.png)
Last month I watched a client screen-share his way through campaign setup, hunting the bid strategy dropdown for Target CPA like it had personally wronged him. It wasn't there. Neither was Target ROAS. His conclusion, delivered with full confidence: Google had killed both strategies and he'd need to rebuild everything before the quarter ended. He was wrong, but I understood the panic. Google renames things the way other companies ship features, and every rename sends a wave of advertisers back to the search bar trying to find a setting that no longer answers to its old name.
Here's the answer you're looking for, so you can close the other tab: Target CPA is now Maximize Conversions with an optional target CPA set inside it. Target ROAS is now Maximize Conversion Value with an optional target ROAS. That's the whole change. Open campaign settings, choose Maximize Conversions, tick the box to set a target cost per action, and you're running exactly what you used to run under the old name. Same mechanics for Maximize Conversion Value: pick it, set your target ROAS, done.
Under the hood, nothing moved. The Smart Bidding algorithm making auction-time decisions is the same one that powered the standalone strategies, reading the same signals: device, location, time of day, query context, the usual pile. Google merged the labels because two of them were redundant. If you can set a ceiling on CPA inside Maximize Conversions, you don't need a separate strategy called Target CPA taking up menu space. The mechanics didn't change; the menu did. And honestly, the rename is the least interesting part of this story. The part that actually costs people money is what comes next: when setting a target helps you, and when it quietly strangles your volume.
Strip the names away and there are only two base instructions you can give Google's bidding. Maximize Conversions says: spend my budget and bring back as many conversions as the auctions allow, whatever each one costs. Maximize Conversion Value says the same thing weighted for revenue, so it chases the $300 cart instead of the $30 one. Setting a target adds a single clause to either instruction: do that, but keep the campaign averaging $45 per conversion, or 400% ROAS, or whatever number you type into the box. That box is the whole story. The standalone strategies were never anything more than this clause with a dedicated menu entry.
The word to pay attention to is averaging. A target CPA is not a ceiling on any individual conversion, and a target ROAS is not a floor under any individual sale. Every auction, the algorithm forecasts how likely that specific impression is to convert and what the conversion is worth, then bids up or down so the campaign's running average drifts toward your number. You will see conversions at $28 sitting next to conversions at $71 in the same week, with a $45 target set. That is not the strategy failing. That is the strategy working exactly as specified, because what it promised you was the average, not the line items. Advertisers who don't understand this spend their Mondays pausing keywords that were never the problem.
Here is the mechanism nobody puts in the sales deck: a target is a throttle, and it only throttles one way. Set your target at or slightly above your historical CPA and nothing dramatic happens. The algorithm keeps entering the auctions it was already winning and your average holds. Set it 30% below history because you'd like costs to come down, and watch what actually occurs: the system forecasts that most auctions will miss your number, declines to enter them, your impression share slides, and volume falls off inside two weeks. The bid strategy did precisely what you told it to do. You told it to stop buying. So the rule I give clients: if your number comes from your margins, set it. If it comes from wishful thinking, leave the target box empty and let the strategy run free until your data earns you a real one.
The process is boring, which is why nobody does it. Pull your last 30 days of conversion value and cost for the campaign. Divide value by cost. That number, adjusted for how stable your volume has been, is your starting target. If the campaign returned 380% ROAS last month, you set 350%, maybe 380% if you feel disciplined. You do not set 600% because a LinkedIn post told you 600% is what good looks like. The algorithm has no idea what you want. It only knows what the auctions have historically delivered at various bid levels, and when you hand it a number 60% above that history, it concludes the rational move is to stop entering auctions. Cause, then effect: unrealistic target, dead campaign, panicked email to me on a Friday.
One floor matters more than any benchmark: your break-even ROAS, which is just 1 divided by your gross margin. Sell at a 40% margin and break-even is 250%, meaning every auction dollar has to return $2.50 in revenue before you've made a cent. Your target needs to sit comfortably above that floor and close to your demonstrated history. If those two numbers conflict, if break-even is 250% and your campaigns have never once cleared 200%, the bid strategy is not your problem. Your pricing, your margin, or your conversion rate is, and no dropdown in Google Ads fixes any of those. This is the diagnosis most accounts need and almost none get, because 'renegotiate your supplier costs' doesn't fit in a bid strategy menu.
Three things reliably knock a target-based strategy sideways, and they're all things you did, not things Google did. Big budget jumps are the first: roughly double a campaign's daily budget and the algorithm has to spend it somewhere, which means entering auctions it previously judged unprofitable, which means your ROAS dips while it recalibrates. The second is target changes themselves. Adjust more than about 20% in one move and you effectively restart the learning period, which runs one to two weeks on a healthy account. The third is anything that changes your conversion rate outside the auction: a site redesign, a price increase, a sale weekend. Smart Bidding adapts to seasonal bidding patterns on its own, but a planned 72-hour promotion is exactly the case for a seasonality adjustment, the one setting built precisely for short, known conversion-rate spikes. The practical rule: one change at a time, sized under 20%, with two weeks of hands-off between them. Tinker daily and you are the learning period's full-time job.
One thing the rename didn't touch: portfolio bid strategies. You can still group multiple campaigns under a single shared target CPA or target ROAS from the Tools menu, and in the right situation it's the better setup. The mechanism is signal pooling. Smart Bidding learns from conversion data, and a campaign pulling 12 conversions a month is statistically starving on its own. Bundle three thin campaigns under one portfolio target and the algorithm learns from their combined 40, which is enough to exit learning and hold a stable average. The catch is that the portfolio optimizes the group's average, so mixing campaigns with different economics invites it to hit the number by over-delivering on your cheap conversions and quietly starving your most profitable campaign. Portfolio for similar margins and similar conversion definitions. Separate targets the moment one campaign's economics genuinely differ.
The rename itself broke nothing, but it triggered a wave of self-inflicted damage. The patterns I keep seeing:
Everything I've described so far is a loop: pull 30 days of data, set the target near history, adjust in 20% increments, wait two weeks, repeat. It requires actual judgment exactly twice — once when you derive the initial number from your margins, and once when you decide whether a dip is noise or signal. The rest is a schedule. I know, because I billed for that schedule for years, and I stopped being able to justify it around the time the tooling got good enough to run the loop without me. What the deck calls 'ongoing bid strategy optimization,' I call a cron job with an invoice.
So here's the split I'd draw. Hand off the mechanical layer: target calibration, budget pacing, learning-period discipline, the daily restraint of not touching things. Keep the economics. No bidding algorithm knows your supplier raised prices and your gross margin dropped six points, or that half of last month's 'conversions' were a duplicated purchase tag, or that your 400% average is being quietly propped up by brand search while your generic terms lose money on every sale. The target is only as honest as the conversion data and margins behind it, and those stay your job whether you run the account yourself or not.
Full disclosure, since it colors the next sentence: I work with groas, and the fully managed side exists precisely because of this split. The engine runs the calibration loop around the clock — bids, budgets, target adjustments, trained on $500B+ in ad spend — while a dedicated strategist owns the economics conversations with you, for a flat monthly fee with no percentage of spend and no lock-in. You can absolutely do all of this by hand; nothing in this article is beyond a disciplined operator. But if your current arrangement is someone charging you 15% of spend to type a number into that target box twice a month, the rename was never the thing costing you money.
