The Wrong Question

"Which platform wastes the most budget?" is the question most business owners ask when a campaign disappoints. It's the wrong question. Meta, Google and every other platform can be run efficiently or wastefully — the platform isn't what determines which one you get.

What actually determines it is a handful of decision-making patterns that show up regardless of platform, budget size, or industry. They're well documented in behavioural economics, and once you can name them, they're much easier to catch in your own account.

The Sunk Cost Fallacy Keeps Dead Campaigns Alive

Psychologists Hal Arkes and Catherine Blumer demonstrated in 1985 that people continue investing in a losing course of action specifically because of what's already been invested — even when that past investment has no bearing on whether continuing is a good idea.

In an ad account, this looks like: "we've already spent £2,000 on this campaign, we can't stop it now, it might turn around." The £2,000 is gone either way. The only question that matters is whether the next pound is likely to perform — and that question gets harder to ask honestly the more has already been spent.

What to watch for: a campaign that's been given "one more week" three weeks running is a sunk cost problem, not a targeting problem.

Action Bias: Doing Something Feels Better Than Waiting

A well-known 2007 study of penalty kicks by Michael Bar-Eli and colleagues found that goalkeepers dive left or right far more often than they stay in the centre — even though staying central saves more penalties on average — because a save that comes from standing still feels like inaction, and a miss while diving feels more forgivable than a miss while standing still.

The same bias shows up in ad accounts as constant, premature tinkering: pausing ad sets before they've left the learning phase, rewriting copy after two days of data, reallocating budget every time a number moves. Every change resets the platform's own optimisation and adds noise that makes the next decision harder to make well.

What to watch for: if your campaign history shows more edits than there were meaningful data points to justify them, action bias — not the algorithm — is the leak.

Vanity Metrics and the Illusion of Progress

Impressions, reach and click-through rate are easy to report and easy to feel good about, because they almost always go up. None of them tell you whether the campaign made money.

This isn't really a psychology bias so much as what it enables — a report full of rising numbers creates the same reassurance a genuinely improving campaign would, without the underlying substance. It's comfortable to look at and easy to present, which is exactly why it survives even when it isn't the metric anyone actually cares about.

What to watch for: if a monthly report leads with reach or impressions rather than booked calls, leads or revenue, that's a sign the account is being optimised for the wrong number — one of the reasons we build conversion funnel tracking around the metric that actually matters before touching ad spend.

The Familiarity Trap

Loss aversion (the same principle behind why hiring decisions stall, covered in our piece on why SMBs delay hiring an agency) also applies to channel allocation. An underperforming platform you already understand can feel safer than a new one you don't — even when the numbers say otherwise — because switching introduces a visible, attributable risk, while staying put just extends an already-diffuse loss.

This is how budget quietly stays parked in a channel that's stopped earning its share, purely because moving it feels like the riskier decision.

A Simple Audit Framework

None of these biases are solved by working harder inside the same decision-making pattern. They're solved by changing the pattern itself.

  • Set a stop-loss before launch, not during. Decide the budget or timeframe a campaign gets before it's judged, while you're not yet emotionally invested in its outcome.
  • Separate "reporting" metrics from "deciding" metrics. Impressions and reach can stay in the report for context. Only cost-per-lead, cost-per-booking or revenue should be allowed to trigger a change.
  • Give every test a minimum run time before touching it. Statistical noise looks like a signal for the first few days of almost any campaign — action taken in that window is usually action bias, not insight.
  • Review channel allocation on a schedule, not a feeling. A quarterly look at where budget sits, compared with where results come from, catches the familiarity trap before it compounds.

The platforms aren't where most ad budget disappears. The decisions around them are. Fix the decision process and the platform usually looks a lot more efficient than it did a month ago.