You set the scope.
Choose the settings and the traffic you are willing to put into a test.
Alongside the Prebid you already run
Continuous experiments for your Prebid.
Ad revenue optimization is the kind of problem AI is good at. BidTune applies that work 24/7: small tests, a 10% holdback on your current configuration, and a ledger of every change with its measured lift.
14 days free after activation. No credit card to get started.
See pricing and trial details.
One script URL Your partners stay 10% holdback, untouched
AutoResearch
Illustrative cycle. Not a customer result.
One change
at a time.Measure. Keep. Repeat.
at a time.Measure. Keep. Repeat.
01 / Observe
Every auction is a signal.
Responses, timeouts, renders, and the settings that reached the page.
Built by the Optimize team at BuySellAds.
Working with publishers since 2008.
Questions the auction can answer
Look at the bid landscape.
Then test one thing.
Bidder health, timeouts, and refresh are ordinary Prebid questions. BidTune keeps the evidence next to the question.
Requests → responsesIllustrative data
Bidder A
Responding
Bidder B
Responding
Bidder C
Review
Requests sent Responses received
See who answers, and who is quiet.
Response, contribution, and request volume, side by side. A quiet adapter is a question for a test.
WHAT TO INVESTIGATEIs this bidder adding demand on this traffic?
Autoresearch
The same question,
run as an experiment.
Autoresearch is continuous experimentation: one question from the auction data, a small test, and a written result. Here is how a timeout question moves through it.
- 01 · Investigate
Is there value past the timeout?
See when bids arrive before spending test traffic on a longer wait.
- 02 · Evaluate
One change. A slice of traffic.
The test runs on 30% of optimized traffic. The 10% holdback is not part of the test.
- 03 · Learn
Keep it only if it measures better.
The ledger records the change and its measured lift. Otherwise it rolls back.
The next question starts from the ledger.
Inspired by Andrej Karpathy’s autoresearch on GitHub .
Fits the stack you have
One script URL.
Your partners stay.
Keep your bidders, ad server, and the people who already work on your Prebid. Change the URL of the script you already load.
Your existing Prebid script
your-prebid.jsYour site’s BidTune URL
See the script changes.bidtune.net/Example URLs. After you add your site, BidTune shows the script tag to install.
Your setup. Your decisions.
You read the ledger.
Every change BidTune made, whether it applied, and its measured lift.
You can stop it.
Pause the work or undo a change. Your current configuration is still there.
Illustrations of publisher controls
Continue with Google and add your site.
Your free trial starts when the tag is collecting data.
Fourteen days on your traffic
Your trial uses
your own auctions.
Fourteen days free after activation. The experiments run on your setup, with the same rules as after the trial.
- 01 · See the auction
Your traffic, as it arrives.
Bidder activity, response timing, and rendered delivery. The same setup you already run.
A view of your auctions. - 02 · Test one change
A slice of traffic.
Tests use 30% of optimized traffic. Each one needs at least 72 hours and 2,000 sessions in each arm.
The 10% holdback stays on your current configuration. - 03 · Read the ledger
Keep only what measures better.
Every change BidTune made, with its measured lift. A change that is not measurably better is rolled back.
No lift number without the measurement behind it.
Some tests need longer than 14 days to reach 2,000 sessions per arm. The holdback comparison is shown when there is enough data.
Put continuous experiments on the setup you already run.
14 days free after activation. No credit card to get started.
See pricing and trial details. Setup review comes first.