Smart Bidding.

The difference from manual bids lies in the timing. A manual bid applies equally to all auctions for a keyword; Smart Bidding decides afresh for each auction and takes into account signals that do not appear as columns in the account: device, location, time of day, browser, language, the exact wording of the query and characteristics of returning users. That is precisely the advantage, and at the same time the reason why results are harder to trace. Anyone using Smart Bidding trades control over the individual bid for computing power across many auctions.

The strategies differ in what they maximise. 'Maximise conversions' gets as many conversions as possible out of the budget, regardless of the cost per conversion. Target CPA instead holds a desired cost per conversion. 'Maximise conversion value' and target ROAS work with revenue rather than numbers of conversions and are suited to cases where conversions are worth different amounts, in retail as well as in services with strongly varying order sizes. All four require conversions to be measured cleanly and completely.

A typical error pattern as an illustration: a trade business counts every submitted contact form as a conversion, including pure price enquiries from abroad and applications for open positions. Target CPA then reliably optimises towards cheap, worthless forms. The account looks better in the report, and the phone rings less often. Only when just qualified enquiries count as conversions and the job application path is excluded does the automation work towards the right goal. Smart Bidding amplifies the definition of the conversion, for better or for worse.

Smart Bidding is the wrong choice when three things are missing: reliable conversion tracking, a sufficient number of conversions and patience for the learning phase. With very few conversions a month, the model has too few data points and overreacts to chance; manual or click-based bidding is then often more stable. Abrupt interventions are also risky: changing targets every day keeps the strategy permanently in the learning phase. Adjustments in steps of around ten to twenty per cent followed by one to two weeks of calm are common practice.