Choose the objective before the strategy
A bidding strategy is an instruction about what Google should optimise in each auction. It cannot repair the wrong conversion action, fictional values or an unprofitable offer. Decide first whether the business needs conversion volume, conversion value, traffic or visibility.
Google’s current bid-strategy guidance maps Maximise Conversions and Target CPA to conversion volume, and Maximise Conversion Value and Target ROAS to value. In June 2026 Google began simplifying some labels: “Maximise conversions with a Target CPA” may appear as “Target CPA”, and the equivalent value strategy as “Target ROAS”. Google states that the underlying bidding behaviour is unchanged.
| Business requirement | Candidate strategy | Evidence required |
|---|---|---|
| Most orders or leads within a fixed budget | Maximise Conversions | Reliable primary actions of broadly comparable value |
| Volume around an allowable acquisition cost | Target CPA | Stable, representative CPA and sufficient headroom |
| Most revenue or weighted lead value within budget | Maximise Conversion Value | Accurate, timely values with meaningful variation |
| Value around a required return | Target ROAS | Values aligned with commercial value, not vanity events |
| Website visits for a non-conversion objective | Maximise Clicks | CPC guardrails and a reason traffic itself has value |
| Visibility in defined Search auctions | Target Impression Share | A deliberate visibility goal and maximum CPC control |
The decision differs by business model
Ecommerce
If every purchase passes accurate revenue but margins vary, revenue-based ROAS can still favour low-margin products. A hypothetical account sells Product A for £100 with £60 contribution before ads and Product B for £100 with £20 contribution. Both report £100 conversion value, so Target ROAS treats them equally. Passing margin-adjusted value, separating economic groups or using value rules where appropriate may produce a better instruction.
Use Maximise Conversion Value when the immediate goal is to use the available budget for the highest reported value. Add a ROAS target when finance has defined a return constraint and the campaign has evidence it can operate near that level. A target far above recent performance usually trades volume for selectivity; it does not force the market to become more profitable.
Lead generation
Maximise Conversions only works as well as the event called a conversion. If the primary event is any form submission, bidding will seek forms, including low-quality ones. Import qualified leads or won revenue through the CRM when possible. If lead values differ materially, value-based bidding can be more faithful than Target CPA.
Suppose Campaign A produces 20 leads at £50, of which two become £2,000-contribution customers. Campaign B produces 10 leads at £80, of which four become equivalent customers. Lead CPA favours A; customer economics favour B. The correct bidding input is the downstream outcome, not the cheapest top-of-funnel event.
Long sales cycles and low volume
There is no universal “30 conversions” switch that makes automation reliable. Google no longer presents that as a general requirement for these strategies. Frequency, delay, variance, campaign similarity and signal quality all matter. With sparse outcomes, pool genuinely comparable campaigns in a portfolio strategy, use earlier qualified stages cautiously, or retain simpler bid controls while improving measurement.
Targets are constraints, not ambitions
Set an initial target from recent representative performance and unit economics. If a campaign achieved a hypothetical 320% ROAS over a stable period, setting 600% because management wants more profit is not a plan; it is a stricter auction constraint. Volume may fall while the system searches for rarer opportunities.
For Target CPA, calculate the allowable CPA from contribution and close rate. For Target ROAS, calculate:
Required ROAS = conversion value reported to Google ÷ allowable advertising cost.
If an £80 order contributes £32 before ads and the business requires £12 contribution after ads, allowable ad cost is £20 and required revenue ROAS is 400%. If repeat purchase is included, document the horizon and discount; do not quietly mix projected lifetime value with first-order revenue.
How to change bidding without losing the diagnosis
- Validate goals: list campaign-specific primary actions and verify values, duplicates and conversion lag.
- Choose a clean baseline: exclude abnormal promotions, outages or tracking failures from the comparison.
- Make one strategic change: avoid changing bidding, budgets, creative and landing pages together.
- Allow for lag: wait until most conversions attributable to the test period have matured.
- Judge business metrics: include qualified lead rate, contribution or new-customer value, not only the strategy’s headline metric.
Where eligible, use a custom experiment. Google describes experiments as a traffic or budget split between a base and trial, making them suitable for testing Smart Bidding changes. Its Experiments documentation recommends allowing enough time for a result rather than reacting to early noise.
Diagnose common failures by symptom
- Budget spends but quality falls: inspect the primary goal and query/placement mix before tightening a target.
- Spend collapses after adding a target: compare the target with realised performance and auction opportunity; the constraint may be unrealistic.
- Reported ROAS rises but profit does not: separate brand, returning customers, product margin and refunds.
- Strategy remains limited or learning repeatedly: review frequent setting changes, budget restrictions, conversion delay and data sparsity.
- Lead volume grows but sales do not: import a deeper qualified or won stage and audit spam handling.
Google’s guidance on changing conversion goals warns that bidding needs time to adjust when the optimisation input changes. Fixing a bad goal may be necessary, but it should be planned and annotated.
Where Manual CPC and Maximise Clicks fit
Traffic-based bidding can be useful when the immediate objective is controlled data collection, when conversion tracking is temporarily unavailable, or when a visibility/traffic task is genuinely separate from acquisition. It should not become a permanent substitute for fixing measurement.
With Manual CPC, document maximum bids and review actual CPC, impression share and query quality. With Maximise Clicks, use an appropriate CPC ceiling where available and monitor whether additional clicks come from lower-value terms. Neither strategy knows which click became a profitable customer unless the account supplies that outcome.
When a portfolio strategy helps
A portfolio strategy can share a bidding objective across several campaigns. It is useful when campaigns serve comparable economics and each is individually sparse. Do not pool campaigns merely to increase data volume if one sells £30 accessories and another £20,000 contracts. Shared data is valuable only when the optimisation target has the same meaning.
Document which campaigns belong, why their outcomes are comparable, target changes and any maximum/minimum bid limits. Review results both at portfolio and campaign level so one high-volume campaign does not conceal another’s failure.
Bidding and budget must be read together
Maximise strategies without a target generally seek the most conversions or value within the available budget and may aim to spend it. A campaign constrained by budget can behave differently from one with spare budget. Before declaring a strategy efficient, compare marginal performance as budget expands and separate “lost impression share due to budget” from “lost due to rank”.
For a hypothetical campaign spending £300 per day at £60 CPA, increasing to £450 does not imply five extra conversions. The next auctions may be less efficient. Set a maximum acceptable marginal CPA and scale in measured increments based on mature cohorts, not a fixed percentage rule.
Exceptional events
Smart Bidding already responds to ordinary seasonal patterns. Google provides seasonality adjustments for short, substantial conversion-rate changes and data exclusions for periods with conversion-data problems. These controls should be exceptional, documented and date-bounded. They are not a way to explain every promotion or weak week.
Before using either, estimate the event window, expected conversion-rate change and affected campaigns. Afterward, reconcile actual outcomes and remove stale assumptions. If a tracking outage occurred, repair the source and annotate the period before making target changes.
Weekly and monthly review
Weekly, check strategy status, budget constraint, goal changes, conversion delay, search/placement mix and large deviations. Monthly, compare realised versus target CPA/ROAS, conversion quality, value reconciliation, marginal performance and experiment results. The ROAS guide shows how to translate reported value into break-even economics.
Bidding worksheet
| Input | Document |
|---|---|
| Primary objective | Volume, value, traffic or visibility |
| Conversion actions | Names, primary status, delay and value source |
| Economic constraint | Allowable CPA or required ROAS calculation |
| Data quality | Reconciliation variance and known outages |
| Campaign similarity | Reason a portfolio is or is not appropriate |
| Review | Mature outcome date and rollback condition |
Complete the worksheet before selecting a label in the interface. It exposes when the real task is measurement, value design or budget allocation rather than bidding.
When not to change bidding
Do not change strategy merely because one week missed target, because Google recommends a different label, or because a competitor claims better automation. First check conversion lag, budget, target changes, traffic mix, tracking and business conditions.
If the optimisation input is valid and the deviation remains statistically ordinary, another strategy change can add noise. The decision to wait should be documented with a review date and threshold, not used as an excuse for inactivity.
When Google changes interface labels, verify the documented bidding behaviour before altering campaigns; a renamed control is not evidence that strategy should change.