Home Data-Driven Thinking There’s A Lot Wrong With Google’s And Meta’s Non-Transparent ‘Refund’ Practices

There’s A Lot Wrong With Google’s And Meta’s Non-Transparent ‘Refund’ Practices


Over the past few weeks, thousands of Google and Meta advertisers have seen refunds trickling back into their accounts.


Well, it’s hard to know for sure, because both companies have a habit of obfuscating the reasons behind campaign refunds. Meanwhile, campaign misappropriation issues seem to be cropping up more frequently on both platforms, and it can be difficult for advertisers to know what really happened with their misspent budgets.

But both Google and Meta are playing with fire. Their opaque refund practices have already exposed them to customer blowback – and could lead to class-action lawsuits by disgruntled advertisers.

They aren’t even refunds, though

One glaring issue with Google’s and Meta’s so-called “refund” practices is that they aren’t actually issuing refunds.

They’re paying advertisers back with ad credits.

I’d argue that any advertiser who receives platform ad credits in return for misspent budgets should file a complaint with the Better Business Bureau.

Ad credits would be fine if we were talking about an earnest effort at campaign reconciliation – but that’s not what we’re talking about here. Instead, the ad platforms are misfiring and causing advertisers to unintentionally overspend by tens – or even hundreds – of millions of dollars. This money is going directly into Google’s and Meta’s respective pockets.

It’s pure misappropriation on the part of the platforms.

Our bug, your problem


AdExchanger Daily

Get our editors’ roundup delivered to your inbox every weekday.

Take Google, which recently sent a memo to many of its DSP customers alerting them that ad credits would show up in their account.

Despite persistent questioning, Google never clarified publicly or even on background why it issued those credits, other than to say it had misspent on a certain subset of inventory between July and December 2023. It also never revealed how much in aggregate was refunded.

The lack of transparency is one thing. But Google and Meta are taking real money from advertisers – real money in that these are not trivial amounts, but also because advertisers are paying in cash. Google and Meta get to hire people, invest in things and pay for all the things they need to pay for.

Their advertisers, meanwhile, get tokens that only work at the Google or Meta arcade.

An advertiser I spoke with last year – a Shopify merchant with a clothes embroidery business – had to take out a loan to bridge the losses in her account after a Facebook ad glitch. The most she’d ever spent on ads in a day was in the low thousands of dollars, while on a typical Sunday it might be just a couple hundred. When the Facebook platform bugged last spring, it spent $12,500 in the wee hours of a Sunday morning and delivered no sales.

Hundreds of other small advertisers have similar stories.

Meta and Google ad platform credits prevent people from paying bills, hiring and investing in their business.

The devil’s advocate would argue that the money in question would be spent on the platform eventually, so what’s the big deal? But that’s not Google’s or Meta’s call. They should be returning misspent ad budgets and let the advertisers choose their next move.

When the Google or Meta ad platform misfires and spends drunkenly with no rationale or ROI, it should not be lucrative to the platform.

The machine learning grift

The dual trends of ad platform glitches and budget misappropriation are exacerbated by Google’s and Meta’s investments in machine learning products. In many cases, when there’s a problem with an automated ad-buying tool, the only party to complain to is an automated rep, because the platforms are replacing human account services with chatbot software.

And advertisers can’t effectively prove their budgets were misspent, because the automated tools are black boxes.

Take Advantage+, Meta’s machine-learning-based ad product. Advertisers don’t know where their ads served or why Meta charges a certain CPM or credits a conversion. It’s up to ad buyers who are deep in the platform every day to even realize when Meta has misappropriated budgets and then gather an army of other account operators to raise complaints in a power-in-numbers play.

When Meta glitched last year and took hundreds of millions of ad dollars, it didn’t bother to alert accounts, although it knew all the affected accounts and how much had been misspent. Meta required accounts to individually submit campaign reconciliation requests – many of which were rejected. It wasn’t until three weeks later as the result of a large advertiser pressure campaign that Meta agreed to refund all accounts and not require them to file reconciliation reports.

Best bad practices

But that was just for one bug. It remains Meta’s practice not to issue public alerts when budgets are misappropriated.

This is an effective strategy, because the advertisers who use Meta’s machine learning products are often the “set-it-and-forget-it” type. They simply opt in when the dashboard suggests a new upgrade or a best practice.

But it’s the best practices themselves that are misfiring.

Facebook recommends that its advertisers set much higher daily budgets than they intend to spend, but with specific rules for how the money is spent – a tactic called cost capping. This is so that a viral event or big day of profitable sales isn’t put on hold by an arbitrary budget limit.

That makes sense. But when there’s a bug with cost caps, huge chunks of budget gets spent way too quickly. A brand that expects to spend $1,000 on a given day might have a daily budget of $50,000.

When cost caps misfired last April, the Meta ad platform spent 125% of advertisers’ daily budgets, which was the upper limit of what the system was authorized to do. But Meta bumped that up to 150% late last year, and this year it was nudged up again to 175%.

Meta hasn’t fixed its ongoing problem with glitches and bugs, by the way – but the stakes are quietly going up and up for advertisers who simply opt into automated ad products.

Where’s the service?

A big issue is that many Meta and Google advertisers are not aware of what they’re signing up for when they opt into Advantage+ or PMax and how the terms and conditions can punish them.

Most Advantage+ advertisers are likely not aware that signing up will fundamentally change how their dollars are spent and the implications moving forward. They just hit “Yes” to a dashboard prompt for what looks like new promising optimization improvements. The same is true for Google’s PMax.

For example, when I do a Google search using Safari on my iPhone, I’m constantly nudged with a prompt to sign in using my known Chrome and Gmail account. This happens at least half a dozen times a day, and I say “no” every time. But if I ever slipped and hit “yes,” I’d never see that notification again. From then until the end of recorded time, it would be taken for granted I want that account linked and logged-in at every opportunity, and Google would do so without ever asking again.

That’s similar to the experience of opting into PMax.

On Google and Meta dashboards, signing up for these machine learning products isn’t just pressed on advertisers; it’s treated as the only and inevitable way forward – despite very few advertisers understanding the nuances and the risks.

The terms of service for these machine learning products grant Meta and Google vast license to determine the prices advertisers pay for ads. Meta and Google ad accounts are also linked directly to an advertiser’s bank account or credit card. When there’s an error in the system, the terms also allow misappropriated dollars to be refunded as credits.

It’s not unlike how Planet Fitness had Ts&Cs that made it practically impossible to cancel a subscription. But in that case, enough customers complained that the FTC eventually voided those contracts.

If enough advertisers were to make the case that they didn’t understand the terms they’d agreed to when signing up for Google’s and Meta’s machine learning products – and those terms are proven to be unfair to their businesses – then the platforms could be in hot water.

An unfair contract isn’t a great defense.

Which brings me back to my earlier point: Report these instances to the Better Business Bureau.

Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media. This column is part of a series of perspectives from AdExchanger’s editorial team.

Follow James Hercher (@JamesHercher) and AdExchanger (@adexchanger) on Twitter.

For more articles featuring James Hercher, click here.

Must Read


Perion Shutters Content IQ, Its Made-For-Advertising Division

Laptop fans can rest a little easier. A network of well-known MFA sites operated by Perion-owned Content IQ have been taken offline.

‘Incrementality’ Is The Buzzword That Stole Prog IO

Well, that’s a wrap on Programmatic IO Las Vegas 2024! The AdExchanger editorial hopped on stage for a live recording of The Big Story to round up all the moments that made us go “a-ha” this week, including observations on commerce media, CTV and generative AI.

Paramount And Shopsense Add Programmatic Demand To Their Shoppable Ad Network

What if the new storefront is a person sitting on their couch and scrolling their phone?

Privacy! Commerce! Connected TV! Read all about it. Subscribe to AdExchanger Newsletters

Scott’s Miracle-Gro Is Seeing Green With Retail Media

It’s lawn season – and you know what that means. Scott’s Miracle-Gro commercials, of course. Except this time, spots for Scott’s will be brought to you by The Home Depot’s retail media network.

Walled Garden Platforms Are Drowning Marketers In Self-Attributed Sales

Sales are way up; ROAS is through the roof across search, social and ecommerce. At least, that’s what the ad platforms say.

Comic: Working Hard or Hardly Working?

Shadier Than Forbes? Premium Publishers Are Partnering With Content Farms To Make A Quick Programmatic Buck

The practice involves monetizing resold subdomains jammed with recycled MFA articles produced by notorious content farms.