Pay for Conversions in Google Ads 2026: Eligibility, Hidden Cost and the Demand Gen Problem
Google Ads has a billing model where you are charged not for clicks and not for impressions, but only for conversions you actually receive. It is called pay for conversions and it runs in Display campaigns on Target CPA bidding. On paper it is the perfect deal: no conversion, no invoice.
In practice the model has three hard eligibility thresholds, several incompatibilities, and one billing behaviour that surprises everyone the first time it happens. More importantly: pay for conversions does not carry over when Display campaigns migrate to Demand Gen — you have to switch the payment event to “interaction” before the move. Below: what is required, how to calculate whether the trade is worth it for you, and what to do once the model is gone.
Pay for conversions does not make advertising cheaper. It transfers the risk of underperformance from you to Google — and Google charges for that by being far more selective about which auctions are worth buying at your target CPA.
What pay for conversions is, and how it differs from ordinary Target CPA
Standard Target CPA is a bidding strategy: Google tries to win conversions at your stated price but bills you for clicks. If the algorithm misses and conversions come in below forecast, your actual CPA lands above target — and you still pay the full click cost.
Pay for conversions changes the payment event. Formally it is still Target CPA, but the charge trigger is a conversion. You are not billed for impressions, interactions, or the clicks that led to the conversion — only for the conversion itself, at exactly your target CPA. Google’s own example: at a $10 target CPA with 10 conversions in a month, you are billed $100.
| Dimension | Target CPA (pay per click) | Pay for conversions |
|---|---|---|
| What triggers a charge | A click | A conversion |
| Who carries the miss risk | The advertiser | |
| Actual CPA | Floats around target | Equals target by construction |
| Campaign types | Nearly all | Display campaigns only |
| Click data in reports | Full, including cost per click | Clicks visible, no click cost |
| Daily budget overshoot | Up to 2x | Can exceed 2x |
The three eligibility thresholds
Access is not a preference setting — it is gated on account metrics.
| Requirement | Threshold | Why it exists |
|---|---|---|
| Conversion volume | more than 100 conversions in the last 30 days | Google needs statistics to price the risk it is taking |
| Conversion speed | 90% of conversions occur in less than 7 days after the click | A long lag makes billing indeterminate |
| Conversion delay | a delay over 7 days disqualifies you | Billing is tied to the recording window |
| Target CPA size | below $200 (or local equivalent) | Above it, warnings and overcharges become possible |
| Store visits | must be excluded from conversion reporting | Modelled conversions cannot be billed |
The store visits requirement is not cosmetic. Store visits are a modelled, extrapolated metric, and including them in billable conversions would mean invoicing for an estimate. So they have to come out of the campaign’s conversion set.
The lag requirement should be checked against your data rather than your intuition. Open the report on conversion windows and conversion lag and look at the distribution: if 15% of your conversions arrive on day eight or later, you are ineligible even if the average lag looks short.
What pay for conversions cannot be combined with
- Shared budgets. A pay-for-conversions campaign cannot share a budget pool.
- Portfolio bid strategies. Campaign-level strategy only — combining it with portfolio bid strategies and shared budgets is not possible.
- Bid adjustments and seasonality adjustments. A pay-for-result model leaves no room for manual multipliers.
- Modelled conversions. Anything not recorded as a fact stays out of the billable set.
The operational consequence: a pay-for-conversions campaign is an isolated unit. It cannot join a budget pool, cannot be nudged with multipliers, cannot be bundled into a portfolio. For agency process that is a downside — such campaigns fall outside shared budget workflows, including Recommended Investment Strategy with its Holistic and Growth modes.
Under the hood: why a fixed CPA is not free
The key to this model is realising Google is not running a charity — it is taking on risk and pricing it.
Under pay-per-click, the algorithm can afford moderately risky auctions: even if no conversion follows, the click has already been paid for. Under pay for conversions, the same algorithm funds impressions and clicks out of its own pocket and only collects a fixed amount when a conversion is recorded. So it only makes sense to buy auctions where conversion probability is high enough to cover the whole chain.
Three observable consequences follow.
Narrower reach. A pay-for-conversions campaign takes narrower, hotter inventory on average than the same campaign paying per click at the same target CPA. You buy price predictability with volume.
A cautious start. While statistics accumulate the algorithm behaves conservatively. That is a normal part of the Smart Bidding learning period, but it feels sharper here because an algorithmic miss costs Google, not you.
Extreme sensitivity to tracking quality. The model bills only conversions Google recorded. Any lost signal — a broken tag, missing cross-domain tracking, duplicates — distorts both the invoice and the algorithm’s behaviour.
Daily budget overshoot: how it shows up on the invoice
The ordinary model allows up to two daily budgets of spend on a given day, offsetting it on other days under the 30.4 monthly multiplier. Under pay for conversions, daily spend may exceed your average daily budget by more than two times — Google states this explicitly, and it follows from the model: conversions arrive unevenly and are billed at the moment they are recorded.
Practically, if you treat budget pacing and the 30.4 rule as a firm daily ceiling, recalibrate that expectation. Control moves to the weekly and monthly level: stop asking “what did we spend yesterday” and start asking “what did we spend over 7 and 30 days”.
Second wrinkle: conversions recorded late are billed late. At month end this creates a tail effect, where campaign spend catches up after you have closed the reporting period.
What breaks when Display moves into Demand Gen
This is the most consequential practical point of 2026. Display campaigns are moving into Demand Gen: from June 2026 the migration tool began rolling out, after which new Display campaigns can only be created inside Demand Gen, and the remainder auto-migrates without advertiser action.
What carries over and what does not:
| Setting | Status in Demand Gen | What to do |
|---|---|---|
| Target CPA | supported | nothing |
| Max conversions / value, Target ROAS, Max clicks | supported | nothing |
| Pay for conversions | not supported | switch the payment event to “interaction” before migrating |
| Manual CPC | not supported | move to Target CPC |
| Viewable impressions | not supported | move to Max clicks |
| Bid adjustments and seasonality adjustments | not supported | remove before migrating |
| Shared budgets, product filters, ad group exclusions, disclaimers, lead forms | not supported | remove before migrating |
Migration mechanics: migrated campaigns are named “[original campaign name] #2”, the originals move to Removed status and are retained for reporting for up to five years. Enabled/paused status carries over, business logos are mandatory (a placeholder is created if missing), and ads are resubmitted for review.
If you are running pay for conversions today, you have a window to prepare the move to interaction-based Target CPA — and that is not a mechanical swap, it is a change in campaign economics. The move itself is covered in the breakdown of Display campaigns migrating to Demand Gen.
Paying for impressions: the Discover and VTC shift to CPM
Another 2026 billing change affects Demand Gen. On 16 June 2026 Google announced that Demand Gen campaigns using view-through conversion (VTC) optimisation on the Discover placement move from cost-per-click to cost-per-mille billing. Effective date: 15 July 2026, applied automatically.
Important qualifiers:
- Only campaigns that both serve on Discover and have VTC optimisation enabled are affected.
- VTC optimisation is off by default, so most advertisers never noticed the change.
- Turning VTC optimisation off in campaign settings reverts the billing model but gives up VTC-optimised bidding.
- Other Demand Gen placements — YouTube, Gmail, Display, Maps — and campaigns without VTC optimisation are untouched.
The meaning of the shift: you pay for attention rather than engagement. For a media buyer that changes the evaluation method — under CPM billing you cannot judge on CPC, and the only defensible efficiency metric becomes cost per result, not cost per click.
Three billing models in 2026: which one are you on?
| Model | Charge trigger | Where it appears | Main risk |
|---|---|---|---|
| CPC / per interaction | click or interaction | Search, Shopping, PMax, Demand Gen, Display | Paying for traffic with no result |
| CPM / per impression | thousand impressions | Video, reach formats, Demand Gen with VTC on Discover | Paying for attention with no click |
| Pay for conversions | recorded conversion | Display campaigns on Target CPA | Narrow reach, and the model leaves with Display |
The question worth asking before every launch: which event triggers a charge, and does it match what I actually need from this campaign? A mismatch here is the root of most arguments about “expensive advertising”.
How to run the numbers yourself: the indifference point
Before changing the model, work out the conversion rate at which you are indifferent to what you pay for. The formula is simple:
CRindifference = CPC ÷ target CPA
At a $0.40 click cost and a $20 target CPA, if you convert better than 2% of clicks (0.40 ÷ 20 = 0.02), pay-per-click wins: you pay less than Google would bill you per conversion. Convert worse than 2% and pay for conversions wins — you push the cost of misses onto the platform.
Then run the second calculation, on volume. Pay for conversions almost always yields fewer conversions in absolute terms at the same CPA. Do it like this:
- Take the campaign’s current weekly conversion volume under pay-per-click.
- Decide how much volume you are willing to give up for price certainty. For planning purposes assume a 20–40% drop — treat that as a rough orientation, not a benchmark, and verify it on your own campaign.
- Compare: the value of eliminating misses versus the margin lost on conversions you never got.
In most accounts with a healthy conversion rate and stable tracking, pay-per-click wins on absolute profit. Pay for conversions wins where spend predictability matters more than volume, and where you are unwilling to fund the algorithm’s learning phase.
Step by step: check eligibility and switch on
- Check volume. More than 100 account conversions in 30 days.
- Check the lag distribution. 90% of conversions must land in under 7 days. Read the report, not the average.
- Remove store visits from the campaign’s conversion set.
- Check the target CPA. It must sit below $200 or local equivalent.
- Pull the campaign out of any shared budget and portfolio strategy.
- Set Target CPA at campaign level and switch the payment event to conversion.
- Freeze a baseline. Spend, conversions, CPA and volume for the 28 days before the switch.
- Leave it alone for two weeks and compare on weekly, not daily, figures.
What the model hides from you
Under pay for conversions you lose part of the analytical picture. There is no click cost, so the familiar cuts disappear: expensive placement, expensive device, expensive audience. There is nothing left to optimise on traffic cost inside the campaign — you only see the final cost per result.
Second, the middle of the funnel goes dark. You cannot tell whether a decline came from traffic or from the landing page, because you no longer have a traffic price. The only compensation is external analytics: paid traffic analysis in GA4 Explorations plus your own on-site behavioural data.
Third, the model never shows you the price of the risk transfer. Google does not publish how much more conservatively it buys auctions under pay for conversions. The only way to find out is to compare conversion volume before and after the switch on your own campaign.
Who should skip it, and when it is not needed
Skip it if you have a short cycle, a stable conversion rate and confidence in your tracking: paying per click almost always wins on absolute profit in that situation. It is premature if you have fewer than 100 conversions a month — you simply do not clear the threshold. And skip it if the campaign lives inside a shared budget you are not willing to dismantle.
It earns its place in two scenarios. First, a hard contract with a fixed cost per lead, where predictability beats volume. Second, a vertical or geography that is new to you, where you are not prepared to fund the algorithm’s learning out of your own budget.
Seven mistakes
- Expecting the model to lower CPA. It pins CPA to target; it does not reduce it.
- Judging eligibility on average lag instead of distribution. The rule is 90% under seven days, not “average lag four days”.
- Leaving store visits in the conversion set. The campaign simply will not qualify.
- Expecting the same volume. Narrower reach is not a bug; it is the cost of transferring risk.
- Policing spend daily. Daily overshoot here can exceed 2x; control weekly only.
- Entering the Demand Gen migration without changing the payment event. A pay-for-conversions campaign will not migrate as-is.
- Changing the target CPA on day two. The algorithm is still pricing the risk; early edits produce oscillation.
What to use instead once the model is gone
- Target CPA with interaction billing in Demand Gen. Same strategy, different economics: the risk comes back to you.
- Clean the inbound traffic. If the billing model used to be your insurance, targeting and exclusions have to become your insurance instead.
- Stricter primary conversions. Separate primary and secondary conversions so the algorithm learns on real deals rather than micro-goals.
- Move to value-based optimisation. Switching to conversion value and value rules gives sturdier control over economics than a payment event ever did.
- Test, do not believe. Push the model change through a split test rather than a month-over-month comparison.
FAQ
How many conversions do I need to enable pay for conversions?
More than 100 conversions in the account over the last 30 days. Below that the model is unavailable.
What is the conversion delay requirement?
90% of conversions must occur in less than 7 days after the click. With a delay over 7 days the model cannot be used.
Which campaign types support pay for conversions?
Display campaigns on Target CPA only. It does not exist in Search, Shopping, Performance Max or Demand Gen.
Why must store visits be excluded?
Store visits are modelled and extrapolated. An estimate cannot be billed, so they are removed from the campaign’s conversion set.
Is there a cap on target CPA?
Yes — keep it below $200 or local equivalent; above that, warnings and overcharges become possible.
How far can daily spend exceed the budget?
Daily spend may exceed your average daily budget by more than two times. Control spend weekly and monthly, not daily.
Can I use it with a shared budget?
No. Pay for conversions is incompatible with shared budgets and portfolio bid strategies.
What happens when my campaign migrates to Demand Gen?
Pay for conversions is not supported in Demand Gen. Before migrating, switch the payment event to “interaction” and move to Target CPA.
Will my CPA change after switching to interaction billing?
The target stays, but actual CPA will start floating around it because the miss risk returns to you. Plan for variance and do not react to week one.
Demand Gen moved to CPM — does that affect everyone?
No. Only campaigns serving on Discover with view-through conversion optimisation enabled, from 15 July 2026. VTC optimisation is off by default.
Does pay for conversions protect me from invalid clicks?
On the billing side, yes — you are only charged for conversions. Traffic quality still matters though: filtering invalid clicks affects the data the algorithm learns from.
How do I tell quickly that the model is wrong for me?
Compare two weeks before and after on conversion volume, not CPA. If volume fell further than you were willing to lose, price predictability cost more than it was worth.
What to do next
The billing model is one of three levers that set campaign economics. The second is where the budget goes between campaigns — that is Recommended Investment Strategy with Holistic and Growth modes. The third is who gets credit for the conversion, namely DDA versus last click attribution models — because the same bidding you configure here trains on attributed credit.
If you are opening new verticals and would rather not test them on a live production account, look at Google Ads agency accounts; a structured walk-through of bidding and billing models is part of the Google Ads guide from PPC Rebels.
