Value-based bidding on Google Ads comes in several distinct forms, and which one you use determines what the algorithm actually learns to do with your budget. The most common confusion in DTC accounts is between Target CPA and Target ROAS — two strategies that sound similar but optimise for fundamentally different things. Target CPA asks Google to find conversions at a fixed cost. Target ROAS asks Google to find revenue at a target return. For any brand with varied order values, that distinction produces meaningfully different customer acquisition outcomes.
The gap becomes clearest with a straightforward example. A wellness brand sells a $45 single supplement and a $119 bundle. Under Target CPA set at $30, both conversions receive identical treatment — both happened, both met the cost target, both get equal weight in the algorithm's learning model. Under Target ROAS at 300%, the $119 bundle purchase is weighted more than twice as heavily as the $45 order. The algorithm learns to find users who spend more, because you told it that revenue is the objective.
This guide covers how value-based bidding on Google Ads works across all three relevant strategies, the precise conditions under which Target CPA is the right choice versus Target ROAS, and when the correct move is to upgrade beyond both toward predictive lifetime value signals. The decision framework is a function of your conversion volume, your product pricing structure, and your data infrastructure — not campaign type.
In This Article
- 1The Core Difference: Cost vs Revenue Optimisation
- 2When Target CPA Is the Right Choice
- 3Why Target CPA Underperforms for Most DTC Brands
- 4Target ROAS: The Value-Based Bidding Default for Ecommerce
- 5The Three-Strategy Comparison: tCPA vs tROAS vs VBB+pLTV
- 6How to Make the Switch: The Correct Sequence
- 7Common Mistakes When Choosing Between Strategies
- 8The Next Layer: Moving From tROAS to pLTV Signals
- 9Further Reading
The Core Difference: Cost vs Revenue Optimisation
Target CPA and Target ROAS share the same underlying Smart Bidding infrastructure but optimise toward different objectives at auction time. When you set a Target CPA, you are telling the algorithm: find as many conversions as possible while keeping the average cost per conversion at or below this number. The algorithm counts conversion events and adjusts bids to hit the cost target. What the conversion was worth in revenue is irrelevant to the optimisation.
Target ROAS, as part of the value-based bidding framework, assigns revenue to each conversion event. The algorithm's objective changes from finding the cheapest conversions to finding conversions that produce the most revenue per dollar spent. A campaign running Target ROAS at 400% will pass on a low-value conversion if winning that impression would push overall ROAS below the target — behaviour that Target CPA would never exhibit, because it has no visibility into what each conversion generates in revenue.
The practical consequence is that Target CPA creates a hidden efficiency ceiling for any business where conversion values vary. Every time your supplement brand acquires a $119 bundle customer at $30 CAC, the algorithm records it as a $30 conversion — identical to a $30 CAC acquisition of a $45 single product. Two different revenue outcomes, one signal. The model learns from both equally, and over time allocates budget toward the profile that converts most cheaply, which is not necessarily the profile that generates the most revenue.
When Target CPA Is the Right Choice
Target CPA is the correct strategy for three specific scenarios. The first is accounts below 30 monthly conversions. At this volume, value-based bidding strategies struggle to identify reliable patterns — the algorithm needs sufficient data to distinguish high-value from low-value conversion profiles, and below this threshold it is essentially guessing. Start with Maximize Conversions to build data density, then transition to Target CPA once the account is producing 30 or more conversions per month with consistent tracking.
The second scenario is fixed-price or single-SKU products where every conversion genuinely generates the same revenue. A wellness brand selling one $49 collagen powder subscription, where 100% of buyers pay the same price, has no variation in order value to exploit. Target CPA and Target ROAS would produce equivalent outcomes here because the revenue signal and the conversion signal carry the same information. In this case, Target CPA is simpler to set up and monitor.
The third scenario is lead generation campaigns where the conversion event is a form fill, a trial sign-up, or a phone call — not a revenue-generating transaction. When no purchase value attaches to the conversion event, there is nothing for value-based bidding to optimise against. Target CPA is the correct default for these campaign types. The issue arises when DTC brands continue using Target CPA on revenue-generating campaigns long after their product mix and conversion volume justify switching.
One exception worth noting
Target CPA is not available for Standard Shopping campaigns. If you are running Shopping and have not moved to Performance Max, your only automated value-based option is Target ROAS. This is one reason Google has been moving advertisers away from Standard Shopping — Performance Max supports the full range of Smart Bidding strategies including value-based variants.
Why Target CPA Underperforms for Most DTC Brands
A legal services business running Target CPA at $70 can maintain consistent acquisition costs for 18 months because every lead is worth approximately the same amount in downstream revenue. An ecommerce brand selling supplements at $45, $67, and $119 runs the identical strategy and gets something different: Google gets better and better at finding cheap conversions, which means more $45 single-product orders and fewer $119 bundles, because the single-product customer converts at lower CPCs.
Research across 312 Google Ads accounts found that ecommerce accounts with varied order values consistently saw Target CPA optimise toward low-value purchases. The algorithm performs exactly as designed — it finds the most conversions at the cheapest cost. The problem is that "cheapest conversion" and "most valuable conversion" are not the same thing for any DTC brand with a product catalogue. Switching from Target CPA to Target ROAS in these accounts produced a median 14% increase in conversion value at similar ROAS levels, according to Google's internal data.
The signal mismatch compounds over time. The longer Target CPA runs on an account with varied order values, the more the algorithm's internal model of "what a good conversion looks like" skews toward lower-value purchasers. When you eventually switch to a value-based strategy, the learning period is longer because the algorithm is re-learning from scratch which audience profiles correlate with revenue, not just with conversion events.
Target ROAS: The Value-Based Bidding Default for Ecommerce
Target ROAS is the entry point for value-based bidding on Google Ads for any DTC account with varied order values and sufficient conversion volume. You set a target return — say, 350% — and the algorithm adjusts bids in every auction based on its prediction of what revenue a given impression is likely to produce. High-revenue-associated user profiles receive higher bids. Lower-revenue profiles receive lower bids or are passed on entirely if winning them would pull ROAS below the target.
The data minimum before switching is 50 conversions per month with purchase values attached to each conversion event. Adalysis research on 16,825 campaigns found that accounts typically make this transition once they exceed approximately 70 conversions and $9,000 in monthly revenue — thresholds at which the algorithm has sufficient value data to make reliable predictions. Below these levels, Maximize Conversion Value without a ROAS target is the correct intermediate step: it builds the value-signal history that Target ROAS needs to function well.
Setting the initial tROAS target correctly matters as much as making the switch. Google's official recommendation is to start 20% below your historical ROAS average to give the algorithm room to learn. An account that has achieved 380% ROAS under Maximize Conversion Value should launch Target ROAS at roughly 305%. Setting the target at or above 380% immediately restricts the auction pool the algorithm can enter, which reduces conversion volume and slows the learning cycle — often causing advertisers to abandon the strategy before it stabilises.
The Three-Strategy Comparison: tCPA vs tROAS vs VBB+pLTV
The practical difference between these three strategies comes down to what each one sees when a conversion fires, and what signal it uses to make future bidding decisions.
A data-driven way to think about this: Target CPA is appropriate when every conversion is binary (it happened or it did not) and worth the same amount. Target ROAS is appropriate when conversion values vary but you are optimising against first-order revenue. VBB with pLTV signals is appropriate when you need the algorithm to optimise against the revenue a customer will generate over 12 months, not the revenue they generated on their first transaction. The distinction between the second and third categories is where most DTC growth gets left on the table.
How to Make the Switch: The Correct Sequence
The bidding progression is not arbitrary — each stage requires the data generated by the previous one. Moving to Target ROAS before 50 monthly conversions with value data means the algorithm is making value predictions with insufficient signal. Moving to pLTV signals before tROAS is running stably means the platform has no baseline to calibrate against. The sequence exists for reasons the algorithm's performance data consistently validates.
Maximize Conversions (no target) — weeks 1 to 6
For any new campaign or any account below 30 conversions per month. No cost constraints. The algorithm spends your budget to find conversions and accumulate data. Do not set a Target CPA during this phase — doing so restricts the auction pool before the algorithm has data to make good decisions. Evaluate CPA during this phase only for trend observation, not for strategy changes.
Verify value tracking is accurate — before moving to tROAS
Confirm that every purchase event fires with the actual transaction value in the conversion value field. Check Google Ads diagnostics: the Conversion Value column should show non-zero, varied amounts — not a uniform number. If all conversions show the same value, the tracking is passing a static figure, not dynamic purchase data. Fix this before switching to any value-based strategy.
Maximize Conversion Value (no tROAS) — weeks 6 to 12
Once 30+ conversions per month with dynamic values are confirmed, switch to Maximize Conversion Value with no ROAS target. This begins value-based optimisation without the constraint of a ROAS floor. Let this run for 4 to 6 weeks. The algorithm is now learning which user profiles correlate with higher revenue — the foundation tROAS needs.
Introduce Target ROAS at 80% of historical ROAS — month 3 onward
Calculate your ROAS from the Maximize Conversion Value phase. Multiply by 0.80. That is your launch tROAS. If Max CV produced 420% ROAS, your starting tROAS is 336%. Do not change bids, budgets, or creative for 6 weeks after this switch. Evaluate performance against revenue and LTV:CAC, not CPA — which will likely increase as the algorithm shifts toward higher-value, more competitive auctions.
Upgrade to pLTV signals once tROAS is stable — month 4 to 6
When tROAS has been stable for two or more consecutive learning cycles, you have the infrastructure to pass predictive lifetime value as the conversion signal. Instead of passing checkout AOV, the signal is a model-generated 12-month or 24-month revenue prediction. The algorithm now optimises toward profiles that will still be buying from you next year — the transition from first-order VBB to full value-based acquisition.
Common Mistakes When Choosing Between Strategies
Running Target CPA on ecommerce with varied AOV
The most widespread error in DTC Google Ads accounts. Supplement brands with $45, $67, and $119 SKUs, skincare brands with single-product and regimen bundles — any catalog with price variation will see Target CPA systematically optimise toward lower-value purchasers over time. The algorithm is performing exactly as instructed. The instruction is the problem.
Setting tROAS immediately without Maximize CV history
Jumping from Target CPA directly to Target ROAS without a Maximize Conversion Value phase means the algorithm has no prior value data to calibrate predictions against. The result is extended learning instability, unpredictable spend distribution, and ROAS that swings widely before settling. Always build 4 to 6 weeks of Maximize CV data first.
Setting the initial tROAS target above historical performance
An initial tROAS above your Maximize Conversion Value baseline restricts the auction pool immediately. The algorithm becomes too selective, impression share drops, and conversion volume falls — which shows up in reports as the strategy underperforming. The fix is starting 20% below historical ROAS, not at or above it.
Evaluating performance during the learning period
Making strategy decisions based on the first 3 to 4 weeks of a new bidding setup produces misleading conclusions. Set a review date for week 7 and observe only. Track conversion value trend and LTV:CAC, not CPA or ROAS against the first-week baseline — both will fluctuate significantly during learning.
Using Target ROAS without verifying dynamic value tracking
Target ROAS running on static conversion values (every order worth $1 or a fixed amount) optimises as if all purchases are identical — which is identical to Target CPA but with worse reporting clarity. Verify that your conversion event passes a different value for each transaction before enabling any value-based strategy.
The Next Layer: Moving From tROAS to pLTV Signals
Target ROAS optimises against the revenue a customer generates on their first transaction. For DTC brands in repeat-purchase categories — supplements, skincare, wellness — this is the second-best available signal. The best signal is predicted 12-month or 24-month customer revenue, calculated at the moment of first conversion and passed to Google as the conversion value through pLTV bidding on Google Ads.
The mechanism is the same as tROAS: a conversion fires, a value is passed, the algorithm learns which profiles produce high-value conversions. What changes is the quality of what "value" means. A $119 supplement bundle that generates a single $119 purchase passes $119 to the algorithm. The same customer whose early behavioral signals predict $680 in 12-month revenue passes $680 via AdZeta's ValueBid™ framework — and the algorithm bids accordingly, acquiring more customers who look like the $680 profile rather than the $119-and-done profile.
Brands consistently using the ValueBid™ platform see 20-30% CAC reduction within the first 60 days and a 15% average ROAS lift from the signal upgrade alone. The prerequisite is a stable tROAS setup with clean dynamic value tracking — exactly the infrastructure this article describes building. The transition from tROAS to pLTV signals is a signal change, not a campaign rebuild. The Beyond ROAS whitepaper covers the full framework for making this transition with vertical benchmarks and implementation timelines.
Key Takeaways
- Target CPA and Target ROAS solve different problems. Target CPA optimises for conversion count at a fixed cost. Target ROAS optimises for revenue at a target return. For DTC brands with varied order values, only Target ROAS can distinguish a $45 order from a $119 one.
- Target CPA is the right choice for three scenarios: below 30 monthly conversions, fixed-price single-SKU products, and lead generation campaigns where no purchase value attaches to the conversion event. Outside these scenarios, it is usually the wrong default.
- Research across accounts confirms the structural problem: ecommerce brands running Target CPA with varied order values see the algorithm optimise toward lower-value purchases over time because those conversions are cheaper to acquire. Switching to tROAS produces a median 14% conversion value lift according to Google internal data.
- The correct sequence is Maximize Conversions → Maximize Conversion Value → Target ROAS (set 20% below historical) → pLTV signals. Skipping stages produces extended instability. Building each stage on the data from the previous one is what lets the algorithm learn accurately.
- The upgrade from tROAS to pLTV signals is where the 14% improvement becomes 20-30% CAC reduction. Instead of optimising against first-order checkout revenue, the algorithm learns which user profiles generate high 12-month revenue — a signal only available through predictive modelling, not through standard transaction tracking.
Further Reading
What Is Value-Based Bidding? A Complete Guide for DTC Brands — how the VBB framework works, the role of conversion values, and what to pass to Google and Meta to activate smart optimisation.
pLTV Bidding on Google Ads: Step-by-Step Guide — the full implementation path for upgrading from tROAS-based VBB to predictive lifetime value signals via OCI and Customer Match.
Beyond ROAS: Predictive LTV for DTC Profitability — AdZeta's whitepaper on the LTV:CAC framework and how to rebuild acquisition strategy around long-term customer value rather than campaign-level return metrics.