Ad refresh loads another impression in a slot the reader is still looking at. In Prebid that has to be a new auction, not a second render of the keys you already sent.
A refresh that only calls the ad server again will send the same hb_ targeting. Prebid's refresh example says the result: the same bid is chosen and rendered over and over, without Prebid being asked for new prices. The documented order is requestBids for that ad unit, then setTargetingForGPTAsync for that unit, then googletag.pubads().refresh on that slot.
Refresh changes how many impressions a session produces. Judging it on CPM alone misses the point. More ads can mean less attention per ad, lower viewability, and a buyer who did not agree to be called that often.
Google's documented minimum time between refreshes is 30 seconds. bidtune will not set a faster interval. A slot that was not refreshing is not given a refresh by this test.
Prebid's example refreshes one unit. PREBID_TIMEOUT is the auction timeout for that call. The ad unit code and the GPT slot have to name the same unit. The demo path in the official example is theirs. Use your slot.
function refreshBid() {
pbjs.que.push(function () {
pbjs.requestBids({
timeout: PREBID_TIMEOUT,
adUnitCodes: ["/19968336/header-bid-tag-0"],
bidsBackHandler: function () {
pbjs.setTargetingForGPTAsync(["/19968336/header-bid-tag-0"]);
googletag.pubads().refresh([slot1]);
}
});
});
}A slot that stays in view can earn another impression if demand is still there and the reader has not left. A cap keeps a long session from refreshing without limit. Pairing refresh with bid caching lets a still-valid bid compete with the new responses.
Refreshing faster than demand or the ad server allows burns viewability and can break a buyer's rules. Refreshing a slot that is off screen, or in a background tab, adds impressions nobody saw. Calling refresh without a new auction re-bills the same creative. A higher impression count with lower revenue per session is a loss.