Maximize Conversions

Maximize Conversions bids automatically at the auction level using signals Google holds about the searcher or viewer - device, location, time of day, browser, operating system, and remarketing list membership among others - to predict how likely a given auction is to convert. Where Maximize Clicks bids toward cheap volume, this strategy bids toward conversion likelihood, so a click that looks unlikely to convert can get a low bid or none at all even if it would have been cheap to win. As with other automated strategies, there is no manual keyword-level bid to manage once it is active; the algorithm recalculates a bid for every auction in real time.

Left with no target, the strategy is built to spend the full daily budget, so budget becomes the primary lever controlling volume. Once conversion data is steady, adding a target CPA gives you a second lever: a cost ceiling the algorithm tries to hold to on average, at the cost of some volume if the target is aggressive. Which conversion actions actually count matters just as much as the bidding logic itself - in the Conversions settings, actions are marked primary or secondary for bidding purposes, and if that is set up loosely, the algorithm can end up optimizing toward a blend of actions that do not have equal business value, some of which may barely matter.

This strategy is the right fit for accounts with steady, accurately tracked conversions - lead-gen and ecommerce accounts that have already validated their tracking, often after a data-gathering phase on Maximize Clicks or plain Manual CPC. It is a poor fit for brand-new accounts with little or no conversion history, since the algorithm has no signal to calibrate against and can behave erratically, sometimes gravitating toward whatever cheap action is easiest to rack up rather than what actually matters to the business.

The recurring failure mode is turning this on while conversion tracking is broken or pointed at the wrong action - a pixel firing on every page load, a form-submit event that fires before validation, or a conversion action counting something trivial like a newsletter signup as if it were a sale lead. None of that stops the algorithm from working; it just trains the system to buy more of whatever it is being told is a conversion, which can mean spend climbs while the metrics that actually matter to the business stay flat or worsen. A second common mistake is adding a target CPA before the account has stabilized, which resets the learning period and produces a stretch of noisy, hard-to-interpret results right when you are trying to judge whether the change worked.

In reporting, the columns to watch are Conversions and Cost/conv, but it is worth checking Impression share lost to budget alongside them - if that number is high, the algorithm is signaling it could deliver more conversions at the current cost if the budget allowed it. Also watch for discontinuities that are not really performance changes: if someone adds a new action to the primary conversion set partway through the month, cost/conv will jump or drop sharply on that reporting date even though nothing about market cost actually shifted - the denominator changed, not the underlying efficiency.