Google Ads Target CPA ignores device bid adjustments. Mobile took 69% of spend
Ten days, enabled campaigns, split by device.
Device | Spend | Share | Clicks | Conversions | Conversion rate | Cost per lead |
|---|---|---|---|---|---|---|
Mobile | $59,010 | 69% | 6,462 | 169.4 | 2.6% | $348 |
Desktop | $19,975 | 23% | 1,547 | 62.4 | 4.0% | $320 |
Tablet | $4,091 | 5% | 404 | 20.0 | 5.0% | $205 |
Connected TV | $2,758 | 3% | 39 | 0.0 | 0.0% | n/a |
Mobile took more than two-thirds of the budget and converted at the lowest rate of the three real devices. Tablet (the device everyone stopped thinking about in 2016), converted at nearly twice mobile's rate, at 59% of the cost per lead, and received a twentieth of the spend.
That's not a small gap. If mobile converted at desktop's rate, the same $59,010 would have produced 258 leads instead of 169.
Why the old fix doesn't work anymore
The manual-bidding answer to this was a device bid adjustment: mobile −30%, tablet +40%, done. It moved money toward the devices that converted.
Under smart bidding (Target CPA, Maximise Conversions, Target ROAS), device bid adjustments are ignored. The algorithm is already meant to be accounting for device at auction time, so the platform discards your adjustment as redundant. The one exception is −100%, which excludes the device entirely.
Bid strategy | Device bid adjustments |
|---|---|
Manual CPC | Applied |
Enhanced CPC | Applied, then modified by the algorithm |
Target CPA / Maximise Conversions / Target ROAS | Ignored, except −100% |
So the campaigns in this account that had just moved to Target CPA had no way to tilt spend away from mobile short of switching mobile off. The lever that used to fix this is gone.
What the algorithm should be doing, and evidently isn't fully
The argument for removing the lever is that the model sees device and bids accordingly. If mobile converts at 2.6% and desktop at 4.0%, a model targeting a fixed cost per lead should bid roughly 35% less on mobile.
Maybe it is. But 69% of spend still went to mobile. Either mobile is 69% of the available auction volume and the model is bidding it down as far as it usefully can, or the model hasn't learned the device pattern yet. In both cases the outcome is the same: most of the money is on the device that converts worst, and there's no bid you can set to change that.
Explanation | What it implies |
|---|---|
Mobile is most of the inventory and the model is already discounting it | The problem is the mobile experience, not the bidding |
The model hasn't learned device effects yet | Wait, but the account can't afford to wait indefinitely |
Mobile traffic is genuinely lower intent | Same conclusion as the first row, fix the page or accept the rate |
The mobile page was the problem
The account's main landing page weighed 245 KB and took 0.62 seconds to load on a fast connection. It carried six separate contact forms. On a phone, on a mobile network, in the car park of a hardware store, that page is a different experience from the one the desktop conversion rate reflects.
A rebuilt version:
Existing page | Rebuilt page | |
|---|---|---|
Weight | 245 KB | 45 KB |
Load time | 0.62 s | 0.26 s |
Forms on page | 6 | 1 |
That's the fix that survives smart bidding. A page that loads in a quarter of a second and asks for one thing converts on mobile. A page that loads in over half a second with six competing forms doesn't. No bid adjustment ever addressed that; it just moved money away from it.
The two options you actually have
Option | When | Cost |
|---|---|---|
Fix the mobile page | Mobile CVR is well below desktop and the page is heavy | Development time; the right answer in most accounts |
Exclude mobile (−100%) | The product genuinely doesn't convert on mobile and never will | Losing 69% of available volume, rarely correct |
Separate mobile campaign with its own target | Mobile converts but at a different economics | Doubles campaign management; fragments conversions |
There is no fourth option that involves a bid. That era is over.
The tablet footnote
Tablet's 5.0% conversion rate on 404 clicks is a small sample, and tablets skew toward older, home-based users who are, for a residential solar business, close to the ideal customer. It may not generalise. But it's the same page, the same ads, the same auctions, and it converts at twice mobile's rate. The device is not the variable. The screen size and the context are.
That's the argument for fixing the page. The traffic isn't bad. The experience it lands on is.
What to check in your account
Segment by device. If mobile is the majority of spend and the minority of conversion rate, you have this. Then load your landing page on a phone, on mobile data, and time it. If it's over a second, or if it asks the visitor to choose between multiple forms, the device report is telling you about the page, and the page is where the fix lives.