Article

Smart Bidding

Google Ads' native automated bidding strategies, including Target ROAS, Target CPA, Maximize Conversions, and Maximize Conversion Value.

Smart Bidding refers to Google Ads’ native automated bidding strategies, which use Google’s portfolio-trained machine-learning models to set bids in real time on individual auctions. The strategies include Target ROAS (tROAS), Target CPA (tCPA), Maximize Conversions, Maximize Conversion Value, and Enhanced CPC. Smart Bidding was introduced in 2016 and has progressively replaced manual bidding for the majority of Google Ads accounts.

How Smart Bidding works

Smart Bidding’s models are trained across the entirety of Google’s advertiser base, then applied to individual accounts. At auction time, the model considers contextual signals — user device, location, time of day, browser, audience attributes, and historical conversion data — and adjusts the bid accordingly. The system optimizes against whatever conversion event the advertiser has identified as the goal.

Because the model is portfolio-trained, Smart Bidding sees far more signal than any single advertiser could provide on their own. This is the structural reason Smart Bidding outperforms manual bidding for most accounts, particularly those with limited historical conversion data.

Strategies

Target ROAS (tROAS). Bids are set to achieve a specific revenue-to-spend ratio. Requires conversion-value tracking. The right strategy for ecom accounts.

Target CPA (tCPA). Bids are set to achieve a specific cost per conversion. Right for lead-gen accounts where conversion value isn’t tracked.

Maximize Conversions. Bids are set to produce the most conversions within the budget, without a CPA target. Useful for budget-constrained accounts learning the cost curve.

Maximize Conversion Value. Similar to Maximize Conversions but optimizes for total revenue rather than count.

Enhanced CPC. A semi-automated strategy that adjusts manual bids based on conversion likelihood. Largely deprecated.

When to use Smart Bidding vs. third-party tools

For accounts below approximately $25,000/month in spend, Smart Bidding usually outperforms third-party bidding tools. The portfolio-trained model has access to far more data than the account itself could produce, which matters most when account-level data is thin.

Above $25,000/month, the math starts to shift. Per-account-trained models from tools like Groas.ai or Albert AI begin to outperform the portfolio model because they can specialize on the account’s specific conversion patterns. The transition point varies by vertical and conversion volume.

Common operating mistakes

Three patterns we see repeatedly:

• Over-intervention during the learning phase. Smart Bidding needs 14–30 days to stabilize after any major change. Operators who intervene during this window reset the model and extend the underperformance.
• Incorrect conversion-event setup. Smart Bidding optimizes against whatever the advertiser tells it is a conversion. If conversion events fire on the wrong action, the model systematically wastes budget on that action.
• Target ROAS set too tight. A target set close to break-even ROAS leaves no headroom for noise. The model will respond by reducing volume rather than absorbing the noise. Set targets at the profitable ROAS, not the break-even.

See also

• Automated bidding software — the parent category
• Performance Max — campaign type built on Smart Bidding
• Manual CPC — the legacy alternative

Maintained by Ruchika Rajput. Methodology and disclosures at methodology.