Every owner who has ever launched a campaign has asked some version of this question, usually around day nine, usually in a slightly tighter voice than they used on day one. The money is going out. Something is happening. Nothing is obviously working.
The honest answer is ninety days. That is the floor, not the target, and it is remarkably consistent across everyone who runs accounts for a living. But ninety days on its own is a useless answer, because it sounds like a stalling tactic and it does not tell you what to watch in the meantime. So here is the more useful version.
Nobody Actually Wants a Timeline
When an owner asks how long ads take, they are not asking for a number. They are asking two questions underneath it, and the two are very different.
The first is when am I allowed to stop worrying. That one has an early answer, and it is much earlier than ninety days. There are things you can see in week two that tell you the campaign is pointed at the right people, long before the cost per lead means anything.
The second is when am I allowed to quit. That one is the expensive question, and it is the one almost everybody gets wrong in both directions. Some owners quit at week three on a campaign that was four weeks from working. Others spend six months feeding a campaign whose arithmetic could never have worked, and nobody ever ran the arithmetic.
Both mistakes come from the same place: watching the wrong clock.
Three Clocks Are Running. Only One of Them Belongs to Google.
This is the thing that makes the whole subject confusing. There is not one timeline. There are three, they run at different speeds, and they are usually all running at once.
The platform clock. Google and Meta both need a stretch of spending before their bidding does anything intelligent. This is the one everybody has heard of and the only one that gets discussed.
The account clock. Separate from the algorithm, you need enough conversions to tell a real pattern from a fluke. A campaign can be fully out of its learning period and still not have given you enough data to make a decision you would defend.
The business clock. However long it takes one of your customers to actually decide. If you sell a $400 repair, that is same day. If you sell a $40,000 system to a committee, that is two quarters. Nothing in the ad platform knows or cares about this number, and it is usually the one that decides when revenue appears.
The reason this matters is that the three clocks fail differently. If you are early on the platform clock, the fix is patience. If you are early on the account clock, the fix is more volume or more time. If you are early on the business clock, there is no fix at all, because the deals are in flight and no amount of optimization pulls them forward.
Google Search Answers Fastest. It Still Takes a Quarter.
Search is the quickest channel to a real answer, because you are catching people who are already looking. Even so, the first stretch is slower and stranger than most owners expect.
Before anything serves at all, the campaign goes through review. Ads sit at "under review" or "pending" for a few hours, sometimes more than a day in regulated categories. Every owner who has not seen this before reads it as a broken campaign, and a fair number of them start changing things to fix a problem that does not exist.
Then the bidding starts learning. Google's own documentation puts this at up to three weeks, or one to two conversion cycles, and lists what restarts the clock: creating or reactivating a strategy, changing one of its settings, or adding and removing campaigns, ad groups or keywords. Notably, Google also says the algorithms keep learning after the "Learning" status disappears. The status going away is not a finish line.
What that looks like in a real account is worth seeing, because the pattern surprises people. One agency published a week-by-week cost per lead on a client campaign after a full rebuild. It went to $145, dropped to $60, jumped back to double the target, then to triple, then settled to $70, then $51, then $57, then into the low $40s, where it stabilized around $45 at a conversion rate near 40 percent. Roughly five weeks of genuine chaos before anything resembling calm.
That is one account and not a benchmark. But the shape is the point, and it is the shape almost every rebuilt campaign traces. Judged at week two, that campaign was a disaster. Judged at week ten, it was the best channel the business had.
One counter-intuitive signal from that same account: as it improved, impressions fell. That is correct and it is what you want. A campaign that is learning who your customer is not stops showing to everybody else. Falling impressions alongside a falling cost per lead is a campaign getting sharper. Falling impressions alongside a rising cost per lead is a campaign getting starved.
Meta Is Not Slower. It Is Hungrier.
Meta's clock works on a completely different principle, and this is where the cross-channel comparison earns its keep. Google's learning period is mostly a matter of time. Meta's is a matter of rate.
Meta's delivery system exits the learning phase after roughly 50 results in the week following an ad set's last significant edit. Note the shape of that: it is fifty per week, not fifty in total. An ad set that produces forty conversions a week never exits, no matter how long you leave it. Meta has a name for that state, "learning limited," and it is refreshingly blunt about what it means: your budget is not being spent effectively because the system cannot optimize with your current setup.
This is the single most important thing for a small advertiser to understand about Meta, and no timeline chart will tell you. If your budget and your offer cannot produce fifty conversions a week in one ad set, waiting does not fix it. More time is the wrong lever. The right levers are the ones Meta itself names: consolidate ad sets instead of splitting them, widen the audience, raise the budget, or optimize for a more frequent event further up the funnel, such as an add to cart or a lead rather than a purchase.
Meta's list of "significant edits" that restart the clock is also longer than most people realize: any change to targeting, any change to creative, any change to the optimization event, adding a new ad to the ad set, pausing it for seven days or more, or changing bid strategy. Budget changes count only if they are large. Meta's own example is that $100 to $101 does nothing, while $100 to $1,000 will likely reset it.
And Meta says the quiet part in its own help pages: during the learning phase, results are "not necessarily indicative of future performance." That is the platform telling you not to judge it yet.
LinkedIn is slower again, for reasons that have nothing to do with algorithms. LinkedIn's own B2B Institute research is built on the finding that around 95 percent of your potential buyers are out of market at any given moment. When the overwhelming majority of the people seeing your ad are not going to buy this quarter regardless of how good the ad is, a ninety-day judgment window is measuring a fraction of the effect. On LinkedIn you judge leading indicators early and revenue late, or you do not judge it at all. We have written separately on advertising to the 95 percent who are not ready, because it changes what the numbers are even for.
The Clock Nobody Puts on the Calendar Is Your Own Sales Cycle
Here is the one that gets missed, including by almost everybody who makes a video about ad timelines. Every one of them talks about the platform. None of them talks about your customer.
Google gives away the point without meaning to. Its stated learning period is "up to three weeks or one to two conversion cycles." That second half is Google admitting that its own clock is set by your sales cycle, not by the calendar.
Work it through. Say your average deal closes 45 days after first contact. You launch on the first of the month. In month three you finally look at revenue and compare it to spend. But the revenue landing in month three came from leads generated in month one, back when the campaign was at its worst and most expensive. You are reading a two-month-old campaign and calling it the current one. If the campaign has been improving the whole time, and it usually has, you are underrating it by exactly one sales cycle.
The same arithmetic runs the other way, and this half is good news. A locksmith or an emergency plumber closes within the hour. For them, lead volume and revenue are the same measurement, and they genuinely can read the account faster than a ninety-day rule implies. The rule is not that everybody waits a quarter. The rule is that you wait one full sales cycle past the point the platform settled down, and for some businesses that is two weeks.
What You Should Actually See at Two Weeks, Six Weeks, and Twelve
Waiting ninety days does not mean flying blind for ninety days. It means watching the right indicator at the right stage. Each of these is available well before the money can tell you anything, and each one is a leading signal that the campaign is pointed correctly.
At two weeks, you are checking aim, not performance. Are the ads actually serving, or still stuck in review. Does the search terms report read like your customers or like strangers, because that tells you within days whether the targeting is anywhere near right. Is conversion tracking firing, and can you prove it by submitting your own form and watching it land. Nothing about cost matters yet. If the search terms are wrong at two weeks, the campaign will not fix itself at twelve.
At six weeks, you are checking direction. Cost per lead should be moving downward on a trend line, even while individual weeks jump around. Conversion rate should be steadying. Impressions may well be falling as targeting tightens, and that is fine as long as the leads are getting cheaper. Ask your team what the last three changes were and why. If nothing changed in six weeks, nobody is optimizing. If everything changed every week, nobody let it learn.
At twelve weeks, you are checking the number. By now you should have a cost per lead you could put in a forecast, a rough sense of what fraction of those leads close, and therefore a cost per customer. That last figure is the entire point of the exercise. Everything before it was the price of finding it out.
Not Finished and Failed Look Identical From the Outside
This is the part that costs the most money, and it is the reason "just be patient" is bad advice on its own. Patience is only correct when there is something to be patient for.
Some campaigns are not slow. They are dead, and they were dead before launch, because the arithmetic never worked. The test is simple and you can run it in five minutes on the back of an envelope.
Take your realistic cost per click in your market. Take a conversion rate you could actually reach, and be honest, not aspirational: 5 to 10 percent is normal for a decent landing page. That gives you a cost per lead. Now take the share of leads you close, and you have a cost per customer. Set it against what a customer is actually worth to you.
If a campaign converts 30 percent of its clicks into leads at a $100 cost per lead, that is genuinely good work. And it is still a dead campaign if the service you are selling bills $250. No amount of optimization rescues that, because the gap is not in the account. It is in the price.
So the distinction is this. A campaign that is unfinished has a credible path to profitability and simply has not walked it yet. You can point at the specific waste you are about to cut and estimate what removing it does to the number. A campaign that has failed has no path at any conversion rate you could plausibly reach. Doubling the conversion rate still leaves it underwater.
The first one deserves the full ninety days and more. The second one should be stopped in week two, and the money moved to a channel where the math works. Being decisive about the second is what buys you the patience for the first.
Fund the Whole Runway or Do Not Start
Which leads to the decision that should be made before any money moves.
If ninety days is the honest minimum, then your entry price is not your monthly budget. It is your monthly budget times three, and you either have it or you do not. An owner who can fund six weeks is not buying a shorter test. They are buying a partial experiment that ends precisely at the point the data becomes worth reading, and they will conclude the channel does not work when what actually happened is that they left before it answered.
This is why we treat the first stretch of spend as tuition rather than an investment. You are not buying customers in month one. You are buying the number that tells you what a customer costs, and it is the number every other decision in your marketing hangs off.
It is also worth being blunt about the alternative. If the runway is not there, the right move is not a smaller, shorter test. It is to wait, or to spend the same money somewhere with a faster feedback loop, and come back when you can fund the full window. Some businesses genuinely should not be running ads yet, and running out of patience halfway through is one of the more expensive ways to find that out.
What This Is Really Buying You
Ninety days sounds like a long time to wait for an answer. It is not a wait, though. It is a purchase.
What you get at the end of it is not just a working campaign. It is a number: what it costs you to acquire one customer, and how reliable that number is. Once you have it, everything downstream gets easier. You know what you can afford to bid. You know whether to add a second channel. You know whether that competitor with the bigger budget is actually a threat or just louder. Every owner we work with who has that number makes calmer decisions than the ones who do not, and it is not because they are calmer people.
The patience is not a virtue. It is the price of the number.
We do this every day, and most of the value we add in the first three months is stopping people from panicking at week two and from persisting at week twenty when the arithmetic said no. If you would rather have someone watching the right clock while you run the business, that is what we are here for. If you would rather learn to read it yourself, everything above is how.
