Switched to tCPA and conversions fell off a cliff: hold the target or loosen it?

Take an account that moves from Maximize Conversions to tCPA at 40 euros, and conversions fall from 25 a week to 9 in the first ten days. The usual advice is to sit tight because learning can take a week or more, but I don't think that's the right default at low volume. With 9 conversions a week the algorithm barely has anything to learn from, so waiting mostly means paying for a few weeks of thin traffic. I'd loosen the target by 15 to 20 percent for a couple of weeks, check the search terms for drift, and only tighten once volume recovers. Curious whether others hold the line or back off faster when volume is this low.

Theo, I'd skip the fixed 15 to 20 percent loosening. That number isn't tied to anything in the account. A better anchor is what the account actually paid per conversion under Maximize Conversions. Say a dental clinic was running at 55 per lead before the switch. A tCPA of 40 asks the bidder to cut cost per lead by almost a third in one step, so it stops bidding in auctions it used to win, and that alone could explain the drop to 9. A target around 50 or 55 is a different test, because it matches what that traffic has actually delivered.

I'd also avoid moving the target repeatedly. Change it twice in a few weeks and you've restarted the learning twice, and at 9 conversions a week that noise is the real cost. Pick a realistic target, hold it for three or four weeks, then tighten in small steps. Waiting only costs money when the target was never realistic to begin with.

Anchoring to the historical CPA is the right call. Just check whether the 40 target was set against the lifetime average or the last 30 days, because a few expensive months can make the old CPA look worse than it really was. And if the drop started right at the switch, look at impression share lost to budget before blaming the bidder. A tight tCPA can quietly cap spend.

I'd start with impression share, because under tCPA the bidder only bids where it expects to hit your target, while Maximize Conversions spends the whole budget regardless. So a budget that looked fine before can suddenly become the bottleneck after a switch. As a rule of thumb, if lost impression share due to budget jumps by more than about 20 points, fix the budget before touching the target. Loosening the target won't help when the budget is the cap, and you'd be adjusting the wrong lever.

Checking budget first is right, but split the lost impression share into budget and rank before you decide anything. Under tCPA a lot of the loss can come from rank, which means the target is the cap and a bigger budget won't fix it. If budget lost is the larger number, raise the budget. If rank lost is larger, move the target. At 9 conversions a week those percentages swing a lot, so read two or three weeks of data before trusting the 20-point line.

Honestly it depends on whether this campaign shares a budget with others. Say a dental clinic runs a brand campaign and a service campaign off one shared budget. The tCPA campaign can pull spend away from the other one, so the budget-lost number may just reflect that shift and not a real cap. Raising the shared budget then mostly feeds the campaign you just changed. So look at lost impression share per campaign first, and only raise the budget if the losses sit in the campaign you moved.

Marco, the budget question changes depending on whether the tCPA campaign is the one losing impressions or the one taking them. If it's gaining share from the brand campaign, the conversions you lost may be coming from brand traffic that got starved, and that tells you something different than a campaign that's just capped. So it comes down to whether total conversions across both campaigns fell or only moved between them. Look at the account-level number before you decide the tCPA is the problem.

Agreed on the account-level check. Before you compare totals, make sure both periods count the same conversions. A bidding switch doesn't change your goals, but the conversion column can still shift if the campaign now reports primary actions only where it used to report all conversions. Also, the first ten days after a change are partly lag, because conversions from recent clicks haven't all been recorded yet. That makes the early numbers look worse than they'll end up.

Agreed on the conversion column. The thing that bites most often is a change in conversion action sets. If the old campaign counted all conversions and the tCPA one only counts primary, a drop from 25 to 9 can be mostly bookkeeping. Compare the Conversions column against the conversion action names in both periods before you read anything into the target.

Agreed, and the conversion action set is the first place I'd look. The "Conversions" column can be switched to a different attribution or counting setting, and moving from Maximize Conversions to tCPA doesn't change that. But the "Include in Conversions" flag on each action does. If a secondary action got turned off for one campaign during the switch, the drop is pure bookkeeping. Pull the breakdown segmented by conversion action for both periods before you touch the target at all.