Target CPA step-downs: 10% at a time or cut it in one go?

Our Target CPA on a lead gen Search campaign is set at 60 but the actual CPA runs around 85, and someone on the team wants to drop the target to 50 today to force it down. I'd step it down about 10% every week or two instead. Smart Bidding needs room to find the converters it's losing right now. Cut hard and the auction filters out anything borderline, so volume can fall off a cliff and you end up with a smaller pool that isn't even cheaper. Say a dental clinic drops from 60 to 40 overnight: impressions can crash and the learning restarts on far fewer signals.

The counterargument has some weight, though. If budget is burning on clicks that never convert, a slow ramp just pays for more waste. So pull the search terms first, and if the waste is obvious, a bigger cut is justified. How are you handling this on your accounts?

Honestly, it depends on conversion volume, and I'd check that before picking a method. If the campaign gets a steady stream of conversions, say 30 or more a month, a 60 to 50 cut mostly just reshuffles the auction, and Smart Bidding adapts within a week or two. If it only gets a handful of leads, a 10% weekly ramp can take a month before you learn anything, and each step is mostly noise. In that case, make one cut, hold for two weeks, and judge it on search terms and lead quality rather than daily CPA.

The volume split makes sense to me, but I'd check conversion lag before anyone cuts. Lead gen conversions often take a week or more to get qualified, so a 30-day CPA of 85 can be missing recent leads that haven't been counted yet. It can also be padded with leads that were never going to close. If your report runs on conversion-entry date, switch to conversion date and look at a longer window. The gap to 60 might shrink, and that changes how big a cut you actually need.

Lag is part of it, but there's another thing pushing that 85 in the other direction. If the target is 60 and the actual is 85, Smart Bidding may be chasing conversions that are still unqualified, which means it's optimizing on the wrong signal. Check whether the conversion action counts raw form submits or only qualified leads (a CRM import or offline conversion upload). If it's form submits, the bidder is happily paying for junk, and no step size fixes that. Switching the primary action to the qualified one often moves the real CPA more than any target change.

I wouldn't read the 85 as proof of junk submits. If the bidder is optimizing on raw form submits, the CPA on those same submits should settle near 60 once it's learned. A gap that size usually means one of three things: the campaign is budget-capped, it's still in learning, or the report is looking at a different conversion action than the one bidding. Confirm which action is actually driving bids before you switch anything. And a qualified-lead action will carry much lower volume, so the target needs its own step-down too.

Fair, and one rule keeps this from getting messy: change one lever per learning cycle. If you confirm the bidding action, switch to a qualified-lead action, and move the target all in the same fortnight, you won't know which change moved the number. And with a client, that's the conversation you don't want, because they'll see lead volume drop and ask why.

So I'd keep it to two steps. First, check whether the campaign is budget-capped or still learning, since cutting the target won't help a budget that's already maxed out. Second, if the bidding action is right, step the target once and leave it alone for two weeks. Write the date and the change in the report before you make it, so the volume dip already has an explanation attached.

Switching the primary conversion action is a much bigger reset than moving a target. The bidder has to relearn on a different signal, whereas a 10 to 15% target change mostly just shifts how you show up in the auction. So if you're keeping the same action, a small step can often stay inside the same learning window.

Priya, I wouldn't treat 10 to 15% as a rule. Neither Google nor Microsoft publishes a threshold for when a target change restarts learning, so it's just a rule of thumb. The easy check is the bid strategy status after you make the change, since it will show whether the strategy has gone back into learning. Also keep in mind that a target cut changes bids on the marginal auctions first, and that's where volume quietly drops. So watch lost impression share (rank) for the next two weeks before you decide the step was small.

Theo, the status check is a good call. I'd add a simple rule on top: if a campaign gets fewer than about 15 conversions a month, make one cut and then wait a full conversion-lag window before you touch it again. On a thin account every step is mostly noise, and that noise is what makes everyone second-guess the target.

Agreed on one cut, but I'd measure the wait in conversions, not days. At 15 a month you're getting roughly four leads a week, so a 10% CPA move is well inside the noise over any two-week window. A full lag window only helps if it also holds enough leads to show a difference, otherwise you're just waiting on the calendar. On a thin account, which would you trust more to judge the cut: CRM close rate or the raw submit count?