Google Ads bidding strategies: tCPA, tROAS or Maximize Conversions?

The question we hear most about Google Ads bidding strategies is still the same: "Should we move to Target CPA, Target ROAS, or stay on Maximize Conversions?" The answer doesn't depend on how big your account is, it depends on the volume of your conversion data and the quality of your value tracking. This guide walks through when each strategy makes sense, how to manage the learning period during a switch, and how to adjust targets with concrete thresholds.

What is Smart Bidding and how does it decide?

Smart Bidding is Google Ads' auction-time bidding system, trained on your conversion data. For every single auction it weighs signals you could never manage by hand, device, location, time of day, audience lists, query intent, browser and more. The system is only as good as the conversion data feeding it: if your tracking is incomplete or wrong, no strategy will bid correctly.

• Maximize Conversions / Maximize Conversion Value, spends the budget to get as many conversions (or as much value) as possible, with no target.
• Target CPA (tCPA), tries to keep your average cost per conversion around the target you set.
• Target ROAS (tROAS), predicts the likely value of each search and tunes bids to hit your return target.
• Manual / Enhanced CPC, defensible today only in very low-volume or highly specific scenarios.

Which bidding strategy should you use when?

Two questions drive the choice: do you have enough conversion volume, and are your conversions equal in value? The flow below shows where to start at campaign level.

• Maximize Conversions (no target), new campaigns or fewer than 30 conversions a month: accumulate data first, then move to a targeted strategy.
• Target CPA, forms, sign-ups, demo requests and other roughly equal-value conversions: you fix the cost and scale the volume.
• Target ROAS, e-commerce where order values range from $200 to $20,000: the system bids high on likely high-value buyers and low on the rest.
• Maximize Conversion Value (no target), you track values but lack the volume for tROAS: use it as the intermediate step.

Target CPA: lock in your cost

Target CPA is for scenarios where you can say "a conversion should cost me at most this much." On individual clicks the system will bid above and below the target; the goal is for the average to land on it. The most important rule: start the target from your actual 30-day CPA, not the CPA you wish you had. If your real CPA is $260 and you enter a $150 target, the system chases an unreachable goal, throttles impressions and volume collapses. And when you evaluate results, look at a window of at least 30 conversions, a verdict based on five conversions is statistically meaningless.

Target ROAS: bid on value

Target ROAS comes into play when conversion values vary, and it requires conversion value tracking: basket values in e-commerce, segment-based proxy values in lead gen. On Search campaigns the technical floor is 15 valued conversions in the last 30 days; in practice 50+ is recommended for stability. If you're below that, don't force tROAS, either consolidate campaigns into a shared portfolio strategy or start with untargeted Maximize Conversion Value. The next level of maturity is targeting profit rather than revenue: if margins differ across product lines, consider thinking in profit terms instead of raw ROAS.

The learning period: what to expect when you switch

When you change strategy or target, the system recalibrates. The official duration is about 7 days; in practice, plan for 7-14 days including conversion lag, and longer on low-volume accounts. During this window it is normal for CPA to run temporarily above target, the illustrative path below shows the typical pattern: volatility first, then stabilization below the old average.

1. Before switching, note your actual 30-day CPA/ROAS average, that's your baseline.
2. Set the initial target at that average (or at most 10-15% more ambitious).
3. Batch your other changes (ad copy, keywords, budget) into the same day so you trigger one learning period, not three.
4. Leave the campaign alone for 7-14 days; watch the daily swings but don't react to them.
5. After stabilization, adjust the target in ±10-15% steps, at most once a week.

How do you adjust targets based on data?

The most common mistake is yanking the target to where it "should" be in one move. An aggressive tROAS target forces the system into only the safest auctions: efficiency looks better on paper while volume suffocates. In the illustrative comparison below, the campaign on an aggressive target slides from 38 to 33 weekly conversions over four weeks, while an identical campaign on a realistic target scales from 41 to 56.

The right rhythm: start the target at your actual average, wait for stabilization, then take one small step per week in a single direction. Accounts that walk gradually toward their efficiency goal avoid both learning resets and volume collapses.

The 5 most common mistakes

• Switching to a targeted strategy with too little data, running tROAS on 10 conversions a month is leaving the outcome to chance.
• Reverting because of learning-period volatility, changing strategy on day 3 because "CPA went up" keeps you in the reset loop forever.
• Setting targets from wishes instead of data, an unreachable target ends with the system throttling impressions.
• Making small tweaks every day, frequent touches mean the system never stabilizes.
• Ignoring conversion tracking issues, with double-counting or missing conversions, your strategy choice stops mattering.

Make bid strategy decisions with data

The hard part of these decisions isn't knowledge, it's follow-through: tracking which campaign has crossed the threshold, which target is choking volume, and whether the change actually worked. Ads Sensor automates that: it unifies your Meta Ads, Google Ads and GA4 data in one panel, and its AI analyzes each campaign's conversion volume and target-vs-actual gap to produce prioritized, reasoned actions. Recommendations you approve are applied through the platform APIs, and before/after results are tracked automatically, so "did moving to tROAS pay off?" gets answered with data, not gut feeling. 24/7 anomaly monitoring also helps you tell learning-period noise from a real problem. Request pre-beta access and be among the first to run the analysis at launch.