Google Ads Bidding Strategies: The Complete Guide

A bid strategy answers one question per auction: what is this specific impression or click worth, and should we buy it. It does not decide who sees the ad (that's audience targeting and match type), how much total money is available (that's the daily budget), or which ad copy shows. Keeping those apart matters because delivery problems get misdiagnosed constantly: an account that isn't spending gets blamed on "the bid strategy" when the real constraint is a Target Impression Share ceiling set below what the position currently costs, or a budget too small to clear the auction even once.

The billing unit is the first real fork. Most Search strategies bill on the click - you pay when someone clicks, whether or not the impression that led there was ever seen. Viewable CPM inverts that: it bills per thousand impressions that met Google's viewability threshold, whether or not anyone clicks, and it exists only on the Display Network as a reach and awareness buy. CPV bidding is a third unit again, specific to YouTube - you pay for a qualifying watch or an interaction, not for the impression or the click. These three units don't compete for the same budget line inside one campaign; picking among them is mostly a campaign-type decision, because the campaign type is what makes the unit available at all.

The second fork is who prices the auction. A manually set bid enters every auction unchanged regardless of device, hour, location, or the specific query behind it. An automated strategy recomputes a number for every auction using signals, some of which - real-time audience composition, for instance - are never exposed for anyone to set by hand even if they wanted to. That second fork is the one the rest of this guide is organized around.

Manual CPC is the baseline every other strategy gets compared against: you set a maximum cost-per-click at the keyword or ad group level, and it holds until you change it by hand. It suits three situations well - tiny budgets that can't generate enough conversion volume for a model to learn from, brand terms where conversion rate is already high and the bid mostly just needs to clear the auction, and diagnostic periods where you're deliberately isolating one variable, like a new ad or a landing page test, and don't want an automated bidder reacting to it mid-test.

Enhanced CPC used to sit between manual and full automation - it took your manual bid and nudged it up or down at auction time based on predicted conversion likelihood. Google retired that adjustment for Search and Display, so campaigns still labeled eCPC in older account exports or shared strategy documents no longer receive any adjustment and behave as plain Manual CPC today. There is no supported halfway setting left. If what you actually wanted was the auction-time nudge, the current equivalent is Maximize Conversions, not a manual bid with a modifier.

The ceiling on manual bidding shows up as accounts grow, not on day one. You cannot react per auction, which means you're bidding blind to the exact signals that competing automated bidders are using against you in that same auction. The common misreading is that manual equals cheaper; in practice it mostly means slower to react, and the gap compounds as competitors' automated strategies get better at spotting which auctions deserve a premium. Manual bidding stays defensible longest at the opposite end from broad match - tight, well-structured exact and phrase match campaigns with a short keyword list, where there's little for an algorithm to differentiate anyway.

Two strategies are automated - Google computes the bid per auction - but neither reads a single conversion signal, which is why neither counts as Smart Bidding even though both sit in the same automated bidding menu. Maximize Clicks spends the available budget hunting the cheapest clicks it can find. Target Impression Share spends toward a chosen page position and a chosen share of impressions, up to a maximum CPC ceiling you set. They chase genuinely different things - traffic volume versus visibility - but they share the same blind spot: nothing in either one has any idea whether the traffic or the position ever produces a sale.

Maximize Clicks needs no conversion history to function, which is exactly why it's an honest choice for a brand-new campaign that has none yet, or for a pure traffic goal. Its only guardrail is an optional max CPC limit, and skipping that limit is a blank cheque - without one, the strategy will pay whatever a single expensive term happens to cost if that's what clearing the budget demands that hour. It also pairs badly with loose match types, since cheap and irrelevant are frequently the same query.

Target Impression Share ignores conversion data by design, which makes it a visibility tool rather than a performance one. Set to a high percentage at "absolute top" or "top," it will keep buying position on a term that has never converted, because nothing in the strategy can see that history. The defensible use is brand-term defense - denying a competitor the top slot on your own name, where conversion rate is already known to be high and presence is genuinely the goal - not a general non-brand strategy, where a loose ceiling just buys expensive impressions with no efficiency check attached.

Smart Bidding is Google's umbrella term for the strategies that price each auction using actual conversion signals - device, location, time of day, the query itself, and audience data, recomputed auction by auction rather than set once. Underneath the umbrella sit two base strategies: Maximize Conversions, which spends the full daily budget chasing the highest conversion count, and Maximize Conversion Value, which chases the highest total value instead of count. Both need working, accurate conversion tracking to function at all - the model trains on whatever you've told it counts as a conversion, and it optimizes toward that definition faithfully, including when the definition is wrong.

The switch from counting to valuing conversions is a data decision before it's a strategy decision. Maximize Conversion Value only behaves differently from Maximize Conversions when the values attached to conversions actually vary - a $40 order and a $400 order need to arrive as $40 and $400, not both hard-coded to a flat number. A hard-coded value turns Maximize Conversion Value into Maximize Conversions wearing a ROAS label: the report shows a confident-looking return figure that means nothing, because every conversion looked identical to the model. Ecommerce accounts with real basket-size variation, and lead-gen accounts that score leads and feed real values back through offline conversion import, are the accounts where this switch is worth making.

The single most common failure mode with either strategy is flipping to it while tracking is broken or counting the wrong action - a thank-you page that also fires on a bounced visit, a lead form action that quietly counts newsletter signups as leads. Smart Bidding will spend the budget finding more of exactly that, confidently and at scale, before anyone notices the conversion count climbed while pipeline didn't. Confirm the conversion action in GA4 conversion tracking or the native Google Ads tag actually reflects the outcome you want more of before handing bidding to it, not after.

Target CPA and Target ROAS are not standalone strategies you pick from a menu - they're optional constraint fields that live inside Maximize Conversions and Maximize Conversion Value respectively. Leave the field blank and the strategy spends the full budget chasing the most conversions or the most value it can find; fill it in and Google starts declining auctions it predicts will miss the number, trading volume for cost discipline. Picking "Target CPA" and picking "Maximize Conversions" are the same decision at two levels of specificity, which is worth knowing because troubleshooting guides that treat them as separate strategies will send you looking in the wrong place.

Target CPA names a currency amount - the average you want to pay per conversion - and it's an average, not a cap; individual conversions land above and below it every week, sometimes by a lot. Suppose an account has been running Maximize Conversions with no target and is averaging $340 per conversion at steady volume. Adding a target CPA of $300 doesn't renegotiate every auction down to $300 - it tells the model to decline auctions it predicts will land meaningfully above that number, shrinking the pool of auctions it's willing to enter. Set the target near what the account already achieves and move it in small increments from there. A target set well below the achievable average - say, $150 against that same $340 baseline - doesn't produce cheaper conversions; it just throttles volume, sometimes down to almost nothing, because the model can't find enough auctions that clear the bar.

Target ROAS works the same way but as a percentage - conversion value divided by spend - and it only means anything when conversion values are real and vary. Suppose an account spends $10,000 and returns $30,000 in tracked conversion value with no target set, a 300% ROAS. Setting a target ROAS of 400% doesn't just re-price the existing traffic; it tells the model to decline auctions predicted to return below a 4-to-1 ratio, cutting into the lower-value orders that were pulling the blended average down. Expect total value and volume to fall, not just the ratio to rise - that's the trade the setting makes on purpose, not a sign it's misconfigured. Aspirational targets an account has never actually reached in unconstrained delivery are the standard way to starve a campaign; both fields punish a number picked from a goal rather than a number pulled from recent performance.

Because tCPA and tROAS sit inside different base strategies, they can't be mixed inside one campaign - a campaign runs Maximize Conversions or Maximize Conversion Value, never both. Where an account genuinely needs both a cost ceiling and a value signal, the usual pattern is splitting traffic across two campaigns rather than looking for a combined setting that doesn't exist. And after any target change, give the account through the learning period before judging it - a target moved sharply asks the model to re-price every auction against a number it has no history near, which shows up first as a delivery drop, not a clean efficiency win.

A portfolio strategy isn't a competing option next to Maximize Conversions or Target ROAS - it's a container, stored at account level, that applies one of those strategies across several campaigns at once and pools their conversion data toward a single shared target. The decision it represents is scope, not strategy: standard, per-campaign application, or shared application across a group. Every portfolio strategy is still a Smart Bidding strategy underneath, with one exception - Target Impression Share is also portfolio-eligible despite not being Smart Bidding at all.

The case for a portfolio is thin data. A campaign generating only a handful of conversions a week rarely gives a per-campaign model enough to learn from; grouped with three or four similar campaigns under one portfolio target, the combined volume can be enough. The cost is control - a portfolio optimizes toward one blended number, and it can genuinely starve one campaign in the group to fund another that's converting more efficiently against the shared target, with no per-campaign override to stop it. It also shares bidding, not budget: each campaign keeps its own daily amount unless a shared budget is layered on separately.

Portfolios expose bid limits and spend targets that campaign-level strategies don't offer, which is useful for managing a group of similar campaigns - several geographies of the same offer, for instance - against one blended CPA or ROAS rather than negotiating each one separately. For a small advertiser running one or two campaigns, a portfolio adds a layer of account-level configuration with nothing yet to pool against. It's worth adopting once there's a genuine group of similar, thin campaigns, not before.

Viewable CPM and CPV bidding don't compete with anything above - they belong to Display and YouTube campaigns built for reach and attention rather than clicks, and neither reads a conversion signal at all. Viewable CPM names the most you'll pay per thousand impressions that met Google's viewability standard, and billing fires only on impressions that qualified; it's a Display awareness buy, full stop. Nothing in it predicts or pursues a click, so pairing it with a conversion goal and then judging it on cost per conversion produces a bad-looking report every time - it was never bidding for a conversion to begin with.

CPV bidding is YouTube's equivalent: you set the most you'll pay per qualifying view - typically a 30-second watch, the full video if it's shorter, or an interaction with the ad - and a skipped pre-roll before that threshold costs nothing. It suits upper-funnel video where completed attention is the actual goal. The trap is reading a low CPV as success on its own; cheap views often come from passive, low-attention inventory, so it needs lift measurement or downstream conversion tracking layered on top to mean anything, not the view count by itself.

Comparing either of these to a click- or conversion-billed strategy on cost per conversion is a category error - the billing unit and the thing being optimized are different in kind, so the fair comparison is something like cost per reached user or cost per site visit, not a shared cost-per-conversion column in the same report.

Sequence, don't jump. A new campaign with zero conversion history has nothing for an automated conversion strategy to learn from, so it starts on Maximize Clicks with a max CPC limit set, or on tight Manual CPC, purely to gather traffic and let conversion tracking prove itself. Once conversion tracking is verified accurate - check a handful of conversions manually against what actually happened, don't just trust the count on the dashboard - move to Maximize Conversions with no target, and let it run unconstrained for a few weeks to find its natural cost per conversion. Only after that number is stable does adding a target CPA make sense, and the target should start close to that observed number, not a number picked from a business goal.

The most common migration mistake is skipping straight from a brand-new campaign to a tight target CPA or target ROAS because a case study said so. With no baseline to set the target near, the number is a guess, and a guessed target that's too aggressive doesn't fail loudly - it just quietly under-delivers, spending less of the budget every day while looking, on the surface, like a normal automated campaign that's simply "finding fewer good auctions." That's often mistaken for a demand problem when it's actually a target problem.

The same logic applies moving from Maximize Conversions to Maximize Conversion Value: don't make the switch until conversion values are real and variable, because flipping the strategy with a flat or hard-coded value changes nothing except the label on the report. And every time a target, a conversion action, or the strategy itself changes, expect a learning period where performance data is unreliable before it's a clean read again - judging a change during that window is the single most common reason accounts bounce between strategies without ever letting one prove itself.