tCPA: Target CPA

Target CPA is an automated ("smart") bidding strategy in which you tell an ad platform the average cost you are willing to pay per conversion, and its machine-learning model sets each auction bid to win as many conversions as possible at or around that target. The platform, not you, adjusts bids in real time.

It works by predicting, for every incoming impression, how likely that user is to convert. When the model expects a high conversion probability it bids up; when it expects a poor fit it bids down or skips. Over thousands of auctions the average cost per conversion gravitates toward your target, even though individual conversions land above or below it.

For a Forex affiliate paid a fixed CPA by the broker, tCPA is a margin-control tool. If the broker pays $250 per qualified funded trader, setting a $100 Target CPA aims to keep acquisition cost low enough to protect roughly $150 of gross margin per acquisition — while removing the daily grind of manual bid tuning so campaigns can scale.

The trade-off is that automation needs conversion data to learn. Google generally recommends a recent history of conversions before tCPA performs well, and setting the target too low can starve the campaign of impressions entirely.

Target CPA sits on top of the ad auction. For each query or impression the model estimates conversion likelihood using signals like device, location, time, query, and audience, then sets a bid designed to hit your target cost on average. High-probability users get aggressive bids; low-probability ones get little or nothing.

The strategy depends on a working conversion signal and enough volume to train on. If your conversion action fires reliably (a verified registration or a funded-account event passed back from the broker or CRM), the model optimises toward genuinely valuable actions. If the signal is noisy or the target is set below what the auction can deliver, the system either overspends chasing conversions or throttles delivery and starves the campaign.