Google Smart Bidding: Which Strategy When

Choosing a bid strategy is the highest-leverage decision left in a Google Ads account, precisely because the platform automated everything downstream of it. You no longer set bids. You choose what the machine optimizes for and what data it learns from, and both of those are decisions the automation cannot make for you. Most underperforming accounts I look at did not pick a bad strategy - they picked a strategy their conversion data could not feed.

This guide covers the four Smart Bidding strategies, what each one actually optimizes, the data volume each one needs by Google’s own numbers, and the order I would try them in. Everything product-specific here links to Google’s documentation, because bid strategy folklore ages badly and the official pages are the only moving part I trust to stay current.

What Smart Bidding is, and what it isn’t

Google’s definition: Smart Bidding refers to bid strategies that use Google AI to optimize for conversions or conversion value in each and every auction, a feature known as auction-time bidding. Four strategies qualify: Target CPA, Target ROAS, Maximize conversions and Maximize conversion value.

Two clarifications spare a lot of confusion. First, Maximize Clicks is automated bidding but not Smart Bidding - it optimizes clicks, not conversions, and I treat it as a diagnostic tool for brand-new accounts rather than a place to live. Second, the four strategies are really two pairs: Maximize conversions with an optional Target CPA constraint, and Maximize conversion value with an optional Target ROAS constraint. Uncapped, the strategy spends your budget chasing volume. With a target, it chases volume at your stated efficiency. Understanding them as pairs is what makes the choice legible.

Maximize conversions: the volume default

Uncapped Maximize conversions has one goal: as many conversions as your daily budget can buy. It is where I start most lead-gen campaigns, for an unglamorous reason - it has no efficiency target to miss, so it works with thin data while the account accumulates the conversion history the target-based strategies need. The cost is that “as many as possible” contains no opinion about price per conversion. Watch the first weeks and expect CPA variance.

The move that matters is knowing when to leave. Once the campaign converts steadily and you can see what conversions have been costing, adding a Target CPA turns volume-chasing into volume-at-a-price. Which brings me to the target itself.

Target CPA: efficiency for lead gen

Target CPA sets bids to get as many conversions as possible at your target cost per action, adjusting for real-time signals like device, browser, location, time of day and remarketing list at auction time. It fits businesses where each conversion is worth roughly the same - demo requests, quote forms, sign-ups.

Two practices decide whether it works. Set the initial target near your recent actual CPA, not your aspiration - a target far below what the account has been achieving just starves delivery, and you walk it down in steps of 10 to 15% afterwards. And judge results only over windows with real volume: Google’s own documentation recommends measuring performance over periods that have at least 30 conversions. A week with six conversions tells you nothing except that you looked too soon. I have seen more Smart Bidding “failures” caused by impatient target changes than by the bidder itself.

Maximize conversion value and Target ROAS: for unequal conversions

When conversions differ in value - an ecommerce catalog, lead tiers you score differently - optimizing for conversion count actively misleads the machine, because it will happily buy a hundred cheap conversions over ten valuable ones. Maximize conversion value optimizes revenue within budget, and Target ROAS adds the efficiency constraint: bids are set to maximize conversion value while aiming at your average return on ad spend.

Target ROAS is the most demanding strategy in the lineup, in two ways. It only knows what your conversion tracking tells it, so value tracking has to be genuinely right - the first thing I check in any account running it is whether the conversion values actually vary, because if every conversion reports the same value, you are running an expensive version of Target CPA. And it needs volume. Google’s documented minimums before using it: at least 15 conversions in the past 30 days for Search and Shopping, and materially more elsewhere - 50 in 35 days for Demand Gen, 30 for Video Action, 50 a week for Hotel campaigns. Below those numbers, the honest choice is Maximize conversion value uncapped, or Target CPA, until the history exists.

Choosing, in one pass

My decision sequence, which covers nearly every account I see:

1. Do conversions have meaningfully different values, and do you track those values? Both yes: the value pair - Maximize conversion value, adding a Target ROAS once past the volume minimums. Otherwise: the conversion pair.
2. Do you have conversion history and a defensible cost target? Yes: Target CPA at roughly your current actual. No: uncapped Maximize conversions until you do.
3. Is the campaign brand new with no conversion tracking yet? Fix the tracking first. No bid strategy rescues a campaign that measures the wrong thing, and Maximize Clicks is acceptable only as scaffolding while conversions start flowing.

The pattern underneath: the constraint (a target) is something you earn with data, not something you start with because it feels prudent.

The learning period, and the thrashing trap

Every change you make sends the bidder back to school. Google’s guidance for its most automated campaign type is explicit about the timescale: ramp up for one to two weeks after launch, run at least six weeks before evaluating, and exclude the ramp from your analysis - and everything worth locking down before that clock starts is in my Performance Max guardrails checklist, with the full build order in the setup walkthrough. For Target ROAS specifically, the documentation advises waiting one to two conversion cycles after changes to how you track value.

The failure mode this implies is the one I actually see: thrashing. The marketer changes the target on Tuesday, panics at Thursday’s numbers, changes it again, and the bidder spends its whole life in a learning period the account’s owner keeps restarting. My working rules: one change at a time, in small steps, on a calendar, with the evaluation window agreed before the change is made. It is the same discipline as any honest holdout test - decide how you will judge before you act, then let the clock run. And when a live strategy is already slumping, my diagnostic order runs before any change gets made, because half of what looks like a bidding problem is a measurement one.

What stays your job

Smart Bidding moved the work, it did not remove it. The machine owns the bids. You own everything it learns from, and that is where accounts are won now:

• Conversion definitions. The bidder optimizes exactly what you count. Count newsletter sign-ups next to purchases and it will buy you newsletter sign-ups. Audit what is marked as a conversion before blaming the algorithm for chasing it.
• Values that mean something. For the value strategies, feeding real margins or scored lead values beats revenue, and beats flat placeholder values by miles.
• The creative and the offer. Auction-time bidding decides how much to pay for a click your ad still has to earn. The ad copy work did not get automated, only the bidding under it - and since the ad itself is now assembled per auction from your asset pool, feeding that combination machine is its own discipline.
• Spend skepticism. My standing bias in paid media applies double here: the platform profits from your spend, and “the algorithm needs more budget to learn” is sometimes true and sometimes a sales pitch. The counterweight is your own measurement - a target grounded in what a conversion is actually worth to you, checked against results the platform did not grade for itself.

FAQ

Is Maximize Clicks a Smart Bidding strategy? No. Google’s Smart Bidding umbrella covers the four conversion-based strategies only. Maximize Clicks automates toward clicks within budget - useful for a week of traffic data on a new account, and a poor place to leave money long-term, because clicks are an input you buy, not an outcome you want.

My account is small. Can I use Target ROAS? Check yourself against Google’s minimums honestly - 15 conversions in 30 days for Search is the floor, not the comfortable amount. Below or near it, run Maximize conversion value without a target, or consolidate campaigns so the conversion data pools instead of fragmenting across five campaigns that each starve.

Should I follow the recommendations tab when it suggests changing my bid strategy? I read those recommendations the way I read any vendor’s ROI calculator: as the platform’s opinion about the platform’s product. Some are sound. Evaluate them against your data and your margins, apply the ones that survive, and never auto-apply bidding changes.

How do I know if Smart Bidding is actually beating what I’d do manually? Run the comparison as an experiment instead of a feeling. Google Ads has built-in campaign experiments that split traffic between two bid strategies - same discipline as any incrementality test: one change, a pre-agreed metric, and enough conversions before judgment.

This post is part of the paid media guides, where the same spend discipline gets applied to ads inside ChatGPT. The copy the bidder delivers is covered in ChatGPT ad copy prompts, the whole seat is mapped in AI for PPC managers, and the wider practice starts at the complete guide to AI marketing.