Google Ads Bidding Strategies: The Full Map And How I Choose

Almost every bidding question I'm asked arrives phrased as a comparison. Target CPA or Target ROAS. Manual or automated. Maximise conversions or maximise conversion value. It's the wrong first question, and answering it directly is how people end up with a strategy that was correct in someone else's account and wrong in theirs.

The right first question is what data the account actually has. Google ads bidding strategies are not interchangeable tools you select by preference or by what performed well for a case study you read. They are consumers of conversion data. Each one has a minimum diet it needs before it can do the thing it claims to do, and feeding a strategy less than it needs doesn't produce a slightly worse version of the intended outcome it produces something closer to random.

This is the hub page for everything I write about bidding. Below is the complete map of what exists, grouped by what each option actually buys rather than by where it sits in the interface. Then the three-question framework I run before choosing anything, the stage-by-stage progression I use as accounts mature, and the part nobody writes down: what the strategy already running in an inherited account tells me about the person who set it up and what they were thinking at the time.

Everything here comes from accounts I've managed or inherited. Where I give a number thirty conversions, fifteen percent, two weeks it's a working threshold I've arrived at by getting it wrong first, not a figure from documentation.

The full map of google ads bidding strategies

Don't group these by name or by the order they appear in the campaign setup flow. Group them by what they buy, because that determines whether a given option can function in your account at all.

Conversion-optimising: the Smart Bidding group

Target CPA, Target ROAS, Maximize Conversions and Maximize Conversion Value. These set bids at auction time using signals you have no manual access to, and all of them require working conversion tracking before they can do anything useful. Google's guidance on determining a bid strategy organises the whole choice around five goal types, and lists Enhanced CPC alongside those four as the conversion-focused options. It also notes that these strategies factor in auction-time signals such as device, location, time of day, language and operating system which is the actual argument for automation over manual control, and a stronger one than the time-saving pitch usually made for it.

The practical division inside this group is simpler than the four names suggest: targets constrain, maximisers spend. Maximize Conversions will spend the entire budget pursuing volume. Target CPA will decline auctions that would break the target, which means it can and will underspend. Neither behaviour is a fault. They're different instructions, and most disappointment with automated bidding strategies traces back to someone issuing one instruction while expecting the other outcome.

Target ROAS and Maximize Conversion Value are the value-based counterparts. They optimise toward return on ad spend rather than volume, which is only meaningful if differentiated conversion values actually reach the account. If every conversion passes the same number, value-based bidding is cost-based bidding wearing a different label, and I've audited plenty of accounts where exactly that was happening for months without anyone noticing.

Click-optimising

Maximize Clicks, Target CPC and Manual CPC. Maximize clicks spends your budget pursuing click volume within whatever ceiling you set. Target CPC steers toward an average click price. Manual CPC hands you keyword-level control over what you're willing to pay.

None of these know what a conversion is. That's not a criticism it's the defining characteristic, and it's why I treat them as scaffolding rather than destinations. A campaign on maximize clicks is a campaign gathering search terms data while something else gets fixed. If it's still there six months later, nobody is managing the account.

Visibility-optimising

Target Impression Share, CPM, tCPM and vCPM. These buy placement or exposure rather than outcomes. Target Impression Share can be set to pursue the absolute top of the page, the top of the page, or anywhere in the search results, and it will buy what you asked for at whatever the auction charges for it.

On brand-defence campaigns this is a legitimate and deliberate choice. On non-brand Search it's the most reliably expensive misconfiguration I encounter, usually set optimistically at launch by someone who read "impression share" as a performance metric rather than a spending instruction, and then never revisited.

View-optimising

CPV bidding for video, where you pay for views and interactions rather than clicks. Different objective, different measurement model, different cluster of decisions entirely and the bidding logic specific to video campaigns does not transfer to Search, no matter how similar the settings screens look.

Portfolio versions

Most target-based options also exist as portfolio strategies applied across multiple campaigns at once. This is not a different algorithm. It's a different data pool, and pooling is frequently the correct fix when several campaigns are each individually too thin to learn anything. Applied across campaigns that share intent and economics, portfolio bidding genuinely works. Applied across a brand campaign and a cold prospecting campaign, it averages two unrelated truths into one number that describes neither.

Bid adjustments sit on top, not inside

Worth stating plainly because it confuses people: bid adjustments are modifiers layered over a bidding method, not a strategy in their own right. Under manual bidding they do exactly what they say. Under most Smart Bidding strategies the system is already making adjustments based on far more granular auction-time context than your device or schedule modifiers can express, and stacking manual multipliers on top mostly adds noise.

The decision framework: three questions, in this order

I ask these in sequence and I don't skip ahead, because question one disqualifies more options than the other two combined and answering questions two and three first wastes everyone's time.

Question 1 Is the conversion data trustworthy?

Not "does tracking exist". Trustworthy: firing once per event, counting the action that matters to the business, not double-counting a thank-you page that visitors reload, not attributing a newsletter signup and a €3,000 purchase to the same conversion action with the same weight.

If the answer is no, every conversion-based option is off the table until it's fixed, because you'd be asking an algorithm to optimise toward a lie and it will comply enthusiastically. This is why I treat conversion tracking as a prerequisite to any bidding decision rather than a parallel workstream to be sorted out later. I've watched an account run Target CPA for four months against a conversion action that was firing on every page load. The reported CPA was beautiful. The business had no idea why the phone wasn't ringing.

The check takes twenty minutes. Fire a test conversion, confirm it appears once, confirm the value that arrives matches the value that should have arrived, and look at conversion rates by conversion action for anything implausible. A landing page converting at forty percent is not a triumph, it's a tracking bug.

Question 2 How much conversion volume does the campaign have?

Roughly thirty conversions in thirty days is where I start trusting a target-based strategy to actually hold a target. Between fifteen and thirty, I use Maximize Conversions without a target and let it accumulate data. Below fifteen, I don't ask a system to learn from noise.

Under that threshold I have two options and I pick between them based on the business, not the account. Either I widen the conversion definition to include a meaningful micro-conversion one that genuinely correlates with revenue, not a scroll-depth event chosen because it fires often or I run maximize clicks deliberately while tracking accumulates, with a firm CPC ceiling so a cold campaign doesn't discover the expensive end of the auction. The specifics of how much data smart bidding actually needs are worth reading before arguing with those numbers, because the threshold moves with how variable your conversion path is.

The number of conversions matters more than the spend behind them. A campaign spending heavily on ten conversions a month is not a well-fed campaign. It's an expensive one.

Question 3 Are all conversions worth the same?

This is the fork between cost-based and value-based bidding, and it has exactly one correct answer per business.

If every lead is worth roughly the same to you a fixed-price service, a single product, a booking that's a booking optimise for cost. Target CPA. Clean, interpretable, and it fails loudly when something breaks.

If a €2,000 order and a €40 order both arrive in the account as "1", you are instructing the system to treat them identically, and it will do so faithfully. That's the case for optimising toward return on ad spend instead. But value-based bidding only works when values are actually passed and actually differentiated, and the way conversion value gets predicted at auction time is worth understanding before you commit to it, because the system is forecasting value from patterns rather than reading your margins.

The answer that falls out

Run those three questions and the shortlist of viable google ads bidding strategies usually collapses to one. Tracking clean, volume thin, values flat: Maximize Conversions. Tracking clean, volume solid, values flat: Target CPA. Tracking clean, volume solid, values genuinely differentiated: Target ROAS. Tracking broken, in any combination: fix it first, run maximize clicks in the meantime, come back in a month.

That decision tree covers roughly eighty-five percent of the accounts I open. The remaining fifteen percent are edge cases with seasonality, extreme conversion lag, or legal constraints on what can be tracked, and those deserve individual reasoning rather than a framework.

Choosing a google ads bidding strategy by account stage

The same setting behaves completely differently depending on how much history sits behind it. This is the axis most advice ignores, and it's why "Target ROAS works brilliantly" and "Target ROAS destroyed my account" are both true statements from people describing different stages.

Stage 1 New build, no history

Maximize Clicks with a maximum CPC ceiling, or Maximize Conversions if tracking was verified live before launch. The ceiling is not optional: without one, a cold campaign will find the expensive queries very efficiently and spend a fortnight's budget proving it.

I hold this position for two to four weeks with one purpose generating search terms data. I'm not judging performance yet. I'm building the negative keyword layer and finding out which queries this offer actually attracts, because no bidding decision made before that information exists is worth defending.

Stage 2 Learning, first conversions arriving

Maximize Conversions, no target attached. The temptation at this stage is to add a target immediately, because a target feels like control and running without one feels reckless.

It isn't control. A target on thin data restricts which auctions the system is willing to enter, which starves it of the exact data it needs to improve. You get fewer conversions, learned from more slowly, and a campaign that looks like it's failing when it's actually just been handcuffed. Let it spend and learn first. The target comes later and it comes from observed reality.

Stage 3 Scaling, volume stable

Now attach the target set at the CPA or ROAS the campaign is already achieving, not the one you'd like it to achieve. This distinction is the single most common error I correct in inherited accounts.

Then move the target in ten to fifteen percent steps, one step per learning cycle, never two variables at once. This is also where portfolio strategies begin to earn their place, and where what actually happens during the learning phase stops being an abstraction and starts costing real money if you react inside it.

Stage 4 Mature, efficiency-constrained

Value-based bidding if the values are real, targets segmented by campaign tier rather than applied uniformly, seasonality adjustments around events you can anticipate.

At this stage the bid strategy is rarely the bottleneck. Feed quality, creative fatigue, landing page experience and offer competitiveness are. I've stopped expecting a strategy switch to rescue a mature account, and I've stopped letting clients expect it either that conversation is more useful than another fortnight of target tuning.

What I infer from the strategy already running

When I open an inherited account, the bidding configuration is the fastest read available on how the account has been managed. Each of these is a pattern I've seen repeatedly, not a rule.

Manual CPC on a live account in 2026: nobody is managing this. Occasionally justified. Almost always a sign the account was configured once and abandoned. The first thing I check is the date of the last change in the history log, and it usually confirms the diagnosis.

Target CPA set nowhere near actual CPA: someone entered an aspiration, not a target. The campaign is typically limited by something other than budget and delivering a fraction of its available volume, while everyone involved believes the market is simply expensive.

Target ROAS with uniform conversion values: value-based bidding in name only. Every conversion passes the same number, so the system is optimising cost while the label claims otherwise. Nobody is lying. Somebody just never checked what was arriving.

Maximize Conversions on heavy spend with no ceiling anywhere: uncapped by design or by neglect. If cost per acquisition is acceptable, this is fine and I leave it alone. If nobody has looked in three months, it's a budget being converted into volume at whatever price the auction happens to ask.

Target Impression Share on non-brand Search: visibility was confused with performance. Reliably the most expensive single misconfiguration in the account, and reliably the easiest to fix.

Every campaign on a different strategy: no framework, just accumulated decisions. Each choice was probably defensible when it was made. Together they make the account impossible to compare against itself, which means nobody can tell what's working.

Aggressive device and schedule bid adjustments layered under Smart Bidding: someone brought manual habits to an automated setup. The system is already making its own adjustments based on richer context. The manual multipliers are mostly fighting it.

A strategy changed within the last fortnight, with performance being discussed: the numbers being argued over aren't real yet. I check change history before I look at any performance data, every time, because judging a transition as though it were a steady state is how good decisions get reversed.

What I stopped doing

Switching strategy to fix a performance problem. Most of the time the problem is the query mix, the offer, or the landing page. A switch resets learning and buys three bad weeks before you discover the cause was somewhere else entirely. I now exhaust the diagnostic before touching the bidding method.

Setting the target I wanted rather than the one the account was hitting. An unreachable target does not make the system try harder. It makes it bid on fewer auctions, which reduces volume, which reduces data, which makes the target even less reachable. It's a spiral and I've watched it run for months.

Changing bid strategy and budget in the same week. Then nothing is attributable and the post-mortem is guesswork. One variable per learning cycle.

Treating interface recommendations as decisions. They optimise for adoption of a feature, not for the economics of your specific business. I read them, and I apply them as offered rarely.

Judging a new configuration before the learning period closes. Two weeks of ugly numbers after a switch is expected behaviour, not evidence of failure. Reacting inside that window resets the clock and guarantees a third ugly week.

Believing a strategy has a fixed personality. The same setting behaves completely differently on twenty conversions a month than on two thousand. Volume is the variable that matters. The name on the setting is almost incidental by comparison.

Optimising toward the metric the strategy reports. A strategy pursuing clicks will report improving click metrics. That tells you the system is working, not that the business is. I judge every configuration on cost per acquisition and revenue, never on the number it was built to make look good.

The practical takeaway

Choose on data first and goal second. Verify the tracking, count the conversions, check whether values genuinely differ and only then pick from the map. A configuration chosen this way survives contact with the account. One chosen by reputation, or by what worked in someone else's ad campaigns, generally doesn't.

Match the choice to the stage rather than to the ambition. Maximisers while you're learning, targets once you're stable, value-based only when the values are real and differentiated. Skipping a stage doesn't accelerate anything; it just moves the learning period somewhere less convenient.

And treat google ads bidding strategies as instructions rather than as performance features. Change one thing at a time and wait a full learning cycle before judging it. The most common reason a bid strategy "doesn't work" is that it was never given long enough to be evaluated fairly and the second most common is that it was evaluated on the wrong number.