Demand Gen CPM Billing: Discover, View-Through Optimisation and the Break-Even Math
One 2026 change went largely undiscussed because it affects a minority of accounts — but for those accounts it rewrote upper-funnel economics. Demand Gen CPM billing: for campaigns serving on Discover inventory with view-through conversion (VTC) optimisation enabled, Google moved from cost-per-click to cost-per-thousand-impressions. The change applied automatically; no advertiser action was requested or required.
Google’s reasoning is coherent: if a campaign optimises toward conversions that happen without a click, charging only for clicks misprices the product — most of the value is created by the impression. The advertiser’s view is different: under CPC you pay for a confirmed action, under CPM you pay for an opportunity. And whether the switch helps or hurts is not a matter of principle. It comes down to one number: your click-through rate.
Below: exactly what changed and for which inventory, the break-even formula with a worked table, how to check whether you were affected, and what to do in three typical scenarios.
Demand Gen CPM billing: what actually changed
The change applies to Demand Gen campaigns when two conditions are true at once:
- the campaign serves on Discover inventory;
- view-through conversion optimisation is enabled.
Meet both and that inventory bills on CPM. Miss either and the old model stands. Importantly, VTC optimisation is off by default, so most accounts were never touched. YouTube, Gmail, Display and Maps are not part of this shift.
| Dimension | Before | After (conditions above) |
|---|---|---|
| Billing unit | Click | 1,000 impressions |
| Who carries creative risk | Advertiser | |
| Spend predictability | Lower — depends on click rate | Higher — spend = reach × bid |
| Effect of a strong CTR | More clicks, more spend | More clicks at the same spend |
| Effect of a weak CTR | Few clicks, low spend | Few clicks at the same spend |
The counterintuitive part: CPM punishes weak creative and rewards strong creative. Under CPC a bad ad simply failed to spend money. Under CPM it spends exactly as much as a good one.
The break-even math
CPC and CPM are not directly comparable — different units. Convert through effective CPM:
eCPM = CPC × CTR × 1,000
At a $0.50 CPC and a 1% CTR you were effectively paying $5 per thousand impressions. If the new price per thousand is below that, you gain; above it, you lose. A reference table at a $0.50 CPC:
| CTR | Effective CPM at $0.50 CPC | Read at a $5 CPM |
|---|---|---|
| 0.3% | $1.50 | CPM materially more expensive — danger zone |
| 0.5% | $2.50 | CPM costs twice as much |
| 0.75% | $3.75 | Close to parity, still worse |
| 1.0% | $5.00 | Break-even |
| 1.5% | $7.50 | CPM 1.5× better |
| 2.5% | $12.50 | CPM substantially better |
Treat those figures as illustrative and substitute your own. The five-minute procedure:
- Export the Demand Gen campaign segmented by inventory; keep only the Discover row, 60–90 days before the change.
- Take the average CPC and CTR for that row.
- Compute eCPM = CPC × CTR × 1,000.
- Compare with your actual post-change CPM (cost ÷ impressions × 1,000 over a comparable window).
- The percentage gap is the price of the change for you.
Why traffic cost is only half the question
Comparing cost per unit of traffic answers half of it. The other half is how many conversions that traffic produces and how real they are — which is where the genuine risk of VTC optimisation sits. A view-through conversion is by definition unconfirmed by a click, and its contribution is easy to overstate.
How much can you trust view-through conversions?
A VTC is credited when a user saw the ad, did not click, and later converted. The problem is obvious: some of those people would have converted anyway. Three factors to weigh before building a strategy around VTC.
1. The view window
The wider the window, the more “conversions” you credit yourself and the less causality each one carries. Sensible practice is to start narrow (1–3 days) and widen only with evidence. The general logic of windows and lag is covered in our piece on conversion windows and conversion lag.
2. Browser limitations
Third-party cookie restrictions in Safari and elsewhere make part of your view-through data invisible and part of it distorted. VTC reporting is therefore uneven across audiences — iOS traffic reads systematically differently. Background on the privacy shifts is in third-party cookies after Privacy Sandbox.
3. Frequency
Under CPM your economics depend directly on how many times you show the same person the same ad. Five impressions to one user cost what five impressions to five users cost, but they are not worth the same. Frequency management stops being hygiene and becomes a spend control — see frequency capping and Target Frequency.
Three scenarios and what to do
Scenario 1 — high Discover CTR (above roughly 1.2–1.5%)
You are probably a winner. CPM lets you buy incremental reach without a proportional rise in click cost. Practical moves:
- leave VTC optimisation alone, but validate its contribution separately;
- invest in creative — every extra point of CTR is now effectively free;
- set a sensible frequency cap so you do not overpay for repeat impressions to the same pocket of audience.
Scenario 2 — low CTR (below roughly 0.5%)
CPM is almost certainly costing you more. Two paths. First, turn off VTC optimisation — it is the single toggle that restores the previous billing model for Discover. Second, fix the creative first and decide afterwards. Treat switching off as a stopgap rather than a conclusion: a weak Discover CTR usually means the creative does not fit the surface, not that the surface is bad.
Scenario 3 — you do not know your Discover CTR separately
The most common case. Demand Gen reports blended numbers by default, and the inventory underneath is not homogeneous: YouTube, Discover and Gmail behave very differently. First action is a placement breakdown and a separate look at the Discover line. Until you have it, every decision is blind.
What to check in the account right now
- Whether you have Demand Gen campaigns at all. If you recently moved Display campaigns, they may have landed in Demand Gen automatically — context in our guide to the Display to Demand Gen migration.
- Whether VTC optimisation is on. The setting lives at campaign level in the goals/optimisation block. Off by default, but easily inherited when a campaign is copied.
- Whether you serve on Discover. Check the inventory breakdown. If Discover is under 5% of spend, the change is immaterial for you.
- How CPM and cost per conversion moved. Compare two comparable windows of equal length with no promotional period inside either.
- Log the date. Any intervention should appear in change history, or in a month you will not be able to reconstruct the sequence — see Google Ads change history.
A worked example, end to end
Take a hypothetical campaign and walk the whole calculation, so it is clear where the decision actually gets made. The numbers are illustrative — substitute yours.
| Metric (Discover row, 60 days pre-change) | Value |
|---|---|
| Impressions | 4,200,000 |
| Clicks | 21,000 |
| CTR | 0.50% |
| Average CPC | $0.42 |
| Cost | $8,820 |
| Conversions (click + view) | 310 |
Effective CPM: 0.42 × 0.005 × 1,000 = $2.10. That is what you were really paying per thousand impressions. Now compare with the actual post-change CPM. Say the next 60 days come in at $9,400 of cost and 2,900,000 impressions → CPM = $3.24. Reach got 54% more expensive.
Here is the part that matters. Do not stop at CPM — look at cost per conversion. If conversions came in at 290 on $9,400, cost per conversion moved from $28.5 to $32.4, up 14%. The gap versus the 54% CPM increase reflects reach shifting toward better placements. Which is exactly why “CPM is more expensive, switch it off” is too fast a conclusion: traffic cost and outcome cost move differently.
Your decision threshold comes from your economics. If allowable cost per conversion is $35, this is survivable and the work is in creative. If it is $25, you either fix CTR or turn off VTC optimisation.
Creative for Discover under CPM
Because CPM raises the price of a creative mistake, treat Discover as its own surface rather than “one more placement”.
- Format. Vertical and square images outperform landscape in feed. Supply the full set of ratios — otherwise the system crops for you, not always well.
- The first frame carries everything. There is no search intent in a feed: the user was not looking for you. The image must work without the text and the text without the image.
- Less dense text on the image. Feeds are scanned fast; a busy composition does not survive a small screen.
- Test in batches. Swap assets in groups of three to five and judge by the asset report, not by instinct.
- Split creative by inventory. What wins on YouTube often dies on Discover — different user posture, different contact length.
If your Demand Gen runs as one campaign across all inventory, evaluating Discover creative separately is impossible. Splitting campaigns by surface costs you some automated optimisation but buys control — and under CPM, control is worth more.
Validating view-through value honestly
Google’s report tells you how many view-through conversions it credited. It does not tell you how many would not have happened without the ad. Three methods, from cheap to reliable.
| Method | What it shows | Cost |
|---|---|---|
| Compare VTC windows | Sensitivity of the metric to settings — if narrowing the window barely loses conversions, they are more likely real | Minimal |
| Pause the campaign for 1–2 weeks | Rough funnel-level read on contribution | Lost reach, contaminated by seasonality |
| Geo experiment | Causal contribution by region | Needs volume and 3–6 weeks |
For accounts with sufficient traffic, a geo test is the only way to get an answer that does not depend on the attribution model. The method is laid out in incrementality and geo experiments. Where volume is thin, a narrow view window plus a branded-demand baseline comparison is the workable substitute.
Where to find the numbers
Half the difficulty with this change is that the default Demand Gen view averages everything together. Seeing Discover on its own takes a few deliberate views.
| What you need to see | Where | Columns to add |
|---|---|---|
| Inventory breakdown | Campaign → segment / placement report | Impressions, clicks, CTR, avg CPC, cost, conversions |
| Actual CPM | Any report | Avg cost per thousand impressions (or cost ÷ impressions × 1,000 manually) |
| View-through share | Conversion columns | Conversions, view-through conversions, all conversions |
| Frequency | Reach and frequency report | Unique users, average frequency |
| Asset contribution | Campaign asset report | Impressions, per-asset performance |
Save this as a named column set and review it weekly instead of rebuilding it during every investigation. How to assemble views like this without manual exports is covered in custom columns and the Report Editor.
Pay particular attention to the view-through conversions column: it is not in the default set, and its absence is why accounts run VTC optimisation for years without seeing what share of results it produces. If view-through conversions exceed half of a campaign’s total conversions, that alone warrants an experiment — regardless of the billing model.
Rebuilding budget planning for CPM
Under CPC, budget was consumed in proportion to interest: a weak day spent less. Under CPM, spend becomes a function of reach, and planning changes with it.
- Plan in reach, not clicks. Your planning unit is a thousand impressions to a defined audience, not a number of sessions.
- Set frequency caps explicitly. Without one, the algorithm can hit volume targets by re-serving a narrow audience.
- Separate upper and lower funnel into different campaigns. Mixing CPM reach with conversion traffic in one campaign means losing the ability to evaluate either.
- Recalculate what you can pay. Start from the conversion value your margin supports, not from the CPC you are used to.
The wider pattern
Moving Discover to CPM is part of a broader direction of travel: Google is steadily shifting the upper funnel from paying for clicks to paying for attention, because a growing share of conversions now happens without an ad click. Shoppable CTV in Demand Gen points the same way, as does the general move of measurement toward modelling. The practical consequence for a media buyer: creative and frequency become economic levers, not just quality concerns.
The lower funnel is shifting in parallel — see the move from call-only ads to call assets, and how offers now reach your ads through automated promotions pulled straight off your landing page. The through-line is the same everywhere: fewer manual levers, and a higher price on preparation — creative, site, measurement.
If you want the break-even run on your own numbers, that is what a PPC Rebels consultation is for; account and access setup is described on the agency account page.
FAQ: Demand Gen CPM billing
Was I affected at all?
Only if you run Demand Gen campaigns serving on Discover with view-through conversion optimisation enabled. VTC optimisation is off by default, so most accounts were not touched.
How do I go back to click-based billing?
Disable view-through conversion optimisation in campaign settings. It is the single toggle that restores the previous model for Discover inventory — and it removes the benefits of VTC optimisation along with it.
Does this affect YouTube and Gmail?
No. This change moves only Discover inventory to CPM, and only with VTC optimisation on. YouTube, Gmail, Display and Maps keep the previous model.
What is the fastest way to tell if it got more expensive?
Compute effective CPM as CPC × CTR × 1,000 for a pre-change period and compare it with your actual CPM afterwards. The gap is the price of the change for your account.
At what CTR does CPM win?
Break-even depends on your CPM rate. The rule of thumb is simple: the higher your CTR, the better CPM looks. Parity tends to land somewhere near a 1% CTR at typical Discover rates, but verify against your own numbers.
Do I need new creative?
If you are staying on CPM, yes — creative priority rises. When you pay per impression, a weak ad costs exactly what a strong one costs, so creative investment pays back directly rather than indirectly.
What should I do about frequency?
Set an explicit cap. Under CPM, repeat impressions to the same person are billed the same as first impressions, and without a limit spend drifts into over-contacting a narrow audience.
Can I trust view-through conversions?
With caveats. Some are not causal and some are distorted by browser restrictions. The workable approach is to start with a narrow view window and validate contribution with an experiment rather than accepting the report at face value.
How does this affect overall attribution?
View-through conversions can overlap with click conversions from other campaigns. If VTCs feed the same goals as everything else, you risk double-counting value. Goal separation and window awareness are prerequisites for an honest read.
Should I run Demand Gen with VTC optimisation at all?
It makes sense once you have evidence that views carry value — a geo test result, or a stable link between impressions and branded demand. Without that evidence, enabling VTC optimisation (and with it CPM billing) is premature.
How do I budget under CPM?
In reach units: how many thousand impressions against your target audience, at what frequency. Clicks become a derived metric rather than a planning input.
The honest conclusion: a change of billing unit does not make Demand Gen a good or bad channel. It changes who carries the risk of weak creative. If you are confident in your creative and can prove the value of views, CPM works in your favour. If you cannot, fix those two things before arguing about the billing unit.
