How Smart Bidding works
Rather than one bid for a keyword, Google calculates a bid for each individual auction as it happens. It uses signals a person could never react to in time: the device, the location, the time of day, the exact query, the browser and language, whether the searcher is on one of your remarketing lists, and how combinations of those things have converted before. Bids rise where a conversion looks likely and fall where it does not.
The strategies differ in what they are told to chase. Maximise Conversions goes for volume within the budget, Maximise Conversion Value goes for total value, and each can carry a target — a cost per conversion or a return on ad spend — that steers how aggressive it is. All of them learn from the conversions you send, which makes conversion tracking the foundation rather than a reporting detail.
Why Smart Bidding matters
Auction-time signals are the point. Two people typing the same words at the same moment are not equally likely to buy, and a manual bid treats them identically because it has to be set in advance. That advantage grows as accounts get larger and the number of meaningful combinations goes beyond what anyone could manage by hand.
It also changes where your effort goes. You stop adjusting bids and start steering: what counts as a conversion, what a conversion is worth, how the campaigns are grouped, how much budget is available and what the target says. Those decisions matter more under automation, not less, because the system will pursue whatever you told it to value with complete literal-mindedness.
Where Smart Bidding goes wrong
Most failures are tracking failures. Counting every form view, every phone-number click and every newsletter sign-up as a primary conversion teaches the system that cheap, meaningless actions are the goal, and it will duly find you more of them. Duplicate conversions, a thank-you page that reloads, or a tag firing on the wrong pages all corrupt the same signal.
Thin data is the second problem. A campaign that produces only a trickle of conversions gives the model very little to learn from, which is a genuine constraint on smaller budgets and narrow local markets. Constant tinkering is the third: every meaningful change restarts the learning phase, so an account edited every day never gets past it. And no bidding strategy repairs a weak offer, a broken landing page or targeting aimed at the wrong people.
How to use it well
Fix measurement first. Decide which single action means money to the business, mark that as the primary conversion, and demote the rest to secondary so they are visible but not chased. Then give the strategy enough data to work with by consolidating fragmented campaigns instead of splitting them further.
Change targets in small steps and leave them alone long enough for the results to mean something. Keep the budget steady while it settles, because sharp budget swings unsettle the model as much as target changes do. Judge the outcome on cost per acquisition and lead quality over a full buying cycle rather than on a bad week.