Quality Score
The relevance signal Smart Bidding inherits when entering each auction. Bid strategy does not replace structural alignment.
Read the entry →Stan Consulting · Marketing Atlas · Reference · Google Ads
Google's machine-learning bid strategies. Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value. The cost of giving the algorithm control.
Section 02 · Quick definition
Smart Bidding is the family of automated bid strategies inside Google Ads that uses machine learning to set bids at auction time, with the goal of optimizing toward a conversion outcome the operator selects. The primary strategies are Target CPA, Target ROAS, Maximize Conversions, and Maximize Conversion Value, plus the older Enhanced CPC strategy that predates them. Smart Bidding consumes the account's conversion signal, contextual data Google holds at query time, and historical performance to set per-auction bids the operator does not see. The strategy decides what to bid; it does not decide what to bid against.
Section 03 · Why it matters
Smart Bidding matters because it changed where the operator's control sits. In a manual-CPC account, the bid was the operator's primary lever. In a Smart Bidding account, the bid is delegated; the remaining levers are the conversion signal, the audience targeting, the keyword set, and the structural decisions about what the algorithm is allowed to allocate against. The work moved upstream.
It also matters because the failure mode is non-obvious. A Smart Bidding strategy fed a corrupted conversion signal does not error; it optimizes against the corruption. An account with double-counted conversions, soft events weighted as sales, or unrepresentative buyer segments included in the conversion definition produces a bidding model that finds more of the wrong thing, faster than a manual operator could.
For an operator, the practical stake is that Smart Bidding fails on under-structured accounts. The algorithm optimizes against the signal you give it; if the signal is junk, the spend is junk faster than under manual bidding. The remedy is upstream of bid strategy.
Section 04 · How it works
Smart Bidding is a family of strategies. Each one optimizes toward a different specification, but all of them share the same underlying mechanic: at the moment of an auction, the algorithm sets a bid using the account's conversion data and Google's contextual signals.
Sets bids to produce conversions at an average cost the operator specifies. Suited to lead-generation accounts where each conversion has roughly equal value.
Sets bids to produce a return on ad spend the operator specifies, using the conversion-value signal sent into the account. Suited to e-commerce and revenue-weighted lead accounts.
Spends the campaign budget to produce as many conversions as possible. No target; the budget is the constraint.
Spends the campaign budget to produce as much conversion value as possible. Used when revenue per conversion varies and the operator wants the algorithm optimizing toward dollars rather than count.
The older strategy that adjusts manual bids upward or downward based on conversion likelihood. Limited adoption since the rollout of the four primary strategies, but still present in older accounts.
The conversion actions selected for the bid signal, conversion values when sent, historical account performance, real-time auction context (device, location, time, query, audience), and seasonality data Google holds.
Strategy changes trigger a learning period during which the bidding model rebuilds. Performance is unstable for one to two weeks; comparing the new strategy against the old before learning completes is unreliable.
Section 05 · Common misunderstandings
“Smart Bidding will figure out the structure for me.”
It will not. The algorithm sets bids inside the structure the operator built. Match types, ad groups, conversion definitions, and negatives are upstream of bidding and untouched by it.
“Switching strategies will fix performance.”
Switching strategies triggers a learning period and changes which signal is being optimized. It does not change a corrupted conversion signal, a permissive keyword set, or an under-curated negative list.
“Smart Bidding does not need conversion tracking to work.”
Three of the four primary strategies require conversion data. Without conversions or with broken tracking, the model has nothing to optimize against and the spend allocation degrades immediately.
“Tighter targets always produce better results.”
Aggressive Target CPA or Target ROAS values throttle delivery. The model bids lower, wins fewer auctions, and starves the campaign of the volume it needs to learn. Targets are tuned, not chased.
“Smart Bidding handles negative keywords.”
It does not. Smart Bidding decides what to bid on the queries the campaign is eligible for. Negative keywords decide eligibility. The two surfaces answer different questions.
Section 06 · Diagnostic questions
Which Smart Bidding strategy is each campaign on, and why was that strategy chosen?
What conversion actions feed the bid signal, and at what counting and weighting rules?
Are micro-conversions, soft events, or non-revenue actions included in the optimization signal?
What is the conversion volume per campaign per 30 days, and is it sufficient for the chosen strategy to learn?
When was the last strategy change, and was the comparison made before or after the learning period closed?
How are Target CPA or Target ROAS values set, reviewed, and adjusted?
For each campaign on Smart Bidding, what structural prerequisites (negatives, conversion hygiene, keyword set) were verified before the strategy was activated?
Section 07 · Related Atlas entries
Section 08 · Five Cents
Smart Bidding fails on under-structured accounts and the failure mode does not look like a failure. The campaigns hit their targets, the dashboard turns green, and the budget grows. What grew was the algorithm's skill at finding more of whatever the conversion signal told it to find. If the conversion signal includes add-to-carts at full weight, the algorithm finds more cart-abandoners. If the signal counts a free trial as a paid customer, the algorithm finds more free-trial tourists. The bid strategy is honest. It optimizes against the signal you give it. The only way to fail with Smart Bidding is to give it junk and let it scale. Scale runs faster than a manual account, which is why the bill arrives faster too.
Stan · Marketing AtlasSection 09 · Sources
Continue in the Atlas