Smart Bidding has been around long enough that most advertisers have formed opinions — often based on bad early experiences. The product has changed significantly, and if you’re still manually setting bids or relying on outdated rules about when automated strategies “don’t work,” you’re probably leaving money on the table.
How Smart Bidding Actually Works
Smart Bidding uses Google’s machine learning to set bids at auction time, factoring in signals that aren’t available to manual bidders: device, location, time of day, browser, search query intent, remarketing list membership, and dozens more. The key insight is that these signals are evaluated in combination — not one at a time — making the bid landscape far more nuanced than any manual strategy can replicate at scale.
tCPA vs tROAS: Choosing the Right Strategy
Target CPA is best when you have a single conversion type, relatively consistent conversion values, and at least 30–50 conversions per month per campaign. It’s ideal for lead generation, app installs, and subscription sign-ups.
Target ROAS is best for e-commerce and any scenario where conversion values vary significantly between transactions. It requires more conversion data (50–100 conversions per month is a reliable minimum) but optimises for revenue rather than just volume.
A common mistake is setting targets too aggressively at launch. Start with a target 10–20% looser than your actual goal and tighten over time as the algorithm learns.
Feeding the Algorithm Better Signals
Smart Bidding is only as good as the conversion data you feed it. Prioritise these improvements:
- Import offline conversions. If your sales happen off-platform (phone calls, CRM deals, in-store), import those back into Google Ads. The algorithm optimises for what you measure.
- Use enhanced conversions. This improves conversion matching when cookies are blocked or delayed, giving the algorithm better data to work with.
- Set conversion values accurately. If all your conversions are tagged as the same value, tROAS has nothing meaningful to optimise toward.
Common Smart Bidding Mistakes
Making too many changes too quickly. Every time you significantly change a target or budget, the algorithm re-enters a learning phase. Give strategies 2–4 weeks before evaluating performance, and avoid major changes during this period.
Evaluating on too short a window. Look at 30-day rolling performance rather than week-over-week, especially for lower-volume campaigns where statistical variance is high.
Ignoring the budget constraint. Smart Bidding can’t perform if your budget runs out by midday. Ensure your budget is large enough that it’s not the binding constraint — the strategy should be.
When to Override with Manual Bidding
Manual bidding still makes sense for brand new campaigns with zero conversion history, for extremely low-volume campaigns that can’t generate enough data, and for highly seasonal promotions where historical data is misleading. In all other cases, Smart Bidding — with proper setup — should outperform manual over a sufficient time window.
