Smart Bidding is Google’s set of automated bid strategies that use machine learning to optimise for conversions or conversion value at auction time. Unlike manual bidding, it evaluates a wide range of contextual signals for every single auction: device, location, time of day, browser, remarketing list membership, ad creative, and dozens more. When conditions are right, it outperforms manual bidding consistently.
But “when conditions are right” is doing a lot of heavy lifting. Smart Bidding needs data to learn, realistic targets to aim for, and clean conversion tracking to optimise against. Get those wrong and you’ll burn through budget while the algorithm flails. This guide covers each strategy honestly, when to avoid automated bidding entirely, and what to do when performance falls apart. If you want hands-on help, Gorilla Marketing’s Google Ads team works with these strategies daily across multiple industries.
How Does Smart Bidding Actually Work?
The defining feature is auction-time bidding. Every time someone triggers an ad auction, Smart Bidding evaluates the context of that specific user, moment, and query, then sets a bid accordingly. It doesn’t apply the same bid to every auction the way manual CPC does. Google has confirmed it factors in device type, physical location, day of week, time of day, remarketing list status, browser, operating system, actual search query (not just the keyword), ad creative, and the interactions between these signals.
This lets Smart Bidding capture patterns that manual bidding can’t replicate at scale. A human might bid down 20% on all mobile traffic. Smart Bidding can recognise that mobile converts worse on weekdays but better on Saturday evenings for a specific audience segment, and adjust per auction accordingly.
The trade-off is control. You tell the algorithm what you want to achieve and trust it to figure out how. That requires a certain amount of faith in the data, and the data needs to be good. If your conversion tracking is unreliable or your targets are unrealistic, the algorithm will optimise confidently in the wrong direction.
The Four Core Strategies
Target CPA (Cost Per Acquisition)
Tells Google to get conversions at or near a specific cost per conversion. The algorithm adjusts bids auction by auction to hit that average over time.
When to use it: You know your target acquisition cost and have consistent volume (15-30+ conversions per month). Strong for lead gen where you know what a lead is worth.
What it needs: Reliable conversion tracking, enough volume, and a realistic CPA target based on actual historical data, not aspirational numbers.
Strengths: Predictable costs. Good at finding efficient traffic pockets you’d miss manually.
Weaknesses: Treats all conversions equally. A £50 sale and a £5,000 sale count the same. Can throttle spend aggressively if targets are too tight, meaning you miss profitable volume at slightly higher CPAs. Performance also deteriorates quickly when conversion volume drops because the algorithm has less data to work with.
Target ROAS (Return on Ad Spend)
Optimises toward a specific return on spend. Set 400% and you’re telling Google that for every £1 spent, you want £4 in conversion value back.
When to use it: E-commerce or any account where conversion values vary significantly. Stops Google treating a £20 sale the same way it treats a £2,000 one.
What it needs: Dynamic conversion values flowing into Google Ads. For lead gen, that means assigning values to different conversion actions or importing revenue from your CRM.
Strengths: Focuses spend where the money is. Naturally allocates budget toward higher-value products or leads.
Weaknesses: Gets very conservative if targets are too ambitious. Requires accurate value data. Wrong values mean the algorithm optimises toward wrong outcomes.
Maximise Conversions
Gets you the most conversions possible within your daily budget. No CPA target. Spends your full budget every day.
When to use it: New campaigns without enough data to set a realistic CPA target. Testing phases where you need conversion data quickly.
What it needs: A budget you’re comfortable spending in full every day. This strategy will use all of it.
Strengths: Gets volume up fast. Useful for gathering data before transitioning to Target CPA.
Weaknesses: No cost guardrail. It’ll spend £100/day whether conversions cost £10 or £50 each. A stepping stone, not usually a long-term strategy.
Maximise Conversion Value
Like Maximise Conversions but optimises for total value rather than total volume. Spends your full budget chasing the highest possible conversion value.
When to use it: E-commerce campaigns where you want to maximise revenue within a fixed budget.
What it needs: Accurate conversion value tracking and a fixed daily budget.
Strengths: Revenue-focused. Prioritises high-value transactions over volume.
Weaknesses: Same lack of efficiency guardrail. You might generate £5,000 in revenue from £3,000 in spend. Better treated as a data-gathering phase before moving to Target ROAS.
How Smart Bidding Differs by Campaign Type
Search is where Smart Bidding has the most signal to work with. Clear intent from search queries, full strategy selection, and the most transparent relationship between targets and results. All four strategies work here. Target CPA and Maximise Conversions are the most common for lead gen. Target ROAS and Maximise Conversion Value suit e-commerce search campaigns where transaction values vary.
Shopping benefits from the product feed giving the algorithm structured data to optimise against. Target ROAS is the most common choice because Shopping campaigns naturally involve products at different price points and you want the algorithm prioritising profitable ones. Feed quality matters more than most advertisers realise. Poor product titles or missing attributes limit what the algorithm can do, because it’s making decisions based on that data.
Performance Max uses Smart Bidding by default, but you’re trusting channel allocation alongside bid optimisation. The algorithm decides whether to spend on Search, Shopping, Display, YouTube, or Discovery. Target ROAS is generally safest for e-commerce PMax because it prevents the algorithm chasing cheap, low-value Display conversions. Maximise Conversions without a target gives it free rein across channels with no efficiency constraint, which is risky.
Display is the hardest environment for Smart Bidding. Lower-intent traffic means the algorithm works harder to find users who actually convert. Target CPA with realistic expectations is usually the smartest approach. Maximise Conversions on Display without a target frequently leads to spend on low-quality placements that generate clicks but few real conversions.
The Learning Phase
Every time you apply a new strategy, change your target, or make a significant structural change, the algorithm enters a learning phase, typically one to two weeks. Performance will fluctuate during this window. CPAs might spike, ROAS might dip. This is normal.
What restarts the learning phase:
Changing your bid target or strategy
Budget changes over 20%
Adding or removing conversion actions
Major structural changes (pausing large numbers of keywords/ad groups)
Pausing the campaign for an extended period
The rule: don’t panic and don’t tinker. Changing your target three days in because the numbers look bad restarts the entire process. Give it the full window, then evaluate.
When NOT to Use Smart Bidding
Low conversion volume. Fewer than 15 conversions per month means the algorithm is working with thin signal. Performance will be erratic. Manual CPC or Enhanced CPC gives you more control until volume builds.
Brand campaigns. Branded terms convert at high rates regardless of bid strategy. Smart Bidding often bids higher than necessary on brand traffic. Manual CPC with a modest bid captures it cheaper.
Tight margins with no room for learning. If a two-week efficiency dip would cause real damage, the learning period introduces risk you may not be able to absorb.
Unreliable conversion tracking. The algorithm optimises toward whatever tracking tells it. Bad data in, bad bidding out. Fix tracking first.
New accounts with no history. Zero conversion data means the algorithm is bidding blind. Build a baseline with manual bidding first.
Setting Up Smart Bidding Properly
Conversion tracking first. Make sure you’re measuring actions that matter, not page views or chatbot opens. Check your counting settings: “Every” for e-commerce (each purchase counts), “One” for lead gen (multiple form fills from the same person aren’t multiple leads).
Realistic targets. Look at your actual CPA or ROAS over the last 30 days. Set your target at or slightly above that, let the campaign stabilise, then tighten gradually. A 10-15% adjustment over weeks is reasonable. A 50% overnight jump is not.
Budget alignment. Your daily budget should be at least 5-10 times your Target CPA. If your Target CPA is £50 and your budget is £30/day, the algorithm can barely afford one conversion per day. That’s not enough to learn from.
Common Mistakes
Changing targets too often. Every adjustment resets learning. Set a target, wait two weeks minimum, then evaluate.
Not enough data. Launching Target CPA on a campaign with eight monthly conversions will produce inconsistent results and lead you to blame Smart Bidding when the real problem is volume.
Misaligned conversion actions. If the campaign optimises for “all conversions” including newsletter sign-ups alongside actual purchases, the algorithm chases the cheapest ones. Set primary conversion actions deliberately.
Wrong strategy for the goal. Maximise Conversions when you care about efficiency. Target ROAS without conversion values set up. Match the strategy to what actually matters.
Rolling out everywhere at once. Switching your entire account to Smart Bidding simultaneously means every campaign enters learning at the same time. Roll it out campaign by campaign.
Ignoring search terms. Smart Bidding adjusts bids, not targeting. Irrelevant search queries still need negative keywords. That’s still your job.
What to Do When Smart Bidding Isn’t Working
Don’t scrap it immediately. Work through these steps first.
Assess the timeline. If it’s been less than two weeks since your last change, you may still be in the learning phase. Check for the “Learning” label on your bid strategy. If it’s still showing, give it more time before making any decisions.
Check data inputs. Is conversion tracking still working? Has anything changed on the site (broken forms, new checkout flow, slower load times)? Has a conversion action been accidentally removed or duplicated? A surprising number of “Smart Bidding failures” are actually tracking failures that have nothing to do with the bidding algorithm itself.
Review targets against reality. If your Target CPA is £15 but your actual CPA over the last 30 days is £35, the target was never achievable with current traffic quality and conversion rates. Pull up historical data and set a target the algorithm can actually reach.
Check the competitive environment. CPC increases, new competitors entering the auction, seasonal demand shifts. All of these affect Smart Bidding performance. If CPCs have risen 30% in your category, your previous target may simply not be achievable without changes to landing pages or conversion rates.
Reset, don’t abandon. If the strategy is right but targets were wrong, adjust to something achievable and give the campaign a fresh two-week learning window. If you need a full reset, switch briefly to Maximise Conversions with a reduced budget. This lets the algorithm re-establish conversion patterns without efficiency pressure. Once you’ve accumulated a couple of weeks of fresh data, move back to a target-based strategy with realistic numbers informed by the new baseline.
Know when to walk away. If you’ve checked everything, tested realistic targets, given adequate learning time, and performance still isn’t there, it might genuinely be the wrong strategy for that campaign. Low-volume campaigns, niche audiences, and irregular conversion patterns are all poor fits. Switch to manual CPC and revisit when conditions change.
Seasonality Adjustments
Smart Bidding learns from historical data, so it reacts slowly to sudden conversion rate changes. Google’s seasonality adjustments let you specify a date range and expected change in conversion rate, telling the algorithm to bid more aggressively during a flash sale or pull back during a known dip.
A few practical points. Seasonality adjustments work best for short, sharp changes of 1-7 days. For longer seasonal trends, the algorithm typically adapts on its own without intervention. Apply them to the specific campaigns that will be affected, not blanket across everything. Be conservative with your estimates: if you say conversion rates will double and they only increase 30%, the algorithm overbids during the adjustment window and performance suffers. For recurring events like Black Friday or end-of-financial-year pushes, review your settings annually. Last year’s conversion rate lift might not repeat exactly.
Audience Signals and Smart Bidding
Smart Bidding already factors audience data into its bid calculations. Someone on your remarketing list gets a different bid than a new visitor. You don’t set these adjustments manually; the algorithm handles it.
But you can improve performance by feeding it better signals. Upload customer match lists so the algorithm knows who’s already bought from you. Segment remarketing audiences by behaviour (“added to cart” is more useful than “visited site”). Build custom segments based on search behaviour or sites visited. Layer in-market audiences on Display and YouTube campaigns where intent signals are weaker.
These aren’t hard targeting restrictions. They’re inputs the algorithm uses alongside every other contextual signal when setting bids. You don’t need to set manual bid adjustments for each audience segment. Smart Bidding evaluates membership alongside device, location, time, and everything else, then sets the bid accordingly. The better your signals, the faster and more accurately it learns.
Getting the Most from Automated Bidding
Smart Bidding works best when you treat it as a tool that needs the right inputs, not a replacement for strategy.
Start with clean conversion tracking, realistic targets, and budgets that give the algorithm room to learn. These aren’t exciting steps, but they’re the ones that determine whether Smart Bidding succeeds or fails. Most underperformance isn’t an algorithm problem. It’s an input problem.
Match the strategy to the goal. Target CPA for predictable acquisition costs. Target ROAS when conversion values vary. Maximise Conversions to gather data quickly in new campaigns. Get the match wrong and you’ll blame the automation for what was really a decision-making error.
Respect the learning phase. Two weeks minimum, no tinkering, no panic adjustments. If you can’t leave a campaign alone for a fortnight, automated bidding will frustrate you constantly, and manual CPC is a perfectly reasonable alternative.
Monitor regularly but intervene rarely. Check performance weekly, review search terms, keep negative keyword lists current, and update audience signals as you learn more about your customers. But don’t change bid targets every time you have a bad day. The algorithm needs stability to perform.
And when performance drops, follow the recovery steps before reaching for the manual override. Smart Bidding failures are almost always diagnosable, and the fix is usually simpler than you think: a target that drifted from reality, a tracking issue that crept in, or a competitive shift that changed the maths. Find the root cause, fix it, and give the algorithm time to recover. That methodical approach will serve you better than cycling between automated and manual bidding whenever the numbers dip.
