Feeding Better Signals to Google Ads Smart Bidding

Smart Bidding's Dependency on Conversion Signals

Google Ads Smart Bidding (Target CPA, Target ROAS, Maximise Conversions, Maximise Conversion Value) is a machine learning system that predicts the probability of conversion for each auction and bids accordingly. The quality of its predictions depends entirely on the quality and completeness of the conversion data you feed it. Poor conversion signals → poor bidding decisions → wasted budget.

Signal Quality Factor 1: Conversion Volume

Smart Bidding needs sufficient conversion volume to learn effectively. Google recommends:

• Target CPA: at least 30-50 conversions per month per campaign
• Target ROAS: at least 50 conversions per month per campaign

Below these thresholds, the algorithm has insufficient data to optimise reliably. If you are below these volumes, consider:

• Optimising for a higher-funnel conversion event (add to cart instead of purchase) to increase volume, then switching to purchase when volume grows
• Consolidating campaigns to pool conversion signals

Signal Quality Factor 2: Completeness (Recovering Missing Conversions)

If 30% of conversions are invisible to Google Ads (due to ad blockers, Safari ITP, cross-device journeys), Smart Bidding optimises on 70% of the actual signal. The algorithm thinks campaigns are performing worse than they are.

Fixes:

• Enhanced Conversions: send hashed email/phone to Google, enabling conversion matching beyond browser cookies
• Server-side conversion import: send conversions directly from your backend to the Google Ads API, bypassing browser-side tracking entirely
• Store gclid server-side: preserve click attribution beyond the 7-day Safari cookie limit

Signal Quality Factor 3: Conversion Values (for ROAS bidding)

For Target ROAS and Maximise Conversion Value campaigns, the revenue values you send with each conversion are critical. If all purchases send the same fixed value (£50) instead of the actual order value, value optimisation cannot function — there is nothing to optimise.

Requirement: every purchase event must include the actual order value in the conversion value field, not a fixed placeholder.

Signal Quality Factor 4: Conversion Attribution Window

Smart Bidding uses conversions attributed within your selected window for its training data. If your window is too short (7 days) but many purchases happen after 14 days, Smart Bidding is trained on an incomplete subset of your conversions.

Check your Time Lag report (Google Ads → Attribution → Time Lag) and set the attribution window to cover 95%+ of your conversion distribution.

Signal Quality Factor 5: Avoid Micro-Conversions as Primary

If you use micro-conversions (page views, add-to-carts, newsletter signups) as primary conversion actions for Smart Bidding on high-intent campaigns, the algorithm optimises for the wrong thing. Use purchase/lead as primary; add lower-funnel events as secondary (observed but not bid-optimised).

Summary

Smart Bidding quality comes from five signal factors: sufficient volume (30-50 conversions/month minimum), completeness (recover missing conversions via Enhanced Conversions), accurate values (real order values, not fixed placeholders), correct attribution windows (covering 95%+ of conversions), and appropriate primary conversion actions (final business outcomes, not micro-conversions). Improving all five simultaneously produces the strongest compounding improvement in Smart Bidding performance.

See our Conversion Tracking Fix service for bidding signal optimisation.

Need Smart Bidding signals improved? Contact Adslytics.