Smart Bidding on Google Ads: Automated Bid Strategies Explained

Every time someone searches for a product or service on Google, an auction opens. In that auction, Google decides who to show, in what order, and at what cost. Until a few years ago, you set the bid yourself, by hand, keyword by keyword: €1.20 on this one, €0.80 on that one, tweaking on the fly whenever something went wrong. Today that job is done by smart bidding, a set of automated bidding strategies that calculate, for every single auction, how much to offer by reading dozens of signals no human could evaluate in real time.

The debate is no longer "automated versus manual". That argument is settled, and on the vast majority of accounts automation has won. The real question is different: which smart bidding strategy to choose, and above all how to make the algorithm optimize toward the right customers rather than the next click that happens to pass by. Because a poorly fed smart bidding strategy brings you a steady stream of leads that never close, with the nagging feeling that "Google Ads just spends money and nothing else".

In this guide we cover the three strategies that actually matter (Target CPA, Target ROAS, Maximize Conversions), when to use one or the other depending on whether you do lead generation or ecommerce, and the part almost nobody manages well: how to feed the algorithm clean conversions from your CRM, not inflated events.

What smart bidding is (and isn't)

Smart bidding is Google's name for machine-learning-based bidding strategies that optimize for conversions or conversion value at auction time. The key difference from the old automated rules is exactly this: the decision is made at the precise instant the user searches, evaluating that specific person's context.

Which signals does it use? Just to get a sense of it: device, operating system, time of day and day of week, language and location, on-site behavior, browser, which remarketing list the user belongs to, query characteristics, plus hundreds of combinations of these elements. These are signals that, with old manual rules, you could only touch in a coarse way: a +20% on mobile, a -30% at night. The algorithm instead cross-references them granularly for every single auction.

What smart bidding is not: it's not magic, and it doesn't optimize toward your business goals. It optimizes toward whatever conversion signal you give it. If you tell it a conversion is "form submitted", it will maximize form submissions, junk ones included. If you tell it a conversion is "qualified paying customer", it aims for those. Output quality depends brutally on input signal quality. We'll come back to this, because it's 70% of the game.

Why manual bidding has almost always lost

A human manager looks at aggregated data and reacts a few days late. The algorithm evaluates every single impression in real time. Even at modest volumes, this granularity advantage almost always beats the expert's hand, especially now that the newest campaign types (Performance Max, Demand Gen) only run on automated bidding. Manual bidding survives in specific niches: tiny budgets with very few conversions a month, or early phases where you want to buy data at a controlled cost. For everything else, the question isn't "whether" to use smart bidding, but "how".

The three strategies that matter

Google offers several strategies, but in daily practice you use three. The others (Maximize Clicks, Target Impression Share) serve goals other than conversions, so we leave them out of the discussion.

Maximize Conversions

Here you tell Google: "spend the whole budget and bring me as many conversions as possible". You don't set a target cost; the algorithm looks for the maximum number of conversions within the daily budget. It's the strategy you almost always start with, because you're not imposing a cost constraint before you even know what a conversion actually costs you.

The risk: if the budget is generous and competition is expensive, cost per conversion can spike. Maximize Conversions will do everything to spend the entire budget, even by paying for expensive conversions. Keep an eye on it, and use it as a launch pad to gather data, not as a long-term regime if you have tight cost constraints.

Target CPA (target cost per action)

Here you add the missing constraint: "bring me conversions, but keep the average cost per conversion around X euros". If you set a Target CPA of €40, the algorithm tries to generate as many conversions as possible while keeping average spend close to those €40. It's the go-to strategy for lead generation on Google Ads, where every conversion (a lead) has roughly the same value to you and what matters is the cost of acquiring it.

A technical note: in recent versions of Google Ads, Target CPA is no longer a standalone strategy but a setting within "Maximize Conversions" to which you add an optional target CPA. The logic doesn't change: with a target you set a soft ceiling on cost; without one you maximize volume.

Target ROAS (target return on ad spend)

Here the goal changes: no longer the number of conversions or their cost, but the value generated relative to spend. ROAS stands for Return On Ad Spend. A Target ROAS of 400% means "for every euro spent I want €4 of tracked revenue". It's the strategy for ecommerce and Shopping campaigns, where every conversion is worth something different: selling a €300 product isn't the same as selling a €20 one, and you want the algorithm to know that and aim for value, not headcount.

Here too, in recent versions, Target ROAS is a setting within "Maximize Conversion Value". The concept is identical to the previous one, just shifted from cost to value.

Quick reference

Strategy	Optimizes for	When to use it	Requires
Maximize Conversions	Number of conversions	Campaign launch, data gathering	Conversion tracking only
Target CPA	Conversions at a target cost	Lead generation, services	Known sustainable cost per lead
Maximize Conversion Value	Total conversion value	Ecommerce with no rigid target	Tracked conversion values
Target ROAS	Value/spend at a target ROAS	Ecommerce with a margin goal	Values plus conversion history

Lead generation or ecommerce: which to choose

The starting rule is simple, and it's worth repeating because it's the choice most often gotten wrong by people setting up campaigns on their own.

Do you run lead generation (agencies, professional firms, services, B2B, quotes)? Every conversion is a contact, and at the entry stage contacts have roughly the same value. What matters to you is the cost of acquiring them. The right pairing is: Maximize Conversions to launch and gather data, then Target CPA once you know what you can afford to pay for a lead. The number you need before setting the target comes from reasoning through your sustainable cost per lead: if a lead costs you €40 but the average value of an acquired customer is high and your close rate is decent, that €40 is a bargain; if you close one contact out of fifty, it's a bleeding wound.

Do you run ecommerce? Every conversion is a sale with a different value. Using Target CPA here would be a mistake, because you'd treat a €30 order the same as a €300 order. The right pairing is Maximize Conversion Value to start, then Target ROAS once you know the margin you need to protect. Be careful with ROAS as a metric, though: on its own it can mislead, because it ignores real margin and cost of goods. Before setting a target, it's worth understanding the difference between MER and ROAS and tying the number to customer lifetime value, not just the first order.

The mistake of setting too aggressive a target

The fastest way to break smart bidding is to impose a Target CPA that's too low, or a Target ROAS that's too high, relative to actual data. If conversions historically cost you €50 and you set a target of €20, you're not "asking to be more efficient": you're telling the algorithm not to enter auctions that cost more than that level. The typical result is that volume collapses, impressions disappear, and the campaign strangles itself. Set the target close to the real cost you're already achieving, then tighten it gradually (10-15% at a time), watching what happens to volume. Sharp moves equal an unstable campaign.

The part almost nobody manages: feeding the algorithm

This is where the difference lies between an account that works and one that burns budget. Smart bidding is hungry for data and optimizes toward whatever you tell it is a conversion. If the conversion signal is dirty, the algorithm learns to bring you junk extremely efficiently.

1. Track the right conversion, not the click on the button

The most common mistake in lead generation is counting every form submission as a conversion. Google learns to maximize form submissions. Too bad half of them are curious browsers, competitors, people who filled it in by mistake. You pay for volume of forms, not for customers. The foundation for avoiding this is a properly set up conversion tracking setup that distinguishes between "action taken" and "valuable action".

2. Feed offline conversions back from the CRM

Here's the step that separates the professionals. Not all leads are equal, and only you know, inside the CRM, days after the click, who actually closes. The technique is called offline conversion import: when a lead becomes a qualified opportunity or a paying customer, you send that information back to Google Ads (via file upload, API, or CRM integration), linking it to the original click through the GCLID. This teaches the algorithm to target not leads in general, but leads who become customers. It's the move we cover in detail in the guide on offline conversions and CRM on Google Ads, and it's probably the single highest-ROI move on a lead generation account.

The mechanism, in short:

• The visitor clicks the ad and Google assigns an identifier (the GCLID), which ends up in the form and then in the CRM.
• The lead enters the CRM with that GCLID saved.
• The salesperson works it. The lead becomes "qualified", then "customer" (or "junk").
• You send the status, and where relevant the economic value, back to Google, linked to the GCLID.
• Smart bidding recalibrates: it looks for other users similar to the ones who closed and avoids profiles similar to those who only generated noise.

Without this feedback loop, you're optimizing blind. With it, every euro spent gets smarter than the last. Naturally, this requires the flow from form to CRM to be clean and reliable: if the GCLID gets lost along the way or the data arrives incomplete, the signal never makes it through. This is where a well-integrated CRM stops being just a contact list and becomes the source that trains your advertising.