Value-based bidding

Ordinary conversion bidding treats every conversion as equal. Value-based bidding lets the advertiser say how much each one was worth, and the platform then bids more for people likely to produce high values. Google describes two forms: maximise conversion value, with or without a Target ROAS. Meta's equivalents are value optimisation and its ROAS goal.

Search arbitrage is an obvious fit, because the value of a visit varies widely. A visitor who clicks a legal-services ad may earn many times more than one who clicks a recipe ad. If the tracker sends each visit's revenue to the platform through a Conversions API or postback, the algorithm learns to seek the valuable visitors and to pay less for the rest.

The approach is only as sound as its inputs. Values are usually estimated revenue, modelled per click and sometimes late, so the platform is learning from an approximation. There is also a quality trap: if inflated values come from clicks later removed as invalid, the algorithm has been trained to find more of exactly the wrong visitors. Sending realistic values, and correcting for sources with heavy deductions, keeps it pointed the right way.

An example

Say audience A converts 30% of the time at $0.50 per conversion ($0.15 per visit) and audience B converts 20% of the time at $1.50 ($0.30 per visit). Conversion-count bidding prefers A. Value-based bidding prefers B, which earns twice as much per visit.