Choose target CPA when conversions represent sufficiently comparable outcomes; consider target ROAS when their values differ and those differences are measured reliably. Start with what the account records as success, rather than the bidding label. Automating a weak signal does not turn it into a profitability measure.
A job application and a sales opportunity do not have equal value simply because both complete a form. A high-revenue order does not necessarily produce more margin either. Before choosing cost per acquisition or return on ad spend, define the outcome you need and how it relates to the business economics.
What target CPA and target ROAS do differently
Google’s target CPA documentation describes seeking conversions at a target average cost. This does not guarantee identical acquisition costs or impose a cap on each transaction. Target ROAS uses conversion value to guide bids toward a target relationship between value and advertising spend.
| Question | Target CPA | Target ROAS |
|---|---|---|
| Which signal matters? | The selected conversion and its volume. | The value assigned to conversions. |
| What needs checking? | Commercially meaningful, comparable actions. | Values reflecting useful differences without duplication. |
| Interpretation risk | Cheap actions that sales cannot accept. | Mistaking attributed revenue for profit. |
| Business question afterwards | What does a valid opportunity or customer cost? | What contribution remains after costs and advertising? |
When CPA may make sense
It can help when a conversion represents a relatively similar unit of value, such as a qualified enquiry for one service. Definition matters. Combining guide downloads, WhatsApp clicks and signed contracts as though they were equivalent makes the average cost less useful for decisions.
Review the next step too. If two campaigns generate equal form counts but only one produces serviceable requests, their dashboard CPAs do not tell the whole story. Compare accepted enquiries and sales, and investigate how dependable commercial signals could be incorporated into measurement.
When ROAS may make sense
Evaluate it when meaningful value differences exist and tracking can represent them. For a store, check currency, tax and shipping against the chosen definition, plus duplicates, cancellations and refunds. A technically transmitted value can still be misleading if it differs from what the team believes it is optimising.
For leads, assigning the same arbitrary value to everyone adds no information. When estimating value by opportunity type, document its basis and update it as commercial conditions change. Do not describe an estimate as collected revenue. Our guide to what ROAS means helps interpret the metric before using it as a target.
Requirements are not one universal conversion threshold
Current target CPA documentation allows starting without conversion history; this does not make an account with no data immediately easy to evaluate. Target ROAS requirements depend on campaign type. Verify the relevant strategy conditions and eligibility before following a generic recommendation.
Separate availability, measurement readiness and enough evidence to assess outcomes. An account can meet the first condition while failing the other two. No single waiting period makes every sample representative; volume and the delay before a commercial result matter.
A hypothetical high ROAS with limited margin
Imagine an order generating COP 500,000 in revenue with COP 350,000 of variable costs before advertising. Contribution before media is COP 150,000. If acquiring the order costs COP 100,000, revenue ROAS is 5, yet contribution after advertising is COP 50,000 before other costs. These are educational assumptions, not market prices or client results.
If the business needs to retain COP 80,000 of that contribution, compatible advertising cost would be no more than COP 70,000 under these assumptions. The calculation defines an economic constraint; setting a bid target does not guarantee orders at that cost. A different product mix changes the relationship too.
Preparing a bidding strategy change
- Audit the conversion: primary action, duplication, value, currency and commercial quality.
- Define the economics: contribution, service capacity and sustainable acquisition cost.
- Review history: include delayed outcomes and record offer changes.
- Design a test: target, observation budget, period and evaluation criteria.
- Record and evaluate: avoid changing landing pages, targeting and bidding simultaneously.
Keep metric definitions consistent when comparing periods. If you change what counts as a conversion halfway through a test, a CPA movement may come from that definition. Explain that change before attributing it to campaign performance.
Errors to resolve before optimising
- Treating target CPA as a guaranteed maximum price per customer.
- Using revenue as a synonym for margin or collected cash.
- Assigning arbitrary values and never revisiting them.
- Judging a strategy before delayed conversions arrive.
- Choosing an aspirational target without checking business economics.
For ecommerce, the decision connects to the catalogue, prices and refunds. For services, it connects to qualification and sales follow-up. Our paid media management starts with those conditions before recommending a bid strategy.
Write a measurement contract before choosing bidding
A shared definition prevents debates about targets using numbers with different meanings. For every action, record what happened, which system confirms it, when it becomes valid and what value it receives. Identify who may change the definition. An unannounced rule change can make history incomparable even when the event name remains the same.
Distinguish an observation from an estimate. A paid order, qualified opportunity and expected closing value are different signals. An analysis may need all three, but should not present them as collected revenue. Label each according to its nature and explain how it relates to the account’s optimisation goal.
Validate the whole journey with a sample: source event, value, currency, date, receipt and commercial outcome. A configuration can accept a field without that field representing the intended economics. Validation needs both the implementation owner and someone who understands how the business recognises sales and costs.
Estimate lead values without inventing revenue
For opportunities with different values, begin with cohorts evaluated consistently. Ask what share became sales, what contribution they produced and how long that took. Document limited history or product changes. A precise spreadsheet cannot remove uncertainty from the evidence supplied to it.
In a hypothetical example, one opportunity category has an estimated 20% close probability and COP 600,000 contribution per completed sale before advertising. Expected contribution would be COP 120,000 per opportunity under those assumptions. This is an internal planning model, not an actual sale, a benchmark or a universal instruction for configuring Google Ads values.
Before using an estimate in the account, validate that the chosen method and definitions fit the measurement design. Define when it will be recalculated too. A changed qualification process or increased cancellations can make expected value obsolete. Do not retain old figures merely because they produce an attractive reported ROAS.
Examine the mix of products and services
An average can hide important differences. Two product lines with equal revenue may leave different contributions. Two lead types may require very different amounts of sales work. Review whether the aggregate objective makes those differences understandable or whether a family-level economic analysis is needed before allocating investment.
| Situation | Economic question | Prior check |
|---|---|---|
| Products with different margins | Does the observed value represent business priorities? | Compare sales mix and contribution |
| Services with different handling costs | Does a cheap lead require excessive follow-up work? | Record qualification and sales effort |
| Purchases with frequent cancellations | Does initially reported revenue remain? | Evaluate mature outcomes and refunds |
| Slow-closing opportunities | Does the period include their outcome? | Compare cohorts with equivalent follow-up |
Do not use the review to assign arbitrary priority to a product someone prefers selling. Relate decisions to margin, capacity, demand and actual conditions. Show additional strategic reasons, such as introducing a new line, separately so they are not confused with demonstrated profitability.
Test how sensitive the target is to assumptions
Repeat the economic calculation while changing one assumption at a time. What happens if delivery cost rises, closing rate falls or average order composition changes? A sustainable-looking target under one combination may be fragile under a small variation. Sensitivity analysis helps the business decide how much risk it can accept.
In the COP 500,000 order example, a refund or an omitted cost can reduce available contribution. Define what the chosen margin includes and excludes. Comparing ROAS between businesses with different revenue, tax or refund definitions can create a misleading impression of which campaign performs better.
Sensitivity analysis does not predict an auction. It establishes business conditions and assessment limits. If the economic target requires results the account has not demonstrated, design a proportionate test and acknowledge the gap. A bidding setting cannot create the necessary demand by itself.
Prepare a change log and a subsequent reading
Before changing strategy, preserve the previous target, included actions, values, budget, period and recent website changes. State what the team expects to learn and which conditions should remain stable during observation, except where a fault requires correction. This baseline lets the team interpret differences without reconstructing account history from memory.
Afterwards, compare the same outcome definition and show pending conversions where possible. If quality changes, investigate its source rather than declaring success from an average cost. If product mix changes, compare contribution. The report should answer the economic question that motivated the test.
Document rejected interpretations as well as the selected decision. For example, an apparent CPA improvement may have coincided with a broader form definition, or a ROAS increase may reflect a short promotion on higher-priced items. Recording that context stops later reports from treating a temporary or definitional movement as proof of a durable bidding advantage.
When to postpone the CPA versus ROAS decision
Postpone a definitive recommendation when genuine conversions cannot be separated from duplicates, reliable values are missing or sales staff interpret statuses inconsistently. This does not require stopping every advertisement. It means recognising that information must be repaired before the bidding choice can be assessed properly.
Agree a short plan with owners: correct the event, reconcile a sample, stabilise classification and revisit the decision. Check operational capacity too. More enquiries at an acceptable acquisition cost can be unhelpful when nobody can handle them. Bidding belongs within an acquisition process the business can sustain and evaluate.
Keep the next decision explicit. A repaired event may justify a new measurement period rather than an immediate target change. A stable qualification rule may justify comparing accepted opportunities before attempting value-based optimisation. The appropriate next step depends on the missing evidence, not on a requirement to use whichever strategy sounds more sophisticated.
Version estimated lead values separately from observed revenue. Record the model date, assumptions and intended use, and keep earlier versions available for comparison. Otherwise a later value adjustment can make reported performance appear to change even when the underlying commercial outcomes have not changed.
Your first Google Ads audit is free. Let’s review what limits your account and what to improve first. Request my free audit.
Frequently asked questions
Does target CPA cap the cost of each conversion?
No. It is an average cost target; individual conversions can cost more or less. It is not automatically cost per customer when the measured action is a form.
Does a high ROAS mean a campaign is profitable?
Not by itself. Variable costs, refunds, margins and other costs matter. Attributed revenue is not profit.
Do all strategies require the same conversion count?
No universal minimum applies to every strategy and campaign. Check current requirements for the campaign type and the quality of its measurement.
Is the first SEOMOS bidding review free?
Yes. Your first Google Ads audit is free and can review measurement, context and priorities. Further implementation is agreed according to scope.