Google Ads Smart Bidding: How to Set It Up to Actually Work

Google's smart bidding strategies — Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value — use machine learning to set bids at auction time based on a wide range of signals: device, location, time, audience, search query, and more. When they work, they outperform manual bidding by a significant margin. When they don't, they can burn budget fast.

The most common reason smart bidding fails is insufficient conversion data. Google's algorithms need at least 30–50 conversions per month per campaign to learn effectively. Campaigns with fewer conversions should use Maximize Conversions without a target, or stay on manual CPC with enhanced CPC, until the data threshold is met.

Conversion quality matters as much as conversion volume. If your conversion actions include low-intent signals — page views, time on site, scroll depth — the algorithm will optimize for those instead of the outcomes you actually care about. Audit your conversion actions before switching to smart bidding. Only include actions that represent real business value.

The ramp-up period is critical and often mismanaged. When you switch to a smart bidding strategy, Google enters a learning phase — typically 1–2 weeks — during which performance may be volatile. Resist the urge to make changes during this period. Changing targets, pausing keywords, or adjusting budgets resets the learning phase and extends instability.

Target setting is where most accounts go wrong. Setting a Target CPA that's 40% below your historical CPA will cause the algorithm to restrict impressions aggressively, starving the campaign of the data it needs to learn. Start with a target close to your historical average, let the algorithm stabilize, then tighten gradually — no more than 10–15% at a time.

Portfolio bid strategies — applying a single smart bidding strategy across multiple campaigns — can improve performance when individual campaigns lack sufficient conversion volume. By pooling data across campaigns, the algorithm has more signal to work with. This is particularly useful for accounts with many small campaigns targeting related products or services.