New market: the campaign that learns comes before the campaign that sells
Automated bidding optimizes from a history of conversions. In a new market that history does not exist, and Google documents volume thresholds that a test budget rarely reaches. The setting that looks technical is in fact a budget constraint, and it decides what your campaign is able to learn.
- Google states that you generally need at least 15 conversions in 30 days, and recommends aiming for 30 per month per ad group to hold a target consistently.
- The initial learning period requires at least two weeks without changes. Every change of target, strategy or conversion action restarts it.
- A test budget that is too small produces a vicious circle: too few conversions to exit learning, so unstable results, so adjustments that restart learning.
- The sequence that works uses two campaigns separated in time: one that learns with no target constraint, one that then optimizes on a target drawn from reality.
- Google notes that the algorithm also draws on aggregated data and on the advertiser domain's history. You are not starting from nothing, but you are not starting from your home market either.
On this page
Useful conversion threshold
The useful conversion threshold is the monthly number of conversions below which an automated bidding strategy does not hold enough data to reach a target reliably. It is not a technical block: the campaign still delivers. It is a statistical threshold below which results become too unstable for a management team to draw a decision from them. It therefore sets the minimum budget of a market test, before any consideration of reach.
What Google documents, and what it implies
The documentation is more precise than its reputation suggests. Google states that you generally need at least 15 conversions over a 30-day period for a target cost per acquisition strategy to work, while noting that some advertisers with relevant historical data may qualify sooner. For a business that wants to hit its target consistently over 30 days, Google recommends aiming for at least 30 conversions per month per ad group.
The initial learning period requires at least two weeks without changes, and the documentation warns that performance can be more unstable during that phase. On the display network, Google refers to a learning window of two to four weeks.
One point deserves keeping because it qualifies everything else. Google explains that the algorithm also learns from aggregated data and from historical performance at the advertiser domain level. So you are not starting completely blind in a new market. But most of what the system knows about you comes from your home market, where buying behaviour, competition and pricing are not the same.
Finally, the documentation adds a condition that test plans almost always ignore: automated bidding works better with a budget that is not limited by spend. A throttled campaign does not compete in the auctions most likely to produce a conversion, which degrades precisely what it is supposed to learn.
The vicious circle of a small budget
It closes in three movements, and a management team that does not know it blames the platform or the market.
First movement: the test budget is set at an amount that looks reasonable, often a few thousand dollars a month. That amount produces a number of conversions below the threshold, because the cost per enquiry in a market where your brand means nothing is higher than elsewhere.
Second movement: results are unstable, exactly as the documentation announces. One week looks good, the next looks bad, and no trend emerges.
Third movement: somebody adjusts. The target gets lowered, the strategy gets changed, the primary conversion action gets modified, an ad group gets cut. Each of those moves restarts learning. The campaign heads into another two weeks of instability, and the cycle begins again.
Three months later the business has spent its test budget, never left learning, and concludes that the market does not respond. That is the most expensive conclusion available, because it is false and it closes a territory for years.
What to settle before opening the campaign
The question is not how much you want to invest, it is whether the amount you are willing to commit can produce an answer. Below a certain threshold, a test does not cost less, it costs the same for no information.
- Can our monthly budget produce at least 15 conversions in this market, at the estimated cost per enquiry?
- If not, are we willing to wait and concentrate the budget on a single market rather than two?
- Who has authority to modify the campaign, and does that person know an adjustment restarts learning?
- Which conversion action are we optimizing on, and does it correspond to a real commercial opportunity?
- On what date do we read the results, explicitly excluding the first two weeks?
A solid answer sets a reading date and forbids changes before it. An answer that promises weekly adjustments is promising permanent learning.
Calculating whether a test budget can produce an answer, before committing it, is part of what we cover in a paid audit.
Two campaigns, not one
The sequence that works separates in time two objectives that almost always get conflated.
The campaign that learns
One objective: accumulate conversions and discover what the market searches for. No numeric target throttling delivery, an unconstrained budget, and above all no changes during the learning window. You do not judge its return, you judge what it learned.
The reading
Once the threshold is reached, you read what the campaign produced: which queries convert, at what real cost, from which regions and which profiles. That is the only honest basis for setting a target.
The campaign that sells
You then set a target drawn from observed cost, not from desired cost. Google recommends anchoring the initial target on real historical performance, accounting for the delay between click and conversion.
The tightening
The target then tightens in small steps, with enough time between moves for learning to settle. A brutal tightening sends the campaign back to the starting line.
That sequence has a cost, the cost of the first phase, and it is exactly what many businesses refuse to name in their plan. They budget a selling campaign and discover they paid for a learning campaign. The difference between the two approaches is not the amount spent, it is knowing which of the two you are buying.
A campaign that never leaves learning is not badly optimized. It is underfunded, and no optimization repairs a volume problem.
Falia analysis gridCalculating the budget backwards
The calculation starts from the threshold, not from the available envelope. It is counterintuitive and it is the only way to get a defensible figure.
Start from 15 monthly conversions as a floor, or 30 if you want a target held consistently. Multiply by the cost per conversion you estimate in that market. That estimate comes from your current cost marked up, because a market where your brand is unknown converts less well: the click costs the same and it takes more clicks to get an enquiry.
The result is your minimum monthly budget. If it exceeds what you are willing to commit, you have three choices, and you have to take one. Shrink the territory, aiming at a region rather than a country. Reduce the number of campaigns, concentrating on one market instead of two. Or postpone, until you have the budget that allows an answer.
The fourth choice, launching anyway on an insufficient budget, is the only one that guarantees spending without learning.
The discipline most often missing is a schedule of changes. Write down, before launch, the date on which the campaign will be read and the list of moves allowed until then. That list fits on one line: none. Well-intentioned adjustments made mid-learning are the leading cause of inconclusive tests, and they almost always come from a good person looking at the numbers every morning meaning to help.
Before launching the entry campaign
The geographic setting that makes the test readable is detailed in your test market campaign is measuring people who are not in the market, and the platform choice in the second engine as a market's first test. The overall sequence is in your export plan does not say where the customers will come from, the profitability calculation in the break-even threshold of a campaign, and the effect of changes at peak in the Black Friday and holiday calendar. When to cut is covered in recognizing a market that will not take.
Working the budget back from the number of conversions required is detailed in the threshold below which a market test proves nothing.
Opening a market is the subject of the Develop a new market goal.
The lever for this work is paid advertising.
Frequently asked questions about entering a market through advertising
How many conversions does automated bidding need?
Google states that you generally need at least 15 conversions in 30 days, and recommends aiming for at least 30 per month per ad group to hold a target consistently. Some advertisers with relevant historical data may qualify sooner.
How long does the learning period last?
At least two weeks without changes for the initial phase, according to Google's documentation, with a window that can reach four weeks on some networks. Performance is more unstable during that period.
What restarts learning?
A change of bidding strategy, a change of target, a change of primary conversion action, and large budget swings. Each of those moves returns the campaign to an unstable phase.
Can you launch a campaign with no conversion history at all?
Yes. Google explains that the algorithm also draws on aggregated data and on the advertiser domain's history. You are not starting from nothing, but what the system knows about you comes mostly from your home market.
What if our budget cannot reach the threshold?
Shrink the target territory, concentrate on one market instead of two, or postpone. Launching on an insufficient budget spends the same amount without producing a usable answer.
Should the campaign be adjusted during learning?
No. Every adjustment restarts the phase. The discipline is to write the reading date before launch and touch nothing until then, even when one week's numbers look bad.
- Google Ads, Set up Smart Bidding, help centre, consulted August 2026. The threshold of 15 conversions in 30 days, the recommendation of 30 conversions per month per ad group, the two-week minimum for the learning period, and the effect of a budget limited by spend.
- Google, Display Smart Bidding Guide, official document, consulted August 2026. Learning from aggregated data and from advertiser domain level history, learning window of two to four weeks.
- Google Ads, About advanced location options, help centre, consulted August 2026.
Gabriel almost always takes your first call and carries out your audit. He builds the strategy starting from your growth goal: where to put your budget, which market to test and how to connect each lead to a real sale in your CRM. He mainly leads engagements for three goals: Optimize the profitability of your digital campaigns, Develop a new market, and Generate demand and growth. With Geneviève, he also works on organic search (SEO), AI visibility (GEO) and conversion rate optimization (CRO). The sales a Google Ads or Meta Ads campaign brings in depend on the page that receives the click. He writes mainly about marketing strategy, paid advertising and measurement.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design