The Google Ads bidding change is a tax on lazy targets

On August 17, Google is quietly rewriting the deal for a big chunk of ecommerce advertisers, and most of them haven't noticed yet. If you run target CPA or target ROAS campaigns that are limited by budget — which is a huge slice of every ecommerce account I've ever audited — the algorithm is about to stop giving you free efficiency. It's going to start actually chasing the target you set.

That sounds like Google being helpful. It isn't. It's Google exposing the fact that most bid targets in most accounts were never real numbers to begin with. They were placeholders someone typed in during setup and nobody ever tuned, because the campaign kept beating them and nobody wants to poke a working campaign.

The gap between the placeholder and the reality is about to become a line item on your P&L.

What's actually changing

The mechanic is simple, even if the second-order effects aren't. Today, a campaign with a £100 target CPA that's capped by budget will often deliver conversions at £50 or £60. The budget is doing the work, not the target. The algorithm buys the cheapest conversions it can inside the cap, then stops, because the cap is where it stops. You get the leftover target headroom as free efficiency.

After August 17, that headroom becomes fair game. Google will spend up toward the target you set. The £50 CPA drifts toward £100. The 600% ROAS drifts toward the 400% you set as your ceiling. Google frames this as "more predictable performance as you scale." That framing is technically true. It's also selling you a haircut and calling it a spa treatment.

The change applies to Search, Shopping, Performance Max, Demand Gen, Travel, and Display. App, video reach, and video view are excluded. For PMax and Demand Gen specifically, Google has also flagged that you may see traffic shift between channels as the system rebalances — which means the reporting picture gets messier at the same moment the economics get tighter.

Your target stops being a ceiling you rarely touched and becomes a destination the algorithm actively aims for.

Why so many campaigns were quietly overdelivering

Here's the thing nobody at Google is saying out loud: the overdelivery wasn't the algorithm being generous. It was a structural artefact of how budget-limited targeting works.

The target could be £200 and the campaign would perform identically, because the £2,000 daily budget is doing all the actual limiting.

When budget is the binding constraint, the algorithm optimises within the cap. It doesn't try to spend the full target — it can't, because there's no runway to spend into. So it picks off the cheap conversions and leaves everything else on the table. That produces a CPA well below target, which produces a happy quarterly review, which produces the enduring belief that your bidding is finely tuned. It isn't. Your budget is small, and your target is loose.

I see this constantly in ecommerce audits. A client will show me a campaign with a £75 target CPA that's been running at £38 for a year, and they'll tell me it's their best campaign. And it is — but not because the target is doing anything. The target could be £200 and the campaign would perform identically, because the £2,000 daily budget is doing all the actual limiting. The target is decorative.

August 17 turns the decoration into a load-bearing wall.

The three types of accounts and what happens to each

Not every account gets hit the same way. In my experience there are roughly three cohorts here, and the outcomes are very different.

The first is the tuned account. Targets that reflect real break-even numbers, revisited quarterly, defended against dilution. These accounts will feel the change least. Their targets already sit close to where the algorithm was landing anyway, so the drift is small. If you're in this group you've earned an easy few weeks.

The second is the aspirational-target account. Someone set a target once as a stretch goal — "wouldn't it be nice if CPA came in at £60" — and the campaign happily overshot it. That target was never a real constraint. It was a wish. After August 17 the wish becomes an instruction. The algorithm will spend up to £60 on conversions that previously came in at £35, and your blended CPA rises accordingly. Same conversion volume, worse economics.

The third is the placeholder account. This is the largest group and the one that concerns me most. The target was set at setup, never revised, and the campaign has been beating it by 40-60% for two years. Nobody knows what the current target actually represents because the person who set it has left the company. In these accounts, August 17 is going to produce a step-change in cost per acquisition that shows up in September reporting as an unexplained margin hit. And the reporting story will be brutal, because the answer to "what changed" is "we did nothing." Which is exactly the problem.

What Google is really doing here

Set aside the mechanics for a moment and look at the incentive structure.

Google has spent the last three years telling advertisers that Smart Bidding is more sophisticated than manual bidding, that targets should be trusted, that budget caps are the honest lever. The pitch has been: set your target, set your budget, let the machine optimise. Fine. Except the machine has been quietly saving advertisers money by leaving target headroom untouched — and that headroom is spend Google isn't capturing.

The August 17 change closes that gap. It doesn't raise your budget. It doesn't touch your target. It just removes the structural inefficiency that was benefiting you and returning that spend to Google. Ginny Marvin has been at pains to say this isn't Google telling you to spend more, and technically she's right. Google isn't asking you to spend more. Google is arranging for the money you already committed — the target you already agreed to — to actually get spent.

That's the loop. And it's an entirely legitimate business move on Google's part. Advertisers set a target of £100. Advertisers got £50. The gap was a rounding error the platform tolerated. Now the platform doesn't tolerate it. Fair enough. But let's not pretend it's neutral optimisation. It's Google collecting on targets you never meant to fully honour.

The measurement problem underneath this

The reason so many accounts are exposed is that most ecommerce advertisers don't actually know what their real break-even CPA is. They know last quarter's blended CAC. They know the margin their finance team says the business runs on. They know what the CFO would flip over if they saw. But the specific number that represents "I break even on this exact SKU at exactly this ad cost" — most accounts don't have that number sitting anywhere authoritative.

Which means when you sit down between now and August 17 to review your targets, you're not really reviewing targets. You're doing unit economics work that should have been done two years ago and never was. In my experience the businesses that struggle with this share a common cause: they treated Google Ads as a media buy rather than a cost of goods, so nobody in the business actually owned the maths.

The uncomfortable version of the advice here isn't "review your target CPA." It's "figure out, this week, what your real target CPA should be — the one that reflects current margin, current AOV, current LTV, current returns rate — and use that number, not the one that was aspirational in 2023."

If that exercise reveals your current target is genuinely wrong, fix it. If it reveals the target is right but you're capped by budget, then the honest question is whether to raise budget (accept more volume at target economics) or switch to Maximise Conversions with a hard budget (which is what Ginny Marvin has been signposting as the correct strategy when budget is the real constraint). Both are legitimate answers. What isn't legitimate is doing nothing and being surprised in September.

What to actually do before August 17

The window is short, and Google isn't going to move the date. A few concrete positions I'd hold.

Pull the "limited by budget" filter across your target-based campaigns and list every one that flag has appeared on in the last 90 days. That's your exposed inventory. For each one, look at actual CPA or ROAS versus target CPA or target ROAS over the last 30-60 days. If actual is more than 20% better than target, that campaign is going to move materially after August 17.

For each exposed campaign, make a deliberate decision. Either the target reflects real economics and you accept the drift toward it (which is fine, but do it consciously), or the target was a placeholder and you retune it to reflect current unit economics, or the campaign should switch to Maximise Conversions with a hard budget because budget is the honest constraint. Google's Bid Target Adjustment Tool went live on July 6 specifically to help with the second option. Use it.

Do not, under any circumstances, treat this as a nudge to loosen budgets. That's the reading Google is politely denying and that Ginny Marvin has been explicitly pushing back against. The change is about making stated targets honest, not about growing spend. If you use August 17 as cover to raise budgets across the account without doing the unit economics work first, you'll compound the problem, not fix it.

And brief your stakeholders now. Whoever gets the September performance report — finance, the CMO, the founder — needs to know before the numbers land that CPA is moving up on a subset of campaigns, why, and what you're doing about it. A conversation you have in early August is a strategy update. The same conversation in mid-September is a defence.

The honest limits

A few things this analysis doesn't cover.

I don't know how aggressively Google's algorithm will actually pursue the target versus continuing to prefer cheaper conversions where they're available. The announcement says performance will track "closer to" the target, not "at" the target. So the drift may be gentler than the worst-case reading suggests. Early data after August 17 will tell us more.

I also don't know how PMax and Demand Gen will handle the promised channel-mix rebalancing. If traffic shifts between Search, Shopping, and Display within a PMax campaign as the system rebalances toward target, attribution gets noisier at exactly the moment you need it to be clean. Anyone leaning heavily on PMax should assume the September reporting picture will require more forensic work than usual.

And there's a reasonable counter-read here: that the change is genuinely a technical improvement, that predictable performance at scale is worth more than opportunistic overdelivery, and that mature accounts will benefit from tighter alignment between stated intent and actual behaviour. That's a fair position. It's just not one that helps the accounts that were quietly benefiting from the old behaviour.

The bigger point

This is, in a small way, part of the same pattern I keep writing about — the measurement layer that marketing runs on quietly changing shape underneath us. Every platform is renegotiating the deal, usually in directions that favour the platform. Google's Q2 numbers already told us the search business is entering a compression phase. This is compression at the ads layer: the same volume of clicks, priced closer to what advertisers said they'd pay, with the structural slack removed.

The advertisers who survive this well won't be the ones with the cleverest bid strategies. They'll be the ones who actually know what a customer is worth to them. That's an unglamorous, un-AI, un-hypeable answer. It's also the only one that holds up when the platform decides your targets should mean what you said they meant.

August 17 isn't a crisis. It's a deadline for doing the maths you should already have done.