On Autopilot, the Platform Wins. Take Back the Controls.

I rarely post on LinkedIn. But after reading the coverage of the FTC's lawsuit against Amazon, I noticed a gap. Plenty of people have explained the allegations. Almost no one has said what brands should do next.

For me, this moment recalls one of the quieter lessons of the Snowden era. Reuters reported that RSA was paid $10 million to make a flawed encryption standard the default in a toolkit other companies built their products on. No one had to break into anyone's systems. The default did the work, and trusted vendors shipped it.

The parallel isn't about intent. It's about how much power sits in a default.

Many of us suspected something like it. Now the FTC filing has brought it into the spotlight. This lawsuit is a similar moment for advertising, and the bigger risk isn't any one auction. It's the defaults.

Why I'm writing

My view: I don't believe Amazon did anything legally wrong. In my experience, Amazon never sold its auction to us as second-price, and we never treated it as one. But the case is a useful reminder. In my view, results can look fine while the platform keeps much of the gain from its own improvements. That's the real lesson for brands.

I worked with some of the people whose names are redacted in the filing. From 2009 to 2012, Amazon was my customer while I was at another major ad platform. I worked at Amazon from 2013 to 2020, and since 2022 I've been building DPG, a tech-enabled agency focused largely on Amazon.

Since starting DPG, we've bid as if every auction were first-price, based on what we saw in our own campaign data. It's one reason our brands often outperform their competitors.

I also want to be fair to the people at Amazon. I think of them in three broad groups: the account teams (sales, account management and go-to-market), product management, and yield management. Nearly everyone I worked with worked tirelessly to help advertisers succeed while protecting the shopping experience. Their jobs just got harder, and they remain some of the best advocates brands and agencies have.

The problem is getting worse, not better

Every ad platform adjusts its auctions to maximize yield. In April 2025, a federal judge ruled that Google illegally monopolized key parts of the ad tech market.

When advertisers managed campaigns by hand in platform consoles, this was easier to spot and manage. The more we hand our budgets to automated tools—whether the platform's, an agency's or a tech provider's—the easier it is to lose control, and the harder it is to see.

Across the industry, platforms are now pitching AI that runs your business for you. Some are saying the quiet part out loud. Mark Zuckerberg put it plainly in 2025: "You're a business, you come to us, you tell us what your objective is, you connect to your bank account, you don't need any creative, you don't need any targeting demographic, you don't need any measurement, except to be able to read the results that we spit out."

Here's how we got here

• Auctions moved toward first-price. Open programmatic moved years ago, and Google Ad Manager switched in 2019. Platform terms generally let them change these mechanics at any time.
• Platforms opened campaign management through APIs, with their own bid and budget recommendations built in.
• Tech providers built on top of those APIs. They added optimization tools, bundled standard reporting, and sold the result as proprietary. Many deliver real value, but how much of the "optimization" is the platform's own recommendation, repackaged? You should ask.
• Now come MCP connections. Amazon's Ads MCP Server has been in open beta since February, and Meta has opened similar connectors. They let any AI agent create and manage campaigns. That makes it easy to launch "AI-powered" products that are often thin layers over the platform's defaults.

Building independent models is hard and expensive. So is the deeper work of growing a brand beyond short-term ROAS. Both hurt margins and valuation multiples, so in my experience, most providers take the easier route.

This is the blind spot. Nobody has to hijack your account. Platform recommendations and defaults flow smoothly through the tools you already trust: your agency's AI, your tech provider's "proprietary optimization," your own agent connected through an MCP. The tools that wrap those defaults become one more vendor shipping them, usually without knowing it.

What Brands Should Do

The fix for a blind spot is knowing what's inside your tools. Ask your agency or tech provider these exact questions:

1. Which parts of how you manage my campaigns are your own technology, and which come from the ad platform?
2. Do you run on a third-party campaign management platform? If so, how does it work, and what powers its AI features?
3. What controls keep my data from leaking? (An NDA doesn't count).
4. How do you protect my budget from margin capture by the platforms?
5. How do you catch platform changes? One retail media network recently expanded how far budgets can automatically flex during peak events. Did your tool catch it?

Push past "it's a blend of platform insights and our proprietary technology." A good partner gives specific answers.

What Agencies and In-House Teams Should Do

The playbook is simple to outline and hard to execute:

• Use platform APIs and MCPs for data and execution, not strategy. They're great for pulling reports, building shell campaigns, and pushing your own changes in bulk.
• Own your definition of success. Platform ROAS is directional at best. It depends on how it's measured and how campaigns are run, and its definition changes—sometimes publicly and sometimes quietly.
• Treat nearly every auction as first-price, and shade your bids accordingly. Put tight guardrails on campaigns, because platforms increasingly roll out automated features that default to broader spending parameters.
• Standardize measurement across retailers. New-to-brand on Amazon, Walmart and Instacart are three different metrics. Normalize them, then move money to where you see the most incremental growth.
• Build your own bidding models where you can. Platform models have to work across tens of thousands of advertisers at once. Yours only has to work for your brands. Use that information advantage, and tie it to your own measurement.
• Keep your listings healthy. Retailers change backend requirements constantly, and gaps open up in your catalog when they do. Most you won’t notice until you look at the individual listing backend, but the papercuts add up.

It isn't easy. If you're hoping to run it all through Claude or another AI agent, you have more work ahead than you might expect. Think hard about whether your agents can do it, or whether you're better off focusing on your business and bringing in a team that does this every day.

DPG couldn't have built this alone. We've had the support of incredible people at Amazon who advise us and help us build on our shared customers' success. Yes, ad platforms focus on maximizing yield, as any business would. But most of the people I know at Amazon, Walmart, Instacart and elsewhere are still focused on helping brands grow.

Our job is to make sure brands keep the benefit, and that the defaults and settings running their business are set for their benefit, not the platforms.