What is the learning phase, and what does leaving it actually tell you?

Both platforms name the period and warn you off reading what happens inside it. Meta puts it in the Delivery column; Google says the Status column on the Bid strategies page under Tools shows your portfolio bid strategy's status. Meta says the Delivery column reads 'Learning' while an ad set is in it, that ad sets are less stable then and usually have a higher CPA, and that your results during it are not necessarily indicative of future performance. Google Ads says a 'Learning' bid strategy status may be shown after a change, that key metrics may vary, and that you may not want to measure performance until the learning period is over. Google says bid strategy statuses are not available for Hotel campaigns at all. So reading a two-day slice of a period the platform has told you is unstable is not impatience, it is reading a number the platform has already said may not be worth measuring yet.

How long it lasts is where the documentation is worth reading rather than remembering, because the two platforms count in different units and Google's duration page gives two of its own. Meta counts results: usually about 50 in the week after the last significant edit, with its own exception for Shops ads, which need 17 purchases through the website and 5 through Meta. Google's duration page publishes no days figure at all: it gives up to around 50 conversion events or 3 conversion cycles for the bid strategy to calibrate, faster with more conversion data, and says the learning period does not apply to manual CPC. A conversion cycle is time, not a count, and the example Google gives for one, on the same page, is how long a click takes to become a conversion. A different Google page gives the figure most people quote, a standard 7 to 14 day learning period, in the course of saying that frequent manual adjustments to budgets, to CPA or ROAS targets, or to conversion goals will reset it.

The practical trap is that significant edits restart the clock, and Meta publishes what counts as one rather than leaving it to folklore: any change to targeting, to ad creative or to the optimisation event, adding a new ad to the ad set, pausing it for seven days or longer, and changing bid strategy. A change to an ad set spending limit, to a bid control, cost per result goal or ROAS goal amount, or to a budget may or may not count, depending on the size of the change, and Meta's own example is that going from 100 US dollars to 101 is unlikely to reset anything while going from 100 to 1,000 may. An account that is adjusted every couple of days can keep several ad sets permanently in learning, which produces exactly the unstable numbers that prompt the next adjustment.

Learning limited is the opposite problem, and it is a Meta status rather than a Google one. Meta says an ad set becomes learning limited when it is unlikely to receive about 50 optimisation events in the week after your last significant edit, that the Delivery column reads Learning limited, and that it is not a penalty but an indication that the budget is not being spent effectively because the delivery system cannot optimise with the current setup. It also publishes the causes, and they are mostly about the setup rather than the ads: a small audience, a low budget, a low bid or cost control, high auction overlap, an infrequent optimisation event, or too many ads running at once. Its own first fix is combining ad sets and campaigns, and one of the others is choosing an optimisation event that happens more often, such as moving from purchases to add to basket.

The error worth naming is treating the end of learning as the verdict, because it looks like patience. People correctly wait for learning to finish before judging, and then treat the finish itself as the judgement. Exiting learning means the numbers have become readable. Whether they are good is a completely separate question, answered against your cost ceiling, and the answer is sometimes no.