MoMo Productions/Getty Images

2021 was a difficult but pivotal year for tech workers and employers alike. We’ve got mixed news: 2022 will likely continue to be difficult but perhaps a little more, well, precedented.

Glassdoor released four predictions for the workplace of 2022 Wednesday based on data it gathered from reviews, salaries and conversations happening on its site, as well as economic trend data. Here’s what the career platform sees in the workplace crystal ball.

Keep Reading Show less
Michelle Ma

Michelle Ma (@himichellema) is a reporter at Protocol, where she writes about management, leadership and workplace issues in tech. Previously, she was a news editor of live journalism and special coverage for The Wall Street Journal. Prior to that, she worked as a staff writer at Wirecutter. She can be reached at

Protocol | Enterprise

COVID-19 kickstarted a war over web accessibility

The pandemic spurred demand for a more accessible web, but experts and practitioners disagree on the best approach to get there.

Experts and practitioners disagree on the best approach to building an accessible web.

Image: alexsl/Getty Images

The pandemic triggered a surge in demand for technology that helps companies adapt their websites for users with disabilities as businesses scrambled to accommodate customers who were now forced to do almost everything online.

This period gave a boost to companies such as AudioEye, EqualWeb and Deque, which offer accessibility services like alternative text that describes images for visually impaired users. But it also sparked a war over the best way to build a more accessible web, with one side arguing the fastest way to achieve change is to put accessible overlays onto existing sites, and the other arguing the web will never be truly accessible until developers build it that way from the start.

Keep Reading Show less
Aisha Counts

Aisha Counts (@aishacounts) is a reporting fellow at Protocol, based out of Los Angeles. Previously, she worked for Ernst & Young, where she researched and wrote about the future of work, emerging technologies and startups. She is a graduate of the University of Southern California, where she studied business and philosophy. She can be reached at

Hirsh Chitkara

Hirsh Chitkara (@ChitkaraHirsh) is a is a reporter at Protocol focused on the intersection of politics, technology and society. Before joining Protocol, he helped write a daily newsletter at Insider that covered all things Big Tech. He's based in New York and can be reached at

Latest Stories