Adaptive bidding, also called smart or automated bidding, is an auction strategy in which the advertising platform sets a different bid for every individual impression opportunity in real time, using machine-learned signals — device, operating system, placement, time of day, user history, predicted conversion probability and predicted value — to decide how much that specific impression is worth to the advertiser.
Definition
Adaptive bidding, also called smart or automated bidding, is an auction strategy in which the advertising platform sets a different bid for every individual impression opportunity in real time, using machine-learned signals — device, operating system, placement, time of day, user history, predicted conversion probability and predicted value — to decide how much that specific impression is worth to the advertiser. The buyer no longer names a fixed price per click or per thousand impressions; instead they hand the platform a target, such as a cost per acquisition or a return on ad spend, and a budget, and the system spends against that target impression by impression.
The mechanism only works well when it is fed a reliable stream of conversion data. The platform learns which impression characteristics precede a conversion and bids up on lookalike opportunities while bidding down or skipping the rest.
If the conversion signal is thin, delayed or mislabelled — for example a postback that fires on registration when the real target is a qualified deposit — the model optimises toward the wrong outcome and can move spend in an unprofitable direction quickly, because it is bidding aggressively on its (wrong) understanding of value.
In context
For iGaming affiliates the practical shift is from manual bid tuning to conversion-data engineering. The work becomes: implement a clean server-to-server postback, pass back the right event (qualified FTD, not raw registration), send a value where possible so the model can chase whales rather than minimum deposits, and accumulate enough events — most platforms need roughly 50 of the target event per week per ad set — before the system can optimise reliably.
Below that threshold, manual or semi-automatic bidding is usually steadier.
The trade-off is transparency and control. The buyer still owns the target, the creative and the audience seed, but not the price paid for any single auction, so a broken tracking setup, a sudden approval-rate drop at the operator, or a mislabelled event are far more dangerous than under manual bidding, where a bad day is capped by the fixed bid.
Experienced buyers run adaptive bidding with tight alerting on cost per result and daily reconciliation against the advertiser's numbers.
Worked example
After a campaign accumulates 60 qualified-FTD events through its postback, the buyer switches the ad set from a $0.35 manual CPC to a target-CPA of $35 and lets the platform bid each auction. Cost per FTD settles at $33 over the next week, with the platform paying anywhere from $8 to $22 per click depending on predicted value.
Frequently asked questions
Browse the full iGaming & affiliate glossary — hundreds of EN/RU terms with examples.
← Back to glossary