Google is now taking your target literally

The Smart Bidding change I wrote about last week went live yesterday. If you missed it, the short version: a budget-capped campaign on Target CPA or Target ROAS used to quietly beat its own target, because Smart Bidding spent the capped budget on the cheapest auctions it could find. From 17 August, it aims for the number you entered instead. The earlier piece has the full pre-change checklist and the four routes out.

That piece was written for the people who manage accounts. This one is for the accounts themselves, specifically the ones spending £500 to £3,000 a month, where Limited by budget is not an edge case but the permanent state of affairs. Because the uncomfortable irony of this change is that the accounts with the least data and the least room for error are the ones with the biggest cushion to lose.

01Why the smallest budgets had the biggest cushion

Think about what a capped budget actually does to a bidding algorithm. It says: you may spend £20 today, and there is £60 of demand you could buy. So it chooses. And left to choose, it picks the auctions where the odds are best and the price is lowest, which is exactly how a £40 target produced £22 leads for a year without anyone touching it.

The tighter the cap relative to demand, the more selective Smart Bidding could afford to be, and the wider that gap grew. A £5,000 a month campaign that occasionally bumps its budget had a small cushion. A £600 a month campaign that hits its budget by lunchtime had a huge one. It has been living off it without knowing.

The smaller the budget, the bigger the cushion. And the cushion is what just went.

None of this means the change is unfair. A target that only holds because Google was quietly under-delivering against it was never really a target. But it does mean the accounts least equipped to absorb a jump in cost per lead are the ones most likely to see one.

02What the next four weeks actually look like

Nothing dramatic happens on day one. Google is staggering the rollout across accounts, so your campaigns may not shift for a week or two, and when they do it is a drift, not a step. Here is the rough sequence on a campaign that has been sitting comfortably under its target.

Give it two conversion cycles before you conclude anything. On a business where an enquiry takes a week to become a booking, that is a month, not a weekend. The first few days after any bidding change are noise, and reacting to noise is how a manageable drift turns into a restart of learning.

03Do you need a target at all?

This is where I want to sharpen something from the earlier piece. I said that dropping the target on a small budget can be a worse outcome than the change you are trying to avoid, and in general that is still true. But there is a specific case worth spelling out, and it is the case most small accounts are in.

If your campaign has spent its full budget every day for the last 90 days, the budget is already the brake. Under the old behaviour the target was doing useful work: it kept Smart Bidding picky within that budget. Under the new behaviour, a target on a capped campaign becomes a price ceiling Google is now content to spend up to. You had two brakes. One of them just stopped braking.

So the realistic options narrow to two:

• Tighten the target to what the campaign has actually delivered over 90 days, with a little headroom. You keep a ceiling, and the ceiling is real. This is the default answer and it is the right one for most accounts.
• Drop the target and run Maximise Conversions with the budget as the only constraint. Same spend, and Smart Bidding goes back to hunting volume within it. The catch is what "conversions" means in your account. If a call click, a page view and a genuine enquiry all count the same, and in a lot of small accounts they do, Maximise Conversions will happily fill the budget with the cheap ones. Do this only if your tracking counts real leads and nothing else.

Whichever you choose, move in 10 to 15 per cent steps rather than one edit, and leave everything else alone for the fortnight that follows. A target change, a budget change and a new ad in the same week leaves you with three suspects and no evidence.

04The September trap

There is a timing problem nobody is talking about. The change lands in the last two weeks of August, and for a lot of service businesses September is when demand comes back. Schools return, people book the things they put off over summer, and competitors who paused for August switch their campaigns back on. CPCs move in September anyway.

So when your cost per lead in mid-September is 30 per cent above mid-August, you will have two explanations on the table, and the honest answer is usually a bit of both. Annotate the account with the date the change visibly reached your campaigns, and compare against last September, not against August. If you don't have a last September, compare against your target and your break-even, which is what you should have been doing anyway.

Compare against last September, not against August. August is the month you were being subsidised.

05Google is about to ask you for more money

Alongside the bidding change, Google has rolled a missed opportunity estimate into the Recommendations tab: modelled lost clicks and conversions from being under-invested, with a suggested fix that almost always involves spending more. The Bid Target Adjustment Tool sits next to it, offering a one-click reset of your target to recent actuals. Neither is malicious. Both are Google's answer to a question you may not have asked.

Limited by budget is a status, not a diagnosis. The right question is what the next £10 a day would actually buy, and whether that lead is profitable at the marginal price rather than the average. On some small accounts the answer is yes, and this change makes the case for more budget easier to argue, because the campaign is now visibly profitable at a number you can defend. On others the answer is that the campaign was only ever profitable at the cheap end of the auction, and more budget just buys the expensive end. You cannot tell which you are from a recommendation card. You can tell from a month of clean data against a target you chose.

06If someone else runs your ads

You do not need to log in to know whether this was handled. Three questions will tell you:

• Which of my campaigns were budget-limited and running on a target when this landed?
• What was the gap between the target and what we were actually paying?
• What did you change, on what date, and is there an annotation on the account?

If they can answer all three in a paragraph, you are in good hands. If the reply is "nothing changed", it may well be true. Campaigns that were never budget-limited, or run on Manual CPC, or on Maximise Conversions with no target, are untouched. But ask them to show you that rather than say it. It is a two minute screenshot.

---

← Back to all articles