Smart Bidding
Google's ML-powered automated bidding strategies
Google's ML-powered automated bidding strategies
Why It Is Critical
Smart Bidding is Google's machine learning-powered automated bidding system. It adjusts your bid in real time for every individual auction based on dozens of contextual signals - device, location, time, search query, audience, and more - that no human can process at auction speed. For most advertisers, Smart Bidding outperforms manual bidding once there is sufficient conversion data to learn from.
How It Works & Underlying Dynamics
The Smart Bidding Strategies: • Target CPA: Optimise for conversions at a specific acquisition cost. Needs 30-50 conversions/month minimum. • Target ROAS: Optimise for conversion value at a target return. Ideal for e-commerce with varying cart values (50+ conversions/month). • Maximise Conversions: Get as many conversions as possible within budget. Used when launching without enough data for Target CPA. • Maximise Conversion Value: Get highest total conversion value within budget. • Enhanced CPC (eCPC): Manual bidding with automated adjustments. • Target Impression Share: Appear in a target % of eligible auctions (brand defense).
Calculation Example & Benchmark Matrix
| Strategy | When To Use |
|---|---|
| Target CPA | Optimise for conversions at a specific cost. Needs 30-50 conv/mo. |
| Target ROAS | Optimise for conversion value at target return. For varying basket sizes (50+ conv/mo). |
| Maximise Conversions | Get maximum conversions within budget. Best for launching new campaigns. |
| Maximise Conversion Value | Get highest revenue within budget without fixed ROAS target. |
| Enhanced CPC (eCPC) | Manual bidding with automated adjustments for transitional control. |
| Target Impression Share | Appear in target % of auctions. For brand protection & competitor defense. |
'Smart Bidding is not a replacement for strategy - it is a tool that executes your strategy at a speed and scale no human can match. Feed it the right conversion signals, give it enough data, and let it learn. The mistake is either not trusting it at all, or trusting it without verifying the conversion data it is optimising toward.'