


The Ad That Got Away
A second path when DNS fails.
An empty ad slot doesn't always mean nobody wanted it. Sometimes, buyers never got the chance to bid.
We saw this when our ad load errors fell sharply. Some devices could no longer find CloudX's server. Our SDK, the ad software inside an app, couldn't start, so those devices stopped requesting ads. The chart looked healthier because fewer requests made it far enough to fail.
For publishers, those missing requests meant lost chances to earn. We had to get those devices connected before either side could take part in an ad auction.
The DNS lookup before the auction
Before the SDK can request an ad, it has to find its server. It knows the server by a name, such as ads.example.com, but a connection needs a numerical IP address.
DNS, short for Domain Name System, is the internet's address book. The device asks a lookup service, called a DNS resolver, to find the address for that name. Then it can try to connect.
DNS can fail for several reasons: outages, timeouts, or replies an app can't use. Network filtering is one possible cause.
For the affected devices, the usual DNS service had stopped returning a usable address for CloudX's startup server.
How CloudX's DNS fallback works
The SDK first tries the device's usual DNS service. If the lookup or connection fails, our fallback asks another service using DNS over HTTPS, or DoH: a DNS lookup carried inside an encrypted web request.
The SDK can then try to connect to the returned address. It still checks the server's identity against the requested name using TLS, the security technology behind HTTPS. Healthy connections keep their existing path.
Recovery can take longer because the normal path has already failed. If the fallback also fails, the SDK still cannot start.
Three paths before the auction
How DNS results change whether the SDK can start.
Usual DNS works
- Address found
- Secure connection
- Startup can proceedAn auction can begin
Usual DNS fails · No fallback
- No usable address
- No server connection
- Startup stopsNo auction
Usual DNS fails · Fallback succeeds
- Ask another DNS service
- Secure connection
- Startup can proceedAn auction can begin
Conceptual request flow. Spacing does not represent elapsed time.
Recovery is conditional; successful startup does not guarantee an ad.
Bidding with the buyers who can deliver
Reaching our server only helps if the winning ad can load. Other ad networks' SDKs fetch their ads from their own servers.
On devices where our SDK regularly falls back to DoH, ads from some networks failed to load 85–97% of the time. When the SDK falls back, it tells the auction how the original DNS lookup failed. The auction uses that to decide whether to include only the networks whose ads tolerate those DNS issues. This is decided request by request, so once DNS resolves normally again, even within the same session, all configured bidders are back in the auction.
Without this, the SDK could start but still not deliver ads as expected.
From a connection to an ad
An Android game publisher ran a one-week A/B test of the fallback. Users on the same version of the app were split evenly between fallback enabled and disabled, with approximately 700,000 in each group. The publisher shared the results anonymously.
The clearest change was among the roughly 6% of test users whose DNS service blocked ad servers. With fallback disabled, 15% of those users saw any ad. With fallback enabled, 69% did. The share seeing an ad was about 4.6× as high.
Among users in that group whose apps tried to start the SDK, success rose from 17% to 98%: about 5.8× the rate with fallback disabled.
Across the full test, startup success among users whose apps attempted it rose from 88.8% to 98.1%, and ads per user rose 2.0%.
For that same full-test group, the publisher's reported lift in revenue per user from ads delivered through CloudX extended up to 12.2% compared with fallback disabled. That is the upper end of the reported interval.
One-week A/B test
One Android game · ~700,000 users per group
≈5.8×
Startup success vs. fallback disabled, among users whose apps attempted it
≈4.6×
Share who saw any ad vs. fallback disabled
Up to12.2%
Upper end of reported lift in CloudX revenue per user, whole test
First two metrics: DNS-filtered users, approximately 6% of the test.
SDK startup success
Among users in the DNS-filtered subgroup whose apps attempted startup
- Fallback disabled
- 17%
- Fallback enabled
- 98%
Users who saw any ad
Share of all users in the DNS-filtered subgroup
- Fallback disabled
- 15%
- Fallback enabled
- 69%
The bidder selection described above was also enabled. These results cover that setup in one game over one week; they don't separate each feature's contribution or predict the same gains for every app.

Publisher test summary
“The rescue works and pays.”
The ad that got away never made it to an auction. Giving the app another way to find our server helped bring those opportunities back. Every recovered request is another chance for a publisher to earn and a buyer to bid. That's why reliable ad delivery starts before the auction does.
The SDK behind CloudX Exchange
That connection is the starting point for CloudX Exchange, too. Exchange brings participating demand partners into publishers' auctions as one bidder.
For publishers, recovering a connection creates another chance to earn from an ad. Exchange brings demand into your auction under the same targeting and floors as your other bidders. Existing CloudX publishers can enable Exchange on eligible line items without adding another SDK or adapter.
For demand partners, connect over OpenRTB to reach mobile in-app inventory without building a mobile SDK of your own. CloudX's SDK renders the winning creative. Its connection recovery helps keep failed lookups from ending ad opportunities before they reach your bidding endpoint.
To join CloudX Exchange as a publisher or demand partner, talk to us.
