Google Ads Bid Strategy: How to Choose One and When to Change It

How to pick a Google Ads bid strategy, what each one optimizes for, and why your conversion volume decides the answer more than the dropdown does.

By Gavin Sevastian · Updated August 4, 2026 · 13 min read

Most contractors asking which Google Ads bid strategy to use are starting at step three. On a small local account, the strategy is rarely the thing holding performance back. The conversion data sitting underneath it is.

The order that works: confirm what your conversion action counts, count how many real conversions you got in the last 30 days, then pick the strategy that matches that number. Jumping straight to the third step is the most common wrong move in this topic, and it stays invisible for weeks.

Two recent changes date most of what has been written about this. In June 2026 Google renamed "Maximize conversions with a Target CPA" to "Target CPA" and "Maximize conversion value with a Target ROAS" to "Target ROAS", stating that "the underlying bidding behavior remains exactly the same". On August 17, 2026 it changes how targets behave on budget-limited campaigns. Anything written in 2024 is partly out of date, including some of what still ranks.

Do this first: tracking, then volume, then strategy

Google states the sequence itself, in its documentation on how our bidding algorithms learn: "After you set up conversion tracking, wait for a few conversion cycles before implementing a new bid strategy so that our algorithms can adjust."

1. 1

   Confirm what your conversion action counts

   Open each conversion action and read what fires it, not what it is called. An action named Lead can be counting a button press, a page view, or a two-second call. A tag on a thank-you page is a real event. A tag on the contact page is a page view wearing a lead's name.

2. 2

   Count real conversions in the last 30 days

   One number, from the conversion action you are bidding toward, over a full 30 days. Not clicks, not form starts, not the total across every action in the account. Write it down.

3. 3

   Pick the strategy that matches that number

   The number you just wrote down is the constraint. The rest of this guide is about matching a strategy to it honestly rather than optimistically.

Step one takes ten minutes and decides more than the dropdown does. A conversion-based strategy is a prediction model. It predicts from the events you told it to care about, and it has no opinion about whether those events are any good.

If the number in step two is small, that is useful information rather than a failure. Most one-truck operations produce a small number. Knowing it is what stops you from choosing a strategy that quietly needs ten times more data than you have.

People skip to step three because it feels like the only step that does anything. It is also the step that produces no visible feedback for weeks, because the system is recalibrating and the numbers move around while it does. By the time performance looks clearly wrong, three or four other things have usually been changed on top of it, and nobody can tell which one caused what.

What each Google Ads bid strategy optimizes for

Seven options is a lot of dropdown for one decision. They get easier to think about in three groups, sorted by what each one is trying to buy. The system will do exactly what you asked, including in the cases where what you asked is not what you meant.

Strategies that buy traffic, not outcomes

Manual CPC, Maximize Clicks and Target Impression Share all buy activity. None of them knows what a lead is.

Manual CPC sets one price before the auction and leaves it there. Google's framing: your max CPC bid is the most you will be charged for a click, and you will often be charged less. Manual CPC and Target Impression Share are the two strategies with no learning period, which matters more than it sounds when you are trying to get a clean read on something else you changed.

Maximize Clicks "adjusts maximum cost-per-click (CPC) and sets your bids to help get as many clicks as possible within your budget," and Google notes it "doesn't optimize towards impression share." You can add a CPC bid limit, and you usually should, but Google warns that if the limit is too low the strategy may not be able to reach its goal at all.

Target Impression Share buys visibility and nothing else. Same trap in a sharper form: its max CPC bid limit will silently stop it hitting its own target if you set it too conservatively, so you end up paying for a strategy that cannot do the one job you gave it.

Not knowing what a lead is turns out to be a feature when you have no conversion data, because there is nothing for a smarter strategy to be smart about. It becomes a liability the moment you do.

Local Services Ads work differently and are not covered by any of this. LSA versus Google Ads deals with them separately.

Strategies that buy conversion count

Maximize Conversions and Target CPA both require conversion tracking. Google is direct about it: you must set up conversion tracking to use them. These, along with Maximize Conversion Value and Target ROAS, are the four strategies Google calls Smart Bidding. Maximize Clicks and Target Impression Share are automated, but they are not Smart Bidding, and the distinction matters because almost every warning written about Smart Bidding does not apply to them.

Maximize Conversions has no cost ceiling of its own. Google describes these strategies as "designed to spend the full daily budget," and adds the sentence most people find out about the hard way: "Maximize conversions will try to fully spend your average daily budget, so if you're currently spending much less than your budget, Maximize conversions could increase spend significantly." If your daily budget was set as a theoretical maximum you never reached, switching to this strategy turns that number into an instruction.

Target CPA adds a constraint the system bids against, which is a different animal. It can also be switched on with nothing behind it, which is worth knowing precisely because of how often the opposite is claimed.

One consequence is reasoning rather than documentation, so take it as reasoning: a strategy maximizing conversion count is indifferent between a small repair inquiry and a full replacement. Both are one conversion. Given the choice it will buy the cheaper one, because that produces more conversions per dollar. For a trade where job sizes differ by an order of magnitude, that is not a small detail.

Strategies that buy conversion value

Maximize Conversion Value and Target ROAS fix exactly that problem, and they need something most contractor accounts never send: values attached to conversions, not just conversions.

Target ROAS carries the only hard numeric entry requirement on Search. Google's wording: "At least 15 conversions in the past 30 days at the conversion tracking level," which applies to Search and Shopping campaigns. You also have to set values before you can apply the strategy at all.

Sending real values back means knowing which inquiries became booked jobs and what those jobs were worth, which lives in your CRM rather than in Google Ads. That loop gets its own section at the end, because it is the part of this work that decides whether any of the rest of it pays.

Conversion volume is the constraint, not the dropdown

Google permitting a strategy and that strategy working well are two different questions, and most writing about bidding blurs them together.

The numbers Google and independent research publish

Three figures get conflated constantly. Separating them clears up most of the confusion.

Fifteen conversions in 30 days is a real requirement, and it belongs to Target ROAS on Search and Shopping. Thirty conversions is not a gate on anything: it is Google's guidance for evaluation, which is to measure over a period long enough to contain at least 30 conversions before drawing conclusions. And Target CPA has no minimum whatsoever. Google says outright that advertisers can start using it with no conversion history.

For where performance becomes predictable rather than merely permitted, the most useful public read comes from Optmyzr's September 2024 study of 14,584 accounts, The Impact of Bidding Strategies on Google Ads Performance: "The threshold for any bidding strategy to be predictably successful is 50+ conversions." The same study found something less convenient, which is that "the jump from under 25 conversions to 25 to 50 conversions doesn't always result in a performance improvement." More data helps, but not smoothly.

What those conversion counts cost

Converted into spend, the thresholds get concrete. The arithmetic below is a published cost per lead multiplied by a conversion count, so you can reproduce it and disagree with it.

Those cost-per-lead figures come from LocaliQ's 2025 home services search advertising benchmarks. They are cross-account US averages across very different markets, offers and account quality, so treat them as a sense of scale rather than a forecast for your business.

What they show clearly is that the spend required to reach the volume where bidding gets predictable varies by roughly four times depending on the trade. Roofing is the extreme case in both directions, which is worth reading alongside how to get roofing leads and the broader question of what Google Ads cost for contractors.

Why a lot of Ontario contractors sit below that line

According to Innovation, Science and Economic Development Canada, 63.3% of Ontario's construction employers have between one and four employees, and 98.9% have fewer than 100 staff.

Nobody publishes conversion volume by firm size, so the bridge from there is an inference, and it is worth making the arithmetic visible rather than asserting it. Fifteen tracked conversions a month at the published home services average implies roughly US$1,364 in monthly ad spend. That is a real commitment for a one-to-four-person operation, and 15 is the smallest of the three thresholds on this page.

None of that argues for giving up. It means the realistic path runs through the simpler strategies for longer than most advice assumes, and that the work in the meantime is accumulating real conversion events rather than shopping for a better dropdown.

Wrong conversion tracking is worse than none

The tracking step is not paperwork you finish before the real work starts. Everything a conversion-based strategy does is a prediction, and that prediction is trained on the events you nominated. Nominate the wrong events and the system does not fail. It succeeds at the wrong task.

Which events steer the bidding is a bigger lever than the strategy

Only the conversion actions included in your bidding objective steer the system. Everything else can still be recorded and reported without pulling on the bids.

Picture an objective assembled from six actions at once: a form submit, a call from the ad, a call from the website, a tap on the phone number, a chat open, and a directions request. That is one objective built from six different meanings, and the system will buy whichever of them is cheapest to produce. It is not making a mistake. It was handed six definitions of success and told to maximize the count.

Pick the one that means "a person asked us to quote a job" and keep the rest out of the objective, where they can still be counted and looked at without steering anything. That single change reshapes what the account is aiming at more than any move between Maximize Clicks and Maximize Conversions.

The events that quietly poison the model

A tap on a phone number counts whether or not the call ever connects, so mobile visitors who bounce off the page look productive. A thank-you page anyone can bookmark or reload counts return visits from customers you already have. A call action with no minimum duration counts hang-ups.

The reasonable expectation, and this is reasoning rather than a documented finding, is that all three of those are cheapest to generate from the least qualified traffic. A mistaken tap costs the same to produce as a real inquiry, and there is far more low-intent traffic than high.

If that holds, the drift is directional rather than random. Six months later the account has an impressive conversion count and a phone that mostly rings with price shoppers, and nothing in the interface points at why. Where calls specifically are the weak spot, why your Google Ads are not getting calls covers that separately.

What misconfigured means when Google flags it

Google has a status for exactly this problem. Its documentation on bid strategy statuses puts it plainly: "if you don't have the right conversion actions set up, it could impact the accuracy of your strategy's automated bids, limiting your conversions."

The status is called Misconfigured (conversion setting), and it lives in a column most people never widen. Open the campaigns table, find the bid strategy column, and read what the status says.

Assume it will not catch everything. A tag on your contact page is a technically valid conversion action, so a clean status is not evidence that your tracking means anything.

What a learning period is, and the three things that trigger one

Two beliefs about learning periods are close to universal and both are wrong. The first is that it lasts seven days. The second is that it is a penalty for having changed something. It is recalibration, and how long it takes depends on your data rather than on a timer.

1. 1

   New strategy

   The bid strategy was recently created or reactivated. Switching a campaign from one strategy to another counts.

2. 2

   Setting change

   A setting for the bid strategy was changed. Changing a target is the documented exception, covered below.

3. 3

   Composition change

   Campaigns, ad groups or keywords have been added to or removed from the bid strategy. Adding a batch of keywords, or moving a campaign in or out of a portfolio strategy, both qualify, so you can restart calibration without ever opening the bidding settings.

Those three are the list for Search campaigns. Shopping carries a fourth.

How long it lasts, and why nobody can give you a number

Google's figure, from its documentation on the duration of the learning period: "It can take up to 3 weeks or 1-2 conversion cycles for the bid strategy to calibrate to the new objective, although it can be faster depending on the amount of conversion data present."

Three things set the length: how many conversions the campaign gets, how long your conversion cycles are, and which strategy you are running.

Google defines a conversion cycle as "the amount of time it takes for a click to result in a conversion." The unit is click to tracked conversion, not click to signed job. If the event you count is a form fill or a phone call, that is often same day for a contractor, so it is rarely the cycle length that stretches your wait.

It is volume. A campaign producing a handful of conversions a month sits at the top of Google's range, and the up-to-three-weeks figure is the one to plan around.

There is no seven-day figure anywhere in Google's documentation for Search. It is a convention people repeat.

Two things that are not true about learning

Learning does not stop when the label disappears. Google: "Our algorithms continue to learn even when the bidding status no longer shows 'Learning'." The label is a notice that calibration is unusually active, not a progress bar that fills up and finishes.

Changing a target does not restart anything. From how our bidding algorithms learn, "Changing a target won't trigger a 'learning' status, and won't reset anything Smart Bidding has already learned about your account."

That is useful, because it means adjusting a target is a cheaper move than switching strategies or reshuffling keywords. It is the one lever you can pull without paying the recalibration cost. Targets carry their own risks, and they get their own section further on, but starting over is not one of them.

Which bid strategy fits where your account is

Here is the decision, made against the conversion number you wrote down earlier rather than the one you wish you had.

No tracking, or tracking you have not opened

Nothing conversion based. Maximize Clicks with a CPC bid limit, or Manual CPC if you would rather set one price and have no calibration to think about at all.

This feels like a downgrade and it is not. At this stage the job is accumulation rather than optimization. You are getting real, correctly defined conversion events into the account so that a smarter strategy eventually has something to be smart about. Buying relevant clicks at a controlled price does that well enough.

Set the CPC bid limit deliberately, because it is doing the job a target would do later. Too high and you fund the expensive end of the auction for no reason. Too low and the strategy cannot reach its own goal. If you are building the account from scratch, Google Ads for small business covers the setup around this.

Tracking working, volume still thin

Maximize Conversions with no target is the defensible position here, and there is independent support for it. The Optmyzr study found that "The majority of advertisers using Max Conversions do not set a target and see better performance on the most important KPIs like ROAS and CPA than those who do."

Google will happily let you set a Target CPA at this stage with no history behind it. A target invented before you have observed performance is a guess wearing authority: the system treats it as a constraint and bids against it, and you have no way of knowing whether it is reachable.

The signal that you are ready to move on is not a date on the calendar. It is when one bad week stops changing your mind about the account.

Enough volume that the numbers mean something

Target CPA, with the target taken from what the account has already done rather than from what you would like it to do. That derivation gets its own section below, because it is where most of the damage happens.

Target ROAS sits further away than most people expect, because it needs conversion values, and values mean knowing which inquiries turned into jobs and what those jobs were worth. The gap between this row and the last one is usually measured in quarters rather than weeks, and it is gated by conversion volume and by whether job values ever make it back into the account at all.

The specific way accounts collapse after a strategy switch

This is the failure people mean when they say Smart Bidding burned through their budget. It is common and it is not mysterious. It is a chain of ordinary things, each documented individually by Google and none of them documented as a chain.

The mechanism, step by step

It starts with a campaign moving to a conversion-based strategy with almost no conversion history behind it.

The model is not blind at that point. Google says its bidding algorithms learn at the query level rather than the keyword level, drawing on conversion data from across ad groups and campaigns, so a new campaign is not starting from nothing. But query-level signal is a weaker starting point than a campaign's own history, and it is why the early weeks are volatile.

Costs per click rise. Where the keyword base is loose and the strategy carries no cost constraint, the campaign starts entering auctions nobody would have chosen deliberately. It is pacing toward its full daily budget rather than toward any cost ceiling, because it does not have one.

Thin data cuts the other way too. Google's page on Target CPA bidding describes the constrained version of the same problem: "Setting a target that's too low, for example, may cause you to forgo clicks that could result in conversions, resulting in fewer total conversions." One route overspends into weak traffic. The other starves. From the outside they look identical for the first two weeks.

The mirror-image trap: one lucky early conversion

Two cheap conversions land in the first week. The apparent cost per acquisition looks excellent. Someone takes that number and makes it the target, on the reasonable grounds that the account just proved it can hit it.

It did not. Two events is not a rate. Under normal conditions the target now sits below anything the account can reach, so the system declines auctions it should be entering, and delivery quietly starves while every setting looks correct.

Google publishes a useful proxy for how little data is too little to trust. The FAQ for its August 2026 bidding change notes that it will not calculate a recommended target at all for campaigns "with fewer than 7 conversions because performance for these campaigns can be unpredictable." If Google declines to derive a number from that sample, you should decline too.

Why the reaction makes it worse

Here is what turns two bad weeks into a bad quarter.

Performance looks wrong, so the target gets changed. Nothing improves quickly, so the budget moves. Still nothing, so the strategy changes again. Somewhere in there a few keywords get paused.

Only one of those, the strategy change, is a documented calibration trigger. Changing the target provably is not: Google says it resets nothing. That is the trap. The reaction that feels least drastic is the one that costs nothing, while the operator assumes everything they touched contributed equally and stops being able to attribute any of it.

The account ends up permanently mid-adjustment. No window is ever clean enough to read, so every result looks ambiguous, so something else gets changed. The original switch is rarely what did the damage. The damage is that nothing after it was left alone long enough to be measured.

How to set a target without wrecking the campaign

Targets are the most misused control on this page, and the deadline in the third part of this section makes a stale one newly expensive.

A target is a constraint on the auction, not a wish

A Target CPA tells the system what an acceptable average cost per conversion looks like. It does not negotiate a better price. It changes which auctions the campaign is willing to enter.

Google's wording on what happens when you set it below what the account can do: a target that is too low "may cause you to forgo clicks that could result in conversions, resulting in fewer total conversions." You get a cheaper average and fewer jobs, and the report looks like an improvement.

So derive the target from what the account has already achieved. Look at cost per conversion over a window with enough conversions in it to mean something, then move from there in small steps. A target should be a slightly firmer version of reality. If you cannot point at the specific 30-day window your number came from, you have a wish rather than a target.

The case for not setting one at all

There is public evidence that targets hurt as often as they help. Optmyzr's read on 14,584 accounts, stated plainly in the study: "Setting targets for bidding strategies has a higher likelihood of hurting accounts than helping them."

The same study found that campaigns using Target CPA achieved the lowest CPA of any bid strategy. Both findings are true, and the tension between them is the lesson. Targets pulled out of real observed performance work. Targets invented from what someone hoped to pay do damage, and there are more accounts in the second group than the first.

Run without one until you can point at the window your number came from.

What changes on August 17, 2026

Google is changing how budget-limited campaigns on Target CPA and Target ROAS behave. Its wording, from changes to target based bid strategies: those campaigns "will more consistently perform toward your bid target."

Google's own worked example is the clearest version. A campaign with a $10 Target CPA that currently achieves $5 "will deliver more closely to a $10 actual CPA." This bites hardest if you have a target sitting well above what your account does, because until now that was a harmless placeholder and it is about to become an instruction.

The action Google names is direct: "To maintain current performance, you should update targets to match recent performance before August 17." The change does not touch Manual CPC or Target Impression Share, and it does not affect campaigns that are not budget-constrained. If your campaign spends its full budget every day and carries a generous target, you are in scope.

How to change a bid strategy without losing your read on it

Most of the damage after a switch is self-inflicted, and it comes from evaluating too early on a window that was never going to be readable.

1. 1

   Change one thing

   One strategy change, or one budget change, or one keyword batch. Not two, and never three on the same afternoon.

2. 2

   Write down the date

   Put it somewhere you will find it in a month. A note in the campaign name works. Comparing across an undated change is how accounts become unreadable.

3. 3

   Set a review date well past the change

   Count from the change, not from when you started worrying. On a low-volume account, plan around Google's up-to-three-weeks figure rather than the bottom of its range.

4. 4

   Leave it alone until then

   This is the hard one. Most material changes in between restart the clock, though changing a target is the documented exception. A restart does not erase what the system has learned, but it does forfeit the clean reading window you were waiting for.

Why one change at a time

The reason is not tidiness. Two simultaneous changes produce a result you cannot attribute to either of them, so whatever you learn is unusable and the next decision is another guess.

There is a specific version worth naming. Switching strategy and raising the budget together is extremely common, because both feel like part of the same fresh start. On a budget-spending strategy those two changes interact directly, and afterwards there is no way to tell whether the extra spend came from the new strategy behaving badly or from the budget you just handed it.

Set the review date on volume, not on the calendar

Two things can make an early read misleading, and they are different problems.

If you import booked jobs or closed work as conversions, the most recent window is structurally incomplete, because those conversions have not happened yet. Every early check reads worse than the truth, which is exactly the condition that provokes another change.

If you track the inquiry, that lag is small and the problem is volume instead. A handful of events is not a result. At small-account volumes a week is a few conversions, and the difference between three and one is noise you can feel but cannot interpret. Either way the fix is the same: count the conversions in the window before you draw anything from it, and if the count would not survive being called a sample, wait.

What a rebuild does and does not cost you

There is a widely repeated belief that rebuilding a campaign wipes out everything the account has learned. Google's documentation on how bidding algorithms learn says otherwise: they "don't have to relearn performance from scratch. Because they learn at the query level rather than the keyword level."

Learning attaches to search queries across ad groups and campaigns, not to the structures you happen to have built. So the case against rebuilding is not that you lose your history.

The genuine costs are the new campaign's own calibration period, reporting that now spans two campaigns and reads badly in every comparison, and the risk that you rebuilt around a root cause you never diagnosed.

What bid strategy cannot fix

Bidding sits downstream of tracking, structure, query relevance and the landing page. Reaching for the strategy dropdown first is the most common wrong first move, and it restarts calibration on top of not solving anything.

Excluding traffic redirects spend, it does not save it

Adding negative keywords on a budget-spending strategy does not reduce what you spend. The campaign still spends its budget. The money moves to the next-cheapest eligible traffic.

That might be an improvement. It might not. It has to be checked afterward, and the check rarely happens, because the mental model says exclusions save money and a lower spend on the excluded terms appears to confirm it.

Only a constrained strategy can genuinely decline to bid. On Target CPA, if the remaining traffic is predicted to land above the target, the system sits out and spend really does fall. This is one of the few places where the strategy you are on changes what an ordinary optimization does, so be sure which case you are in before you spend an afternoon on negatives.

No strategy repairs a mismatch between the search and the page

If someone searches for emergency furnace repair at 11pm and lands on a general heating and cooling homepage, no bidding configuration recovers that visit. The auction did its job. The page did not.

Bidding decides who sees you and what you pay. What happens after the click is decided by whether the page answers the search quickly, whether the phone number is obvious on a phone, and whether the form asks for a reasonable amount of information. Strategy changes cannot reach any of that, and a campaign with a weak page will read as a bidding problem indefinitely.

The number that decides anything is cost per booked job

Cost per booked job, measured against what a job is worth to you, is what tells you whether any of this is working. A lead price that is expensive for a small service call is cheap for a full replacement, and the gap between those two jobs in most trades is enormous.

Google Ads never sees either figure. It cannot optimize toward them and it cannot tell you when you are winning. That reconciliation is yours, and it is the subject of whether Google Ads are worth it for contractors.

Why this takes more attention than it looks like it should

The judgment in this work is not in the dropdown. It is deciding whether an account can support a conversion-based strategy at all, and being willing to say not yet. It is holding still through calibration while the phone is quiet.

The hardest part is the one the interface will never help with. At small-account volumes a week is a handful of events, so telling ordinary noise apart from real degradation is genuinely difficult, and getting it wrong in either direction costs money. Add a seasonal trade and you also have to separate the season from the strategy.

Then there is the work that lives in your CRM and your phone log: which inquiries became quotes, which quotes became jobs, and what those jobs were worth. That is the only route to value-based bidding, and it is what turns a conversion count into a business number. If you are weighing doing this yourself against paying someone, that is the work to price, not the dropdown. What Google Ads agencies charge covers the other side of that comparison.

Your next step

Do the three things from the top of this page, in order, this week. Open your conversion actions and read what fires each one. Count real conversions from the action you bid toward over a full 30 days, and write the number down. Then match a strategy to that number using the table further up.

If the honest read is that you do not have enough conversion data yet, that is a sequencing problem with a known fix: simpler strategy, controlled cost per click, correct tracking, and patience while real events accumulate. Set a review date, and leave it alone until it arrives.

If you would rather have someone else hold that line for you, here is what we do, and how it runs for contractors across the GTA and southern Ontario.