Preview summaries on ad changes: read the numbers or trust the wording?

With a preview-first setup, read the before and after values and skip the summary, because the wording can sound fine while the number is wrong. Say a dental clinic asks the AI to cut mobile bids. The summary reads "reduce mobile bid adjustment," which sounds reasonable, but the diff shows -90% where they meant -9%. That's a very different account.

The obvious objection is that checking every diff makes the preview pointless for small changes, and people stop looking after a week. Fair, so I'd split it: read the numbers closely for budget and bid changes, and just skim the rest. Curious how others draw that line.

Agree on the split, but I'd draw the line by what a wrong number can do, not by the type of change. A device bid adjustment is a percentage on the base bid, and Google accepts anything from -90% to +900%. So a typo that turns it into -90 is still a valid change, and nothing will flag it.

What I'd add is the other side of the number: what share of conversions that device brings in. If mobile is 70% of a dental clinic's calls, a -90% cut isn't a small tweak, it's closing most of the front door, and the diff alone won't show you that. For a device that carries little volume, skimming is fine. For one that carries most of the conversions, read the number against its share before you confirm.

Checking the number against its share of conversions makes sense, and the other thing I'd check is scope, which the change summary tends to blur. On TikTok, a line like "update budget" or "pause ads" can apply at the ad group level or the campaign level. If the campaign uses campaign-level budget, a change that looks like it only touches one ad group can shift spend across every ad group under it. So the same number can be a small tweak on one ad group and a big move on the parent campaign. Before you read the value, look at the target ID in the diff, not just the field name, because that's where the wording is usually vaguest.

Microsoft Ads has the same trap under a different name. Campaign budgets there can be shared, so one budget can fund several campaigns, and an edit that looks like it only touches one campaign may be pulling from a pot the others use too. A 20% raise on a shared budget moves every campaign attached to it, and the change summary rarely tells you that. So before you read the amount in the diff, check whether the budget is shared.

Agree, and the number in a shared-budget diff is the size of the pot, not the size of the change for the campaign you were looking at. A summary saying "raise budget from 100 to 120" on a budget shared by three campaigns just means those three have 20 more a day between them. The campaign you thought you were helping might get almost none of it, since the extra spend goes wherever the bidding finds volume.

So before confirming, check how many campaigns are attached to that budget ID and what each one spent last week. If one campaign already eats most of the pot, a 20% raise mostly lands there. And if the attached list is longer than you expected, that count matters more than the percentage.

I'm not sure the last line holds up. The number of attached campaigns is a weak signal on its own. A budget shared by ten campaigns, where nine are paused or nearly idle, behaves like a one-campaign budget. Sort the attached campaigns by spend and look at the top one's share. If one campaign takes 85% of the pot, a 20% raise is effectively a raise for that campaign, whatever the list length says.

Also look for any attached campaign you didn't expect, because that's the one that quietly soaks up the extra money. And don't trust last week's spend alone. A campaign that was paused for a few days or ran a promo will look smaller or bigger than usual, so pull two or three weeks before you rely on the split.