
Summary
- Smart Bidding is the standard in 2026. Manual bidding only makes sense with a budget under 25,000 CZK (≈ €1,000) a month or for brand-new campaigns with no data. Everywhere else, automated strategies beat manual settings.
- For e-shops the best strategy is Maximize Conversion Value (optimising for order value), not Maximize Conversions (number of orders). Once you've gathered 50+ conversions a month, add a Target ROAS goal.
- The minimum for it to work: 30 conversions in 30 days and a budget of at least 50,000 CZK (≈ €2,000) a month. Below that threshold the algorithm doesn't have enough data and results will fluctuate.
- The most common mistake is impatience. The learning phase takes 1–3 weeks (about 50 conversion events). Change your settings after three days and you're starting over, again and again.
- In 2026, AI tools can watch Smart Bidding for you — automatic alerts on anomalies, budget forecasting, weekly reporting. Your specialist can then focus on strategy instead of checking numbers.
Smart Bidding — automated, AI-driven bidding strategies — is now used by the vast majority of successful advertisers. And no wonder. According to an Optmyzr study that analysed 14,584 accounts, campaigns using Smart Bidding achieve on average 19–27% better ROAS than campaigns with manual settings.
But "switching on Smart Bidding" isn't enough. The wrong strategy, unrealistic targets or a lack of data can make your campaigns worse instead of better. And that's exactly what we see with plenty of e-shops that come to us — Smart Bidding switched on, but no results in sight.
This article is a cheat sheet you can keep coming back to. It gives you an overview of every strategy, a "when to use what" decision table, the most common mistakes and a practical process for setting up Smart Bidding so it saves you money instead of spending it.
What Smart Bidding is and how it works
Smart Bidding is the umbrella name for automated bidding strategies in Google Ads that use machine learning to optimise bids in real time. At every auction — and there are billions daily — the algorithm evaluates dozens of signals and decides how much to bid for a click.
What signals does it take into account? Plenty that you, as the advertiser, have no access to:
- Device — mobile, tablet, desktop (and their specific models)
- Location — not just the city, but the neighbourhood
- Time of day and day of the week
- Remarketing lists — whether the user has already been on your site
- Browser language and operating system
- Demographics — age, gender, household income
- The user's historical behaviour on Google
That's a fundamental advantage over manual bidding. You can set a different bid for mobile and desktop. The algorithm can set a different bid for "a woman, 32, from Brno, on an iPhone, on Tuesday at 2 p.m., who already saw your ad last week." You simply can't do that by hand.
The 4 Smart Bidding strategies: an overview and when to use which
Google Ads has four main Smart Bidding strategies. In reality there are two core ones (Maximize Conversions and Maximize Conversion Value), and to each you can add a goal (Target CPA or Target ROAS). Since 2025 that's how Google displays it too — Target CPA and Target ROAS are no longer standalone strategies but "goals" added to one of the two core ones.
It sounds complicated, but it's simple. Here's the overview:
1. Maximize Conversions
The algorithm tries to get as many conversions as possible within your budget. It doesn't ask about cost — it just wants the most conversions. That's great for the start, when you need data. But be careful — this strategy spends your entire budget, whether the conversions are expensive or cheap.
- When to use it: Launching a new campaign, gathering data, short-term promotions.
- Minimum: None — it works even without history.
- Risk: The algorithm ignores cost per conversion. It might bring you 50 orders at a CPA of 500 CZK even if your target is 200 CZK.
2. Maximize Conversions + Target CPA (target cost per conversion)
Here you tell the algorithm: "I want as many conversions as possible, but I don't want to pay more than X per one." The algorithm then finds a balance between volume and cost.
- When to use it: Lead generation, B2B, services — anywhere every conversion has roughly the same value.
- Minimum: 30+ conversions in the last 30 days.
- Risk: Set the target too low → the algorithm stops spending, because it can't find conversions at your price. Set it to your current CPA or at most 10–20% lower.
3. Maximize Conversion Value
This is the number-one strategy for e-shops. The algorithm doesn't optimise for the number of orders but for their total value. So it would rather bring one order worth 5,000 CZK than five orders worth 200 CZK. And that's exactly what most e-shops want.
- When to use it: E-commerce with varied product prices.
- Minimum: No formal minimum, but 30+ conversions a month for stability.
- Risk: Without a Target ROAS goal, the algorithm may "chase" high-value orders at a disproportionately high cost per click.
An interesting finding from the Optmyzr study: Maximize Conversion Value delivers the best results of all the strategies — better ROAS, better CPA and better CPC. Even for lead generation it beats Maximize Conversions by almost 300% on ROAS. That's a number worth thinking about.
4. Maximize Conversion Value + Target ROAS (target return on ad spend)
The gold standard for mature e-shops. You tell the algorithm: "I want to maximise order value, but for every unit I spend I want at least X back." A Target ROAS of 500%, for example, means that for every 100 CZK in ads you want 500 CZK in revenue.
- When to use it: E-shops with 50+ conversions a month and accurate revenue tracking.
- Minimum: 50+ conversions in 30 days (Google recommends ideally 100+).
- Risk: Too high a ROAS target → the algorithm limits spending because it can't meet your expectations. Start at your current ROAS and raise it gradually by 10–15%.
Decision table
| Your business | Campaign stage | Recommended strategy |
|---|---|---|
| E-shop, new campaign | Start, data gathering | Maximize Conversion Value |
| E-shop, 30+ conversions/mo | Growth | Max Conv. Value + a mild tROAS |
| E-shop, 50+ conversions/mo | Mature campaign | Max Conv. Value + Target ROAS |
| Lead gen / services | Start | Maximize Conversions |
| Lead gen / services, 30+ conversions | Mature campaign | Max Conversions + Target CPA |
| New campaign with no data | The very beginning | Maximize Conversions (no goal) |
| Budget under 25,000 CZK (≈ €1,000)/mo | Any | Consider manual CPC |
The learning phase: why you have to hold on and leave it alone
When you switch on or change a Smart Bidding strategy, the campaign enters what's called the learning phase. The algorithm gathers data and learns how much to bid and to whom. It takes roughly 50 conversion events, which — depending on your campaign volume — means:
- 10+ conversions a day: Learning takes less than a week.
- 2–3 conversions a day: Learning takes 2–3 weeks.
- Fewer than 1 conversion a day: Learning can take a month or more — and results will be unstable.
During the learning phase your CPA and ROAS will fluctuate. Sometimes sharply. That's normal and expected. The algorithm is testing different audience segments and different bid levels to find out what works.
What NOT to do during learning:
- Don't change the bidding strategy. Switching = a restart from zero.
- Don't change the target values (tCPA/tROAS). Every significant change triggers a new learning phase.
- Don't change the budget by more than 20%. The algorithm handles small adjustments; big jumps it doesn't.
- Don't panic after 3 days. I know the itch is real. But three days of data isn't enough to conclude anything.
One thing Google has confirmed that surprises many: you don't need to "warm up" the campaign on a simpler strategy and then switch. You can start straight away with the strategy you want to run. The system learns how conversions come in. So if you're an e-shop, feel free to start right away with Maximize Conversion Value.
The 5 most common mistakes with Smart Bidding
1. Unrealistic targets
This is the most widespread mistake. Your current CPA is 350 CZK and you set a target of 150 CZK because "that would be nice." The algorithm reads it literally — and stops spending, because it simply can't get conversions for 150 CZK. The result: the campaign stalls, no conversions, and you're left feeling Smart Bidding doesn't work.
Rule: set targets no more than 10–20% away from your current values. Once it stabilises, move again by 10–15%. Never jump by more than 20% at once.
2. Optimising for the wrong conversion actions
This one's insidious because you don't see it at first glance. If you have as your primary conversion in Google Ads a visit to the "Thank you for your order" page AND ALSO a newsletter sign-up AND ALSO a phone click — the algorithm treats all three equally. But a newsletter sign-up worth 15 CZK won't keep your e-shop alive.
The fix: keep only purchase as your primary conversion (or purchase + begin checkout). Set everything else as secondary conversions — Google will report them but won't optimise bidding for them.
3. Impatience
You changed the strategy on Monday, on Wednesday your CPA jumped 40%, so on Friday you switched it back. Congratulations — you've just wasted a whole week of data and learned absolutely nothing. Smart Bidding needs time. At least two weeks, ideally three to four, before you evaluate.
One tip: if the nerves get the better of you, set up an automatic alert (say, via a Google Ads script) that emails you when CPA exceeds twice your target. That gives you a safety net without having to check the account every two hours.
4. Too little data
Smart Bidding under 30 conversions a month is a lottery. Under 15 conversions a month it's a permanent "semi-learning" state — the algorithm never properly learns. If your volume is low, either merge several campaigns into one (so the algorithm has enough data) or stick with manual CPC.
A tip from practice: if your e-shop has a low conversion frequency, try setting "Add to cart" as the conversion action instead of "Purchase." You'll have more conversion data for the algorithm and the campaign will learn faster.
5. Poor conversion tracking
This is the be-all and end-all. Smart Bidding is only as good as the data it learns from. If your conversion tracking doesn't count correctly, the algorithm optimises for the wrong signals.
Specifically: use the Google Ads conversion tag, not GA4 imported conversions. GA4 events are delayed; the Google Ads tag provides real-time signals. And set up Enhanced Conversions — campaigns that use them show 40–60% better results.
Smart Bidding for e-shops: a practical process
If you run an e-shop and want to set up Smart Bidding correctly, here's the step-by-step process:
Step 1: Check your conversion tracking
Do you have the Google Ads conversion tag? Does it send correct order values (dynamic, not fixed)? Do you have Enhanced Conversions? If not, this is the first thing to fix — there's no point going further without it.
Step 2: Choose Maximize Conversion Value
Not Maximize Conversions. For an e-shop you want to optimise for order value, not order count. Otherwise the algorithm will bring you orders worth 99 CZK instead of 2,500 CZK.
Step 3: Let it run for 3–4 weeks
No changes. Watch it, but leave it alone. Keep the budget as it is.
Step 4: After 50+ conversions, add Target ROAS
Look at the ROAS the campaign is currently achieving. Set Target ROAS to that value — not to a dream target. For example, if the campaign is achieving a ROAS of 380%, set tROAS to 380%.
Step 5: Optimise gradually
Every 2–4 weeks, raise the ROAS target by 10–15%. Watch whether conversion volume drops. If it does, the target is too aggressive — take a step back.
For orientation: an e-shop with a 40% margin needs at least 300–400% ROAS to be profitable. Top campaigns reach 600–1,200%.
AI tools and Smart Bidding: what the machine can do for you
Smart Bidding is itself AI. But who's watching that the AI does what it should? In 2026 the answer is: more AI.
It sounds like sci-fi, but it's practical. Smart Bidding optimises bids in real time — that's its job. But someone has to watch whether the campaign is behaving oddly, whether CPA has spiked, whether the budget will last to the end of the month, whether a new learning phase went through fine.
Traditionally a PPC specialist does this by hand — opens the account, goes through the numbers, compares with last week. It takes hours. AI tools do it in minutes and only flag you when something's wrong.
What exactly the AI watches:
- Anomalies in CPA/ROAS — if performance deviates more than 25% from the average, you get an alert.
- Budget pacing — a forecast of whether your budget will last to the end of the month or run out in 20 days.
- Learning phase monitoring — tracking whether learning is progressing normally or the campaign has stalled.
- Weekly reporting — an automatic overview comparing against the previous week, identifying problems and suggesting actions.
Paid tools like Optmyzr or Skai cost hundreds of dollars a month. But custom solutions built on AI tools like Claude Code do the same for a fraction of the cost — and exactly to your needs.
At LK Media we use these automations every day. Instead of spending the morning checking dozens of campaigns, we get a report of what needs attention. And we spend the rest of the time on what AI can't handle — strategy, creative and business decisions.
Conclusion
Smart Bidding isn't a magic wand that fixes campaigns on its own. It's a tool that needs the right data, realistic targets and patience. Give it those and your reward is cheaper conversions, better ROAS and more time for the things that need a human mind.
Start simply: check your conversion tracking, choose Maximize Conversion Value, and let it run for three weeks. Then evaluate. You'll see for yourself whether the numbers make sense.
And if you're not sure which strategy is right for your e-shop, get in touch at info@lkmedia.cz. We'll look at your account and tell you straight away what to change.
FAQ
What is Smart Bidding in Google Ads?
Smart Bidding is the umbrella name for automated bidding strategies that use machine learning to optimise your click bids in real time. At every auction the algorithm evaluates dozens of signals (device, location, time, user history) and decides how much to bid for a click. The main strategies are Maximize Conversions, Target CPA, Maximize Conversion Value and Target ROAS.
What's the difference between Target CPA and Target ROAS?
Target CPA (target cost per conversion) tells the algorithm the maximum you're willing to pay for one conversion — it suits services and lead generation, where every lead has roughly the same value. Target ROAS (target return on ad spend) tells the algorithm what return on investment you want — it suits e-shops, where orders have different values. An e-shop should almost always use Target ROAS.
How many conversions do I need for Smart Bidding to work?
The minimum is 30 conversions in the last 30 days for Target CPA strategies. For Target ROAS, Google recommends 50 conversions a month, ideally 100+. Below 15 conversions a month the campaign slips into a permanent "semi-learning" state and results will be unstable. If you have fewer, consider merging campaigns or using a simpler conversion action (add-to-cart instead of purchase).
How long does the Smart Bidding learning phase take?
The learning phase takes roughly 50 conversion events. At 10 conversions a day that's less than a week; at 2–3 conversions a day it's 2–3 weeks. Performance fluctuates during learning — that's normal. Don't change the strategy, targets or budget by more than 20% during this time. Any larger change restarts learning from the beginning.
Is Smart Bidding worth it for small e-shops on a low budget?
It depends on volume. If your budget is over 50,000 CZK (≈ €2,000) a month and you have at least 30 conversions a month, Smart Bidding will most likely help you. Under 25,000 CZK (≈ €1,000) a month and with only a handful of conversions, you're better off staying with manual CPC or the Maximize Clicks strategy and gathering data gradually. Smart Bidding needs data the way a car needs fuel — without it, it won't get anywhere.


