The single biggest predictor of whether Smart Bidding works in a Google Ads account is not which strategy you pick — it is whether you have enough clean conversion data to feed it. Across the accounts we audit, the most common cause of “Smart Bidding isn’t working” is a strategy chosen for an account with too few conversions, a target set below what the account has ever achieved, or conversion tracking that quietly feeds the algorithm noise. Get the data foundation right and the choice between Target CPA, Target ROAS and Maximise Conversions becomes a straightforward decision based on your business model and volume. This guide gives you the thresholds, a decision tree, and the transition rules that protect performance when you change.
Quick reference — Smart Bidding by business model (AU):
- Lead generation → Maximise Conversions first, then Target CPA once you have ~15–30 conversions in 30 days
- E-commerce / revenue goal → Maximise Conversion Value, then Target ROAS once values are flowing and stable
- Volume floor for value-based bidding (tROAS): aim for ~50+ conversions in 30 days before constraining with a target
- Volume floor for tCPA: ~15–30 conversions in the last 30 days per campaign
- Learning period after any change: ~7 days (often 1–2 weeks to stabilise); expect volatility, don’t judge inside it
- Set tCPA at or slightly above your trailing CPA — a target below historical performance throttles delivery
What is Smart Bidding and what does it actually optimise?
Smart Bidding is Google Ads’ set of automated bid strategies that use machine learning to set a bid for every individual auction, using signals — device, location, time, audience, query context and dozens more — that no manual bidder could process in real time. The four strategies that matter for most accounts are Maximise Conversions, Target CPA (tCPA), Maximise Conversion Value, and Target ROAS (tROAS).
They split cleanly into two families. The conversion-count family (Maximise Conversions, Target CPA) treats every conversion as equal — one lead is worth the same as another. The conversion-value family (Maximise Conversion Value, Target ROAS) optimises for the dollar value you assign to conversions, so a $4,000 sale counts for more than a $40 one. Choosing the wrong family is the first and most expensive mistake: running Target CPA on an e-commerce account with wildly different order values tells Google to chase cheap conversions regardless of revenue, and it will happily do exactly that.
Everything that follows assumes one non-negotiable foundation: accurate conversion tracking. If your conversions are double-counted, firing on the wrong page, or missing offline sales, Smart Bidding optimises toward a distorted picture. Fix tracking first — our conversion tracking guide for Australia and enhanced conversions setup cover the groundwork.
How much conversion data do you need before you constrain bidding?
This is the question that decides everything, and most accounts get it wrong by being impatient. Smart Bidding needs a baseline of conversions to model the auction; the tighter the constraint you apply (a target), the more data it needs to hit that target reliably.
A practical, account-tested rule of thumb:
- Maximise Conversions / Maximise Conversion Value — usable almost immediately. These have no target to satisfy, so they are the right starting point for newer campaigns or those with thin data. They spend your full budget chasing the most conversions or value they can find.
- Target CPA — wait for roughly 15–30 conversions in the trailing 30 days at the campaign level before applying a target. Below that, the algorithm doesn’t have enough signal to hit a CPA reliably, and you get erratic spend.
- Target ROAS — the most data-hungry. Aim for 50+ conversions in 30 days with values attached before you constrain with a ROAS target. Value-based optimisation needs both volume and a spread of conversion values to learn from.
These are not Google’s hard minimums — campaigns can technically run a target on less — but they are the levels at which the strategies behave predictably in the accounts we manage. Below them, you are asking a model to hit a precise target with too little evidence, and the result is the stop-start spending that gets blamed on “the algorithm.”
If a single campaign can’t reach these thresholds, consolidate. Three thin campaigns each with eight conversions a month will each struggle; merged into one with 24, Smart Bidding has something to work with. Over-segmented account structures are one of the most common reasons Smart Bidding underperforms.
When should you use Maximise Conversions?
Maximise Conversions tells Google to get as many conversions as possible within your daily budget, with no cost-per-acquisition constraint. Use it when:
- The campaign is new or has fewer than ~15 conversions in 30 days and you need to build data before applying a target.
- You are budget-constrained rather than CPA-constrained — you have a fixed monthly spend and simply want the most leads it can buy.
- You are transitioning an account onto Smart Bidding for the first time and want a clean baseline CPA before deciding what target is realistic.
The risk to manage is that Maximise Conversions will spend your entire budget every day and, without a target, can chase expensive conversions if that’s what hits the count. Keep budgets at a level you’re comfortable spending in full, and watch the resulting average CPA — that number becomes the input for your eventual Target CPA. You can also add an optional Target CPA to Maximise Conversions, which is functionally the modern way to run tCPA.
When should you use Target CPA?
Target CPA optimises for the most conversions at, or below, a cost-per-acquisition you specify. It is the workhorse for lead generation — services, B2B, trades, professional services — where every lead is roughly comparable in value and you know what a lead is worth.
Apply it when you have the ~15–30 conversions/30 days baseline and a defensible target. The critical rule: set the target at or slightly above your trailing CPA, never below it. If your account has historically converted at $80 and you set a $45 target because that’s what you’d like, Google restricts delivery to only the cheapest auctions it’s confident about, volume collapses, and you conclude Smart Bidding “doesn’t spend.” It’s doing exactly what you told it. Lower the target gradually — 10–15% at a time — and let each step stabilise for a week or two before tightening further.
Target CPA suits the service verticals we write about most — see how it plays out in allied health and mortgage broker accounts, where lead quality and cost-per-lead are the metrics that matter.
When should you use Target ROAS or Maximise Conversion Value?
The moment your conversions have meaningfully different dollar values — almost always e-commerce, but also any business that can pass back real revenue — you should be in the value family, not the count family.
Maximise Conversion Value is the starting point: get the most total revenue from your budget, no target. Run this until value is flowing reliably and you’ve established a baseline ROAS.
Target ROAS then constrains that to a return target — a tROAS of 400% tells Google to aim for $4 of revenue per $1 spent. It needs the most data of any strategy (the 50+ conversions guidance) because it’s modelling not just whether someone converts but how much they’re worth. The same discipline as tCPA applies in reverse: set a tROAS at or slightly below your trailing ROAS. Demand 800% from an account averaging 400% and delivery throttles to almost nothing.
Two foundations make value bidding work, and both are commonly missing:
- Accurate conversion values must be sent back to Google. Static or guessed values cripple the strategy. Ideally pass real order value, and where possible feed net-of-returns or margin-based values so the algorithm optimises for profit, not vanity revenue.
- A spread of values to learn from. If every order is $49, value bidding behaves almost like conversion-count bidding. The more your order values vary, the more there is to optimise.
For stores, this connects directly to feed and Shopping setup — our e-commerce Google Ads guide covers the value-tracking groundwork that tROAS depends on.
How do you transition between bid strategies without losing performance?
Changing a bid strategy resets — or at least disrupts — the learning. Every account we move onto a new strategy goes through a learning period of roughly 7 days, sometimes 1–2 weeks, during which performance is volatile and not representative. The mistakes that turn a normal transition into a crisis:
Judging performance inside the learning window. CPA spikes and conversions wobble while the model recalibrates. If you panic and revert after three days, you reset the learning again and never escape the volatility. Decide the change, then leave it alone for a fortnight.
Making big jumps. Going from manual CPC straight to an aggressive Target ROAS in one move is a shock. The smoother path is manual/eCPC → Maximise Conversions (build baseline) → Target CPA or tROAS (apply realistic target) → tighten gradually.
Changing the target too hard. Moving an existing tCPA or tROAS more than ~15–20% in one step re-triggers learning. Step changes, with a week to settle between each, keep delivery stable.
Stacking changes. Don’t change the bid strategy, the budget, the targeting and the ad copy on the same day — when performance moves, you won’t know which lever caused it. Change one thing, measure, then change the next.
Ignoring budget as a constraint. A Target ROAS campaign that’s budget-limited can’t optimise properly — it’s being throttled before the strategy gets to work. If a strategy is “limited by budget,” that flag matters more than the target.
Common Smart Bidding mistakes that waste budget
- Choosing the wrong family — running Target CPA on an e-commerce account with varied order values, so Google chases cheap conversions and ignores revenue.
- Setting targets below historical performance — a tCPA under your trailing CPA, or a tROAS above your trailing ROAS, throttles delivery and gets misread as “won’t spend.”
- Applying a target with too little data — constraining a campaign with eight conversions a month instead of running Maximise Conversions until volume builds.
- Over-segmentation — splitting conversions across so many campaigns that none reaches the volume Smart Bidding needs.
- Feeding noisy conversion data — double-counted conversions, the wrong conversion actions marked “primary,” or guessed values. Smart Bidding amplifies whatever you feed it.
- Reacting inside the learning period — reverting changes after a few days and resetting the model repeatedly.
- Counting low-value actions as conversions — optimising toward newsletter signups or page views dilutes the signal away from the leads or sales that pay the bills.
This last point is worth dwelling on: Smart Bidding is only ever as good as your conversion definitions. If you find it’s optimising toward the wrong outcomes, the fix is usually in the conversion setup, not the bid strategy. The same principle runs through the most common account problems we document in our 14 most common Google Ads mistakes.
If your account is on Smart Bidding but spend is erratic, CPAs are climbing, or a strategy “won’t spend,” the cause is almost always one of the data or target issues above — not the strategy itself. A structured review against our 45-point Google Ads audit checklist surfaces the conversion-tracking gaps and target mis-settings that quietly hold Smart Bidding back.
Book a free 30-minute review at get in touch and we’ll tell you whether your bid strategy, your targets, or your tracking is the thing costing you conversions.
Before you set a budget, use our free Minimum ROAS & break-even calculator to work out the return your campaigns need to be profitable.
Frequently Asked Questions
How many conversions do I need before using Target CPA?
As a practical rule, aim for roughly 15–30 conversions in the trailing 30 days at the campaign level before applying a Target CPA. Below that, the algorithm lacks the signal to hit a CPA target reliably and spend becomes erratic. If a campaign can’t reach that volume, run Maximise Conversions to build data first, or consolidate thin campaigns so the conversions pool together.
Should I use Target CPA or Target ROAS?
Use Target CPA when every conversion is worth roughly the same — typical of lead generation, services and B2B — and you optimise for cost per lead. Use Target ROAS (or Maximise Conversion Value) when conversions have meaningfully different dollar values, which is almost always e-commerce, and you can pass real revenue back to Google. Choosing the count family for a value-driven business is one of the most expensive Smart Bidding mistakes.
Why is my Smart Bidding campaign not spending its budget?
The most common cause is a target set tighter than the account’s historical performance — a Target CPA below your trailing CPA, or a Target ROAS above your trailing ROAS. Google restricts delivery to only the auctions it’s confident can hit that target, so spend collapses. Loosen the target toward your actual historical figure, then tighten gradually, 10–15% at a time, letting each step stabilise.
How long is the Smart Bidding learning period?
Roughly 7 days after a strategy or significant target change, though it can take 1–2 weeks to fully stabilise. Performance is volatile during this window and should not be judged. Avoid reverting changes early, because doing so re-triggers learning and traps the account in permanent recalibration. Make one change, then leave it for at least a fortnight before evaluating.
Can I switch bid strategies without losing performance?
Yes, but switch gradually. Move from manual bidding to Maximise Conversions to build a baseline, then to Target CPA or Target ROAS with a realistic target, tightening in small steps. Avoid changing the bid strategy, budget, targeting and creative all at once, and never judge the result inside the learning period. Abrupt jumps and stacked changes are what cause the performance loss people blame on automation.
Does Smart Bidding work with broad match keywords?
It can work very well, because Smart Bidding sets a per-auction bid that accounts for the wider range of queries broad match brings in — but only with strong conversion tracking, tight negative keywords and enough volume. With weak tracking or limited budget, the combination can burn spend fast. We cover exactly when the pairing helps versus hurts in our guide to broad match in 2026.