Google Ads automated bidding, explained in plain English.

Your Google Ads rep switched your campaign from Maximize Conversions to Target CPA last month, and nobody asked whether that made sense for your account first. That's usually how it happens: a note in the dashboard, a quick email saying "we've optimized your bidding strategy," and a business owner left wondering what actually changed and why the results look different now. Google gives you exactly three automated bidding strategies that matter for a local service business, and each one is simple to understand once someone explains what it's actually optimizing for, instead of just flipping the switch for you. Here is what Maximize Conversions, Target CPA, and Target ROAS really do, which one fits an account your size, and why the strategy that works great for a national ecommerce brand can quietly starve a low-volume local account.

Control, Handed To An AlgorithmThe Dial

Why your rep picked a strategy without asking you.

Open the Recommendations tab in any Google Ads account and you'll find Google actively nudging toward automated bidding, usually with a bold "Apply" button and a promised lift in conversions. Every recommendation an account applies raises something Google calls the Optimization Score, a single number out of 100 that Google displays prominently and that a lot of agencies, and more than a few in-house marketers, have quietly started treating as a report card. Apply enough recommendations and the score climbs. Nobody on the account necessarily checked whether the account had the conversion history to run the new strategy well.

That's the steering problem in a sentence: the switch gets made because the dashboard suggested it and the score went up, not because someone looked at your actual monthly conversion count and decided Target CPA or Target ROAS was the right fit for an account your size. We've taken over accounts where the bidding strategy had been changed three times in four months, each change chasing the next recommendation, with nobody ever asking whether the campaign had enough data to support any of them.

None of this makes automated bidding bad. Google's machine learning genuinely outperforms manual bid adjustments once an account has enough conversion volume to feed it. The problem is sequence: a strategy that needs conversion history applied to an account that doesn't have it yet, because a score wanted to go up, not because the account was ready.

We see a version of this constantly: a campaign gets switched to Target ROAS two months after launch, before the account has logged even its first full month of steady leads. The campaign has no real conversion value tracking in place, so every lead gets passed back to Google tagged at the same flat placeholder value, which means the algorithm has nothing meaningful to tell one lead from another on the revenue side. It isn't optimizing for anything real at that point. It's bidding against a number somebody typed in once and forgot about. Rolling an account like that back to Maximize Conversions while real value tracking gets built out typically brings the cost per lead down noticeably within a few weeks, not because the algorithm got smarter, but because it's finally running the strategy the account's actual data can support.

What each bidding strategy actually does.

Strip away the dashboard language and there are really three automated strategies worth knowing for a local service business. Each one answers a different question, and the question it answers is the whole key to picking the right one.

Maximize Conversions

Maximize Conversions answers one question: spend my whole daily budget and get me as many conversions as possible doing it. You can optionally attach a target CPA to guide it, but in its plain form it has no cost ceiling in mind at all, it's just trying to rack up volume within whatever budget you've set. That makes it the right starting point for a brand-new campaign with little or no conversion history, since it doesn't need a target to function and it builds the conversion data every other strategy depends on later.

Target CPA

Target CPA answers a different question: hit this average cost per conversion while getting as many conversions as you can at that price. You tell Google what you're willing to pay for a lead, roughly, and the algorithm bids up or down across the day and the account trying to land near that number on average. This is the strategy built for lead generation, where you know your breakeven cost and want the algorithm bidding toward it instead of chasing volume with no ceiling.

Target ROAS

Target ROAS answers a third question entirely: maximize the total dollar value of conversions relative to what I spend. That question only makes sense if you're actually tracking dollar value per conversion, not just counting leads. An ecommerce store knows a conversion's value the second the cart closes. A local service business has to build that value tracking manually, tagging each lead with what the closed job was actually worth and feeding that number back to Google. Skip that step and Target ROAS is optimizing against fake or flat numbers, which is worse than not using it at all.

The three strategies, side by side.

Put the three questions next to each other and the decision gets a lot less confusing than the dashboard makes it look. Here's what each one actually needs to work, and who it's actually built for.

Read down that middle column and the pattern is obvious: every strategy past Maximize Conversions asks for something specific before it can do its job well, either a stable conversion count or accurate dollar values. An account that doesn't have what the strategy needs isn't being optimized when it's switched over. It's being handed a target it doesn't have the information to hit.

Only One Fits The LockThe Choice

Why low-volume accounts break automated bidding.

Smart Bidding works by finding patterns across recent conversion events and adjusting bids in real time based on them. That requires a certain density of signal. A plumber generating 20 leads a month, split across four or five campaigns and a dozen ad groups, might be handing the algorithm three or four conversions a month per ad group to learn from. That's not enough pattern for a machine learning system to aim precisely at a dollar target, and the result usually isn't gentle underperformance, it's erratic weeks: a cheap week followed by an expensive one, with the algorithm effectively guessing between the two.

Walk through the math and the problem gets concrete fast. Say an HVAC company is running three campaigns, each generating five to six conversions a month, with a Target CPA of $90 set on all three. Smart Bidding needs a meaningful sample of recent conversions to judge whether a given auction, at a given time of day, in a given zip code, is likely to produce a cheap lead or an expensive one. With five or six data points a month per campaign, the algorithm is essentially re-learning from scratch every few weeks instead of refining a stable pattern. Merge those three campaigns into one, pooling 16 to 18 conversions a month into a single bidding pool, and the algorithm suddenly has three times the signal to work from, often stabilizing the average cost per lead within a single billing cycle.

An algorithm chasing a target it doesn't have the data to hit isn't optimizing. It's guessing with extra confidence.

There's a related trap here worth naming directly, because it causes real confusion: this data problem is about how precisely the bidding algorithm can aim, not about how fast a campaign starts producing leads in the first place. A properly built Google Ads campaign can start putting calls on the phone within 24 to 48 hours of launch, regardless of which bidding strategy it's running. What takes longer is giving an automated bidding strategy enough conversion history to hit a specific cost or revenue target reliably. Those are two different clocks, and mixing them up is how a business owner ends up thinking their ads are "slow" when the real issue is a Target CPA set before the account had the data to support it.

Consolidating ad groups to pool conversion data into fewer, larger buckets helps the algorithm see more signal per decision. So does running Maximize Conversions with no target attached long enough to build a real history before asking the algorithm to hit a number. Switching strategies every few weeks makes this worse in a different way: every change resets the campaign into a fresh learning period, typically one to two weeks of unstable bidding while Google recalibrates, so an account that gets switched repeatedly is almost never out of that unstable window long enough to show what it can really do.

What to actually run, and when to switch.

Strip out the recommendations panel entirely and the decision path is short enough to run from memory.

1. Brand new campaign, little or no conversion history: run Maximize Conversions with no target attached. Let it collect data for at least four to six weeks, or until the campaign has logged somewhere around 30 or more conversions, before touching it again.
2. Steady conversion flow, roughly 15 to 30-plus a month, and a known breakeven cost: add a Target CPA set near your real historic average cost per conversion, not meaningfully below it. Setting the target too aggressively below what the account has actually been achieving is the single most common way Target CPA quietly throttles volume.
3. Accurate per-conversion dollar values flowing back to Google, not just lead counts: this is the only point where Target ROAS belongs in the conversation. If that value tracking doesn't exist yet, build it before the strategy, not after.
4. Any time you change strategy or meaningfully change a target: expect one to two unstable weeks while the campaign relearns, and resist judging or changing it again inside that window.

Most local service accounts we review are further along this path than their bidding strategy reflects, usually a Target CPA set months ago at a number nobody has revisited since, while the account's real conversion volume and average cost have both moved. The fix in that case isn't a new strategy, it's updating the target to match where the account actually is now, then leaving it alone through its next learning period before judging the result.

It's also worth building a habit around when you check in on a bidding strategy versus when you touch it. Checking weekly to understand the trend is healthy. Changing the target, the strategy, or the budget every time a single week looks off is what keeps an account stuck in permanent learning mode, never given long enough on a stable setting to show what it's actually capable of. The accounts that get the best long-term cost per lead out of automated bidding tend to be the ones where the owner or agency makes fewer changes, not more, and gives each one the full learning window before deciding whether it worked.

Settled On The Right CallThe Verdict

If you're running both Google and Meta and want the same discipline applied across both platforms, our guide to reading a Google Ads report covers how to tell whether a strategy is actually working once it's live, and our piece on why Google Ads aren't getting calls is the right next stop if volume is the real problem, not the bidding strategy. If the deeper issue is that you can't tell whether any of this is paying off, our guide to tracking real marketing ROI walks through building the exact value tracking that Target ROAS requires in the first place.

Frequently asked questions.

Why does my ad rep keep switching my bidding strategy without asking me?

Google Ads surfaces automated bidding switches inside its own Recommendations tab and scores an account higher for adopting them, whether or not the account has enough conversion history to run them well. An agency or in-house rep chasing that optimization score will sometimes apply the suggested change without reviewing your actual conversion volume first. The fix isn't to refuse every recommendation, it's to ask what your monthly conversion count is before any bidding strategy changes, since that number is what actually decides whether the switch will help or hurt.

Is Target CPA or Target ROAS better for a local service business?

For most independent local service businesses, Target CPA fits better, because it only needs a conversion count, not a conversion value. Target ROAS needs accurate dollar values passed back to Google for every lead, and most shops either don't track that or fake it with a flat placeholder value, which gives the algorithm bad data to optimize against. Target ROAS becomes the right call once you're genuinely tracking what a closed job is worth and feeding that number back automatically, not before.

How many conversions do I need before automated bidding actually works well?

There's no hard cutoff, but we typically see Target CPA and Target ROAS stabilize once a campaign is generating somewhere around 15 to 30 or more conversions a month. Below that, the algorithm doesn't have enough recent signal to aim precisely at a dollar target and tends to produce erratic, expensive weeks. Campaigns under that range almost always do better on Maximize Conversions with no target attached while they build up history.

Should I ever go back to manual bidding?

Rarely, and usually only for a very low-volume campaign, under roughly 5 to 10 conversions a month, where you want tight manual control over exactly what you bid on which keyword while the account is too small for any automated strategy to have enough data to work with. For nearly every other local service account, Maximize Conversions without a target is the safer default while conversion history builds, not a return to manual bidding.

Mitchell Wolfert Founder, Provision Digital. Running paid media, SEO, and funnels for local and national brands from Queen Creek, Arizona.