Bidding Strategies Explained: Manual, Target CPA, and Maximize Conversions

Switch a $3,000/month account from manual CPC to Target CPA on a Tuesday and by Friday you’ll either be thrilled or furious, because automated bidding doesn’t gently nudge your results — it commits hard to whatever pattern the algorithm thinks will hit your target, and if the account doesn’t have the conversion volume to support that confidence, spend gets erratic fast. Most advertisers pick a bidding strategy based on what a platform rep recommended or what a blog post from three years ago said, without checking whether their account has the one thing every automated strategy actually needs: enough conversion data to learn from.

What Each Strategy Is Actually Optimizing For

Manual CPC (or manual bidding generally, across Google, Microsoft, and Meta’s legacy options) puts you in charge of the bid on every keyword or placement, and the algorithm’s only job is auction mechanics — it doesn’t touch your bid based on predicted conversion likelihood. You’re trading algorithmic leverage for total control, which is valuable when you understand your account’s patterns better than a machine learning model with thirty days of data does.

Target CPA tells the platform “get me conversions at roughly this cost, and vary bids per auction to make that average work.” It raises bids for auctions it predicts will convert and suppresses them for auctions it predicts won’t, using signals — device, time of day, audience overlap, query intent — that you can’t see or manually replicate at that granularity. The tradeoff is that it needs a real sample size to make those predictions confidently; Google’s own guidance suggests at least 15-30 conversions in the last 30 days per campaign before Target CPA has enough to work with, and accounts below that threshold frequently see wild cost swings as the algorithm essentially guesses.

Maximize Conversions drops the cost constraint entirely and just spends your full daily budget trying to get as many conversions as mathematically possible. It’s the most aggressive of the three and the easiest to misuse, because “as many conversions as possible” says nothing about whether those conversions are profitable. An account on Maximize Conversions with no value data attached will happily fill its budget with the cheapest, lowest-intent conversions available, because cheap and plentiful is literally the optimization target.

The Conversion Volume Threshold That Decides Everything

This is the single biggest predictor of which strategy will work, and it’s almost never discussed as clearly as it should be. An account generating 5 conversions a month has no business on Target CPA — there isn’t enough signal for the algorithm to learn a pattern, and you’ll see the classic symptom: costs spike for a week, crash the next, spike again, with no discernible trend, because the system is essentially re-guessing every cycle.

The rule of thumb that holds up in practice: under roughly 15 conversions/month, run manual bidding and optimize by hand using search term reports and bid adjustments. Between 15-30, you can test Target CPA but expect a rocky 2-3 week learning period and don’t panic at the first bad week. Above 30-50 conversions a month, automated bidding usually outperforms manual within a month, because the algorithm is working with more signal per day than a human optimizer could process by hand across hundreds of auctions.

If your account is below threshold and you need automation anyway, the fix isn’t to force Target CPA — it’s to broaden the conversion action you’re bidding to. Bidding to “purchase” with 8 conversions/month performs worse than bidding to “add to cart” with 60/month, even though add-to-cart is a weaker signal, because volume beats precision when the algorithm needs data to function at all. You can layer purchase-value optimization back in later once volume grows.

A Worked Example: The Same $4,000 Budget, Three Ways

Take a hypothetical B2B software account spending $4,000/month, generating 22 demo-request conversions a month at a $180 average CPA. On manual CPC, a hands-on manager might spend 4-5 hours a week adjusting bids by keyword, pulling search term reports, and pausing underperformers — realistically holding CPA flat around $180 with maybe a 5-10% improvement over a quarter through careful keyword-level pruning, but at a real time cost.

Move that same account to Target CPA set at $190 (5% above trailing average, per the “walk it down” approach below), and the typical pattern over the first three weeks looks like this: week one, CPA swings between $140 and $260 as the algorithm samples different auction types; week two, the range tightens to $160-$210; week three, it stabilizes around $175-$185 with volume up 10-15% because the algorithm is finding conversions across device and time-of-day combinations a manual manager wasn’t bidding aggressively on. That’s the outcome when the 15-30 conversion threshold is respected.

Now take the same $4,000 and same 22 conversions but put it on Maximize Conversions with no value data and no conversion-action segmentation. Volume jumps — often 25-40% more conversions — but a chunk of that increase is newsletter signups and content downloads counted in the same conversion action as demo requests, because the campaign wasn’t segmented to only fire the bidding signal off high-intent actions. Blended CPA looks great on a dashboard; pipeline-qualified leads from that traffic often barely move, because the algorithm delivered exactly what it was told to optimize for — more conversions, not better ones. The strategy wasn’t wrong; the setup handed it the wrong optimization target.

Manual Bidding Still Wins in Three Specific Situations

Brand-new accounts with zero conversion history should start manual, full stop — there’s nothing for an algorithm to learn from yet, and letting Maximize Conversions run wild on a cold account with a $50/day budget is a fast way to burn the entire budget on the first three auctions the algorithm happens to see.

Accounts with highly seasonal or event-driven demand (a tax-prep service in March, a fireworks retailer in June) also do better manual, because automated strategies use trailing data to predict forward performance, and trailing data from the off-season is actively misleading during a demand spike. Bumping bids manually 2 weeks ahead of a known seasonal surge outperforms waiting for the algorithm to notice the pattern is already underway.

And any account running a genuinely small, tightly curated keyword list — under 20 keywords, each hand-picked for specific commercial intent — often does fine on manual because the entire point of manual in that context is that a human has already done the targeting precision work the algorithm would otherwise be approximating.

Target CPA Needs a Realistic Target, Not an Aspirational One

The most common Target CPA failure isn’t the strategy itself, it’s setting the target 30-40% below what the account has historically achieved, hoping the algorithm will just find cheaper conversions. It won’t — a CPA target set meaningfully below the account’s actual average cost per conversion typically causes the algorithm to under-spend the daily budget, because it can’t find enough auctions it’s confident will hit that low a cost, and impression volume collapses rather than cost per conversion improving.

The better approach: set your initial target at or slightly above (5-10%) your trailing 30-day average CPA, let it run for two full weeks untouched, then step the target down in small increments — 5% at a time — every 1-2 weeks if performance holds. This “walk it down” approach gets you to an aggressive target eventually without triggering the under-delivery problem that comes from asking for too much too fast.

Maximize Conversions Needs Guardrails You Set Yourself

Because Maximize Conversions has no built-in cost ceiling, the guardrail has to come from campaign structure. Segment campaigns so that Maximize Conversions only runs against conversion actions you’re genuinely happy to pay any reasonable amount for — a demo request or a completed purchase, not a newsletter signup or a PDF download, which are cheap enough that Maximize Conversions will happily hoover up budget on them at scale while contributing little pipeline value.

If your platform supports Maximize Conversions with a target ROAS overlay (value-based bidding), turn it on as soon as you have conversion value data flowing — even directionally accurate values from a rough LTV estimate by product line beat no value data at all, because it gives the algorithm a quality signal instead of a pure volume signal. An account bidding purely on conversion count with no value data attached is, structurally, optimizing for the wrong thing even when it “performs well” by the metrics it can see.

Common Failure Modes That Aren’t About the Strategy Itself

Most of the “automated bidding doesn’t work for us” complaints trace back to setup mistakes rather than the algorithm being wrong for the account. The most frequent one: conversion tracking counting duplicate or low-value events as the primary conversion action — a “form start” firing alongside “form submit,” or a thank-you-page view firing on page refresh, both inflating conversion counts artificially. This looks like plenty of volume to the algorithm, clears the 15-30 threshold on paper, but is training the bidding model on noisy or duplicated signal, and performance never quite tightens up no matter how long you wait.

The second common mistake is changing the bid strategy and the campaign structure (ad groups, keywords, budgets) at the same time. When performance shifts after a change, you can’t tell whether it was the bidding strategy or the structural change that caused it, and teams often blame the wrong variable — killing a bidding strategy that would have worked fine if it had been tested in isolation. Change one variable at a time and give each change a full learning cycle before drawing conclusions.

The third mistake is applying portfolio-level bid strategies (shared across multiple campaigns) too early, before any individual campaign has proven its own conversion pattern. Portfolio strategies pool conversion data across campaigns, which can mask the fact that one campaign in the portfolio has plenty of volume and another has almost none — the algorithm ends up making decisions on the low-volume campaign based on data that actually belongs to a completely different audience and funnel stage. Let campaigns prove themselves individually before pooling them.

Reading the Signals That Tell You It’s Time to Switch

Three signals suggest a manual account is ready to graduate to automated bidding: conversion volume has crossed the 15-30/month threshold for three consecutive months (not just one good month), the account has a stable, non-seasonal demand pattern, and you have at least one full quarter of consistent conversion tracking with no major tagging changes in that window — automated bidding trained on inconsistent tracking data learns the wrong lessons just as confidently as it learns the right ones.

Conversely, the signal that it’s time to step back from automation isn’t just “performance got worse” — automated strategies have natural week-to-week variance and punishing them for a single bad week is how advertisers whipsaw between strategies without ever giving either one a fair test. The real signal is a sustained divergence — 3+ weeks of costs 25%+ above target with no corresponding change in market conditions (increased competition, a landing page change, a tracking break) that would explain it externally.

Running a Fair Test Between Strategies

If you genuinely don’t know which approach fits your account, the only reliable way to find out is a controlled test, not intuition. Split near-identical ad groups or campaigns (same keywords, same budgets, same landing pages) across manual and automated bidding for a minimum of 3-4 weeks — shorter than that and you’re just measuring the automated strategy’s learning phase, not its steady-state performance.

Track cost per conversion, but also track conversion volume and impression share alongside it, because a strategy that hits a lower CPA by simply buying fewer, more certain conversions isn’t actually more efficient — it’s just more conservative, and conservative isn’t always what the business needs if growth is the priority. The account that wins on a blended view of cost, volume, and impression share earned, held over a full month, is the one worth committing budget to going forward — not the one that looked best in the first excitable week after the switch.

How to Prioritize This If You’re Managing Multiple Accounts

Managing several accounts at once, apply the conversion threshold check first, before anything else — it takes five minutes per account (pull trailing 30-day conversions from the platform’s own reporting) and immediately tells you which accounts are candidates for automation and which aren’t, regardless of what strategy each is currently running. Accounts already misassigned relative to their volume (a 6-conversion/month account stuck on Target CPA, or a 60-conversion/month account still on manual) are your highest-leverage fixes and should be addressed before any creative, keyword, or landing page optimization work, because a mismatched bidding strategy caps the ceiling on everything else you do in the account.

After the threshold check, prioritize fixing conversion-tracking hygiene (duplicate events, missing value data) before touching the bid strategy itself, since a strategy change layered on top of broken tracking just produces a second confusing variable to untangle later. Only once volume is confirmed adequate and tracking is clean does it make sense to run the controlled test described above — testing a strategy change against unreliable tracking data wastes the 3-4 week test window on a result you can’t trust anyway.