.png)
For about four years I ended every Sunday the same way: a bid spreadsheet, two coffees, and roughly 200 manual CPC edits pasted into Google Ads Editor before Monday's traffic hit. I was good at it. I also now think most of that work was theatre. Not because I was lazy about it, but because a keyword-level bid set on Sunday night is a single number applied to every auction that keyword enters for the next seven days, and Google is pricing those auctions individually using signals I was never allowed to see.
If you're still on manual CPC, or you've got a folder of bid scripts holding the account together, you already suspect the switch is overdue. What you actually want to know is the order of operations: what to fix first, what to move first, how long the mess lasts, and how to get back if it goes badly. That's what this is. I'll also tell you who should not migrate this month, because there are accounts where switching to Smart Bidding makes performance worse for a reason that has nothing to do with the algorithm.
Here's the mechanism, because the mechanism is the whole argument. When someone searches, Google's auction has access to a stack of signals about that specific query: device, exact search phrase, time of day, physical location, browser, operating system, whether that user is on a remarketing list, what they did on your site three weeks ago. Your manual bid knows none of that. It's one number, negotiated in advance, applied identically to a 9am desktop searcher in your best metro area and an 11pm mobile searcher in a state where you don't ship. Bid modifiers were the old patch for this, and stacking device times location times audience times daypart gets you maybe a dozen crude buckets. Auction-time bidding effectively prices every impression on its own. That gap is not a matter of skill. You cannot out-spreadsheet a system that gets to look at data you're not shown.
Manual bidding lets you compensate for bad tracking with judgment. You know the "Contact" conversion is mostly spam, so you discount it in your head and bid accordingly. Smart Bidding has no head. It optimizes toward whatever you told it counts, at full face value, at speed. Hand it a conversion action that fires on every newsletter signup and it will go buy you newsletter signups, efficiently, until you notice revenue is flat and CPA looks fantastic. Every bad migration I've watched failed here, not at the bid strategy dropdown.
So before you touch bidding, spend a day on the tracking. Open Tools, Conversions, and go line by line:
Google's published minimums have been softened over the years, and I'd ignore them. My working thresholds, learned by breaking accounts: 15 conversions in the trailing 30 days per campaign before I'll run Maximize Conversions without a target, 30 before I'll set a target CPA, and 50 or more with reasonably consistent order values before I'll set a target ROAS. Below 15, the algorithm is guessing from noise, and you'll see the classic failure pattern: spend concentrates into two or three keywords that happened to convert twice last week. If your campaigns are individually thin, consolidate before you migrate. Four campaigns at 8 conversions a month each are useless separately and workable as one. Consolidation is unglamorous and it does more for bidding performance than any target you'll ever type into a box.
The menu confuses people because Google renamed things. Target CPA now lives inside Maximize Conversions as an optional target box, and Target ROAS lives inside Maximize Conversion Value the same way. Same algorithm, different label. The other thing worth knowing before you plan a gradual migration: Enhanced CPC, the old halfway house where you kept your manual bids and let Google nudge them, has been retired. There's no gentle intermediate step left. You go from bids you control to bids you don't, in one move, which is exactly why the pilot campaign matters.
My default for a first switch is Maximize Conversions with a target CPA set at your demonstrated 30-day CPA. Not your desired CPA. Your actual one, rounded up slightly. Here's the mechanism people miss: under manual CPC, most accounts underspend their daily budgets, sometimes by 20 to 40%, because the bids are too conservative to win the volume the budget would allow. Maximize Conversions has one job, which is to spend the budget. Flip a campaign over on a Monday with no target set and you can wake up Tuesday to a full budget spent at a CPA nobody signed off on. The target isn't there to improve performance in week one. It's there to stop the algorithm from discovering how much you were actually willing to pay.
Target ROAS is a second migration, not the first one. It needs stable conversion values flowing in, which means real revenue passed back per transaction and enough volume that a couple of outsized orders don't drag the average. If you sell one product at one price, value-based bidding gives you nothing that a CPA target doesn't. If you sell across a 10x price range with different margins per line, it's worth the work, and you should feed margin-adjusted values rather than gross revenue, or the system will happily optimize toward your least profitable bestseller.
This is the step that bites technical advertisers. A Google Ads automation script that writes keyword-level bids doesn't error out when the campaign moves to Smart Bidding. It runs, it reports success, and the bids it sets are ignored, because auction-time bidding doesn't read the max CPC field anymore. So you get a script that appears healthy, a log full of green, and no effect on anything. Worse are the rule-based pruning scripts: the one that pauses any keyword over your CPA ceiling, or with 100 clicks and no conversions. During the first three weeks after migration those scripts will amputate exactly the exploration the algorithm is doing on purpose, and you'll conclude Smart Bidding is expensive when what actually happened is your own code kept cutting its legs off mid-stride.
Go through the folder before migration day and sort it into three piles:
One honest caveat about the keep pile: monitoring scripts are still code, and code you now depend on more than before. Mine used to break quietly whenever Google renamed a report column, which is a fun thing to discover eleven days late. If you keep them, put a heartbeat on them: a daily email that fires whether or not there's anything to report, so silence means broken rather than fine.
Decide your kill criteria while you're calm, because on day six you will not be. Mine are simple: if at day 21 the campaign's CPA is more than 25% above the pre-migration baseline on comparable or lower conversion volume, something is wrong and I intervene. If conversion volume is up 40% and CPA is up 10%, that's not a failure, that's you buying more at a slightly worse rate, and whether you keep it depends on your margins rather than on your feelings about the CPA column. Write both numbers down before you switch. The version of you watching a $180 CPA on Wednesday morning is not qualified to decide what counts as failure.
Intervening rarely means reverting. Reverting to manual CPC throws away everything the strategy learned and costs you another two weeks when you inevitably switch back, so my first move is almost always to loosen rather than retreat: raise the target CPA by 15 to 20% if volume has collapsed, or add a portfolio strategy with a max CPC limit if a handful of expensive queries are eating the budget. Actual rollback is for real breakage, like conversion tracking failing mid-migration. And if you do want that door open, export every keyword's max CPC to a CSV before you flip the switch. Google won't hand your old bids back to you afterwards, and rebuilding four years of manual bids from memory is a week of your life you won't enjoy.
Once you're through the first campaign, the monitoring cadence should get boring on purpose. Week one: a daily two-minute look at spend pacing and disapprovals, nothing else. From week two: a weekly search terms review with negatives added by hand, and a check that your Primary conversion actions are still firing at the volume you expect. From month two: tighten targets in 10 to 15% steps, one campaign at a time, never two levers in the same week. The discipline that made you good at manual bidding still applies. It just moves up a level, from setting the bids to setting the constraints the bidding operates inside.
Here's the part that surprised me most after I finally migrated everything: the bidding got better and my workload barely moved. Smart Bidding optimizes bids. It does not write the ad, decide which campaign deserves the next $2,000, notice that a search theme is drifting toward the wrong buyer, mine your search terms, or fix the landing page that every one of those expensive clicks lands on. Auction-time bidding will happily buy the perfect click for a page that answers a different question than the one the user typed. You've automated the one job that was mechanical and kept all the jobs that were judgment, plus a few that are just tedious. That's a real improvement. It's also why "we switched to Smart Bidding" stopped being an achievement somewhere around 2019.
I work on the product side at groas, so weigh this accordingly, but the reason I ended up here is that the remaining pile is what we build for. The models take the same actions a human takes in the account, bids, budgets, keywords, targeting, running around the clock rather than on a Sunday-night cadence, and they were trained on more than $500 billion in search ad spend. The piece I'd have wanted most back when I was doing this by hand is the landing page side: groas takes your existing page and deploys dynamic versions that match what each searcher actually asked for, so the arm-hair-trimmer query lands on an arm-hair-trimmer page instead of a category grid. Bidding can't fix that mismatch. Nothing in the bid strategy menu can. As for whether it works, the numbers I'd point at are the unglamorous mid-market ones: a testing kits company at $10-20k a month that cut CPA 27% and scaled LATAM volume in 30 days, a dental practice at similar spend that cut CPA by a quarter in month one.
If you're still bidding by hand, though, don't start there. Start with the boring thing on Monday morning: open your conversion actions, and find out how many of the ones marked Primary you'd actually be willing to buy more of at your current CPA. In most accounts I've audited, it's about half. Fix that list first, migrate one campaign three weeks later, and you'll get more out of the switch than any target you type into the box. My Sunday spreadsheet was never the thing making that account work. It just felt like it was.
