A Consent Banner Failed. Here Is Who Paid for It.

Archi4 min read

privacyregulationconsumers

Contents
  1. The Bug That Kept Consenting
  2. Who Was in the Affected Group
  3. The Asymmetry Nobody Budgets For
  4. What a Better Failure Looks Like
  5. Three practices that would have changed the outcome
  6. The Lesson That Travels

The Bug That Kept Consenting#

On a Friday in late March, a regional retailer in the Midwest pushed a routine update to its cookie consent banner. The new version was supposed to load faster on mobile. Instead, it silently stopped saving the "reject all" choice for any visitor using an older Android browser. For six weeks, roughly 214,000 shoppers who tapped reject were tracked anyway.

The company discovered the fault in May during a scheduled internal audit. It disclosed the issue to its state attorney general, patched the code in four days, and posted a notice on its homepage. By the standards of most privacy failures, this one was small: no credentials leaked, no SSNs, no payment data. The retailer's outside counsel described it as a "technical misconfiguration" in a filing.

That description is accurate and also beside the point. The people who were affected were not a random slice of the customer base. They were disproportionately the ones with the least capacity to notice, complain, or leave.

Who Was in the Affected Group#

Older Android browsers are not evenly distributed across a customer base. They cluster among people on prepaid plans, people using hand-me-down phones, and people in households where a $400 handset is a significant purchase. The retailer's own traffic data, included in its filing, showed that 61 percent of the affected sessions came from devices more than four years old. A third came from zip codes with median household incomes below $38,000.

Those shoppers had done the right thing. They had seen a banner, read enough of it to find the reject button, and tapped it. The system took their choice and then discarded it. The failure was not that the banner was confusing. It was that the banner lied.

Privacy law has a vocabulary for this. Consent must be informed and specific, and it must be revocable. But the law is mostly written for the moment of collection, not the moment of silent reversal. A company can be fully compliant on paper and still override a user's stated preference for six weeks.

The Asymmetry Nobody Budgets For#

The retailer's response was, by the letter of the rules, adequate. Notice was posted. The regulator was informed. The fix shipped. What the response did not include was any direct outreach to the people whose choices were ignored. No email, no text, no in-app message. The homepage notice reached people who happened to visit the homepage again, which is to say, almost nobody in the affected group.

This is the pattern in a lot of privacy failures. The cost of the mistake lands on users who have the fewest ways to route around it. A shopper with a current iPhone and a desk job can install a tracker blocker, switch browsers, or write a complaint that a regulator's office will read. A shopper on a four-year-old prepaid Android with a data cap has none of those options and probably never learns the reject button did nothing.

The company, meanwhile, absorbed the failure as a legal expense. Its filing estimates the cost of outside counsel, the audit, and the patch at under $90,000. It did not estimate the cost to the 214,000 people whose browsing history was collected after they said no.

What a Better Failure Looks Like#

There is a version of this incident that would have ended differently, and it does not require new law. It requires treating a consent failure as a product failure with named users, not a compliance event with a filing number.

Three practices that would have changed the outcome#

  1. Direct notice to the affected cohort. If a company knows which accounts or devices were affected, it can reach them. The retailer had device-level logs. It chose a homepage banner anyway.
  2. A kill switch on the consent path. Consent code should fail closed. If the banner cannot save a choice, it should stop collecting until it can. The retailer's code failed open, which is the default in most banner implementations.
  3. A reporting line for the people who noticed. The retailer's support team logged 47 complaints about the banner during the six weeks. None were escalated to legal or engineering. A single weekly review of consent complaints would have caught the bug in April.

None of these are exotic. Each costs less than the outside counsel invoice. The reason they do not happen is that privacy failures are still measured in regulatory exposure rather than in user harm, and the users who bear the harm are the ones whose complaints are least likely to reach a decision maker.

The Lesson That Travels#

The retailer's case is unremarkable, which is why it is worth studying. Most privacy failures are not breaches. They are quiet reversals of a choice a user made, lasting weeks or months, affecting the people with the least margin to absorb them.

The regulatory outcome, if there is one, will likely be a settlement and a corrective action plan. The shoppers on old Android phones will get nothing, because the law does not currently have a mechanism to give them anything. What they will get is another banner, a little slower, with a reject button that may or may not work. The burden of checking falls on them, again.

More articles