How many conversions do you need before you can judge whether Google Ads is working?

A conversion is the action you want someone to take after clicking your ad: buying, booking a demo, starting a free trial, subscribing to a newsletter, or filling in a contact form. You decide what counts as a conversion when you set up conversion tracking in Google Ads.

Your conversion rate is conversions divided by clicks. If 200 people click your ad and 10 of them sign up, your conversion rate is 10 ÷ 200 = 5%.

Your cost per conversion (often called cost per acquisition or CPA) is what you spent divided by the conversions you got. If those 200 clicks cost $400, your cost per conversion is $400 ÷ 10 = $40.

These are the numbers founders use to decide whether Google Ads is "working." The problem is that with small numbers of conversions, they bounce around a lot from pure chance. That's why the question of how many conversions you need matters so much.

It depends on which question you're asking. Most founders are actually asking one of four different questions, and each needs a different amount of data:

The question you're asking	What it takes	Typical amount of data
Is this particular keyword a dud?	A spending rule, not a conversion count	Spend 3× your target cost per conversion with zero conversions
What is our real conversion rate or cost per conversion?	Enough conversions to narrow the range of uncertainty	About 10 to 64 conversions, depending on how precise you need to be (30 is the common default)
Is ad A (or keyword A, or landing page A) better than B?	A proper comparison test	Roughly 22 to 400+ conversions per version
Can I switch on Google's or Microsoft's automated bidding?	The platform's own minimums	15 to 50 conversions in 30 days, depending on the strategy

The rest of this FAQ explains where each of those numbers comes from and how to adjust them for your business.

It comes from the ad platforms' own guidance on evaluating automated bid strategies:

• Google Ads, Target CPA bidding: Google says advertisers "can start using Target CPA with no conversion history," but for evaluation it recommends you "measure performance for the last 30 days, including at least 30 conversions." (Google Ads Help: About Target CPA bidding)
• Microsoft Advertising: for Target CPA, Maximize Conversions and Target ROAS, Microsoft recommends "at least 30 conversions before evaluating" performance, after a learning period of up to two weeks, and then another two to four weeks of running before you judge. (Microsoft Advertising: Let Microsoft Advertising manage your bids)

So 30 is a real, platform-backed number, but its original purpose is narrow: deciding whether an automated bidding strategy is performing. It gets repeated as a general rule of thumb for everything, which is where founders get misled.

There's also a statistical reason 30 is a sensible default. As the table in "How do I calculate how many conversions I need?" shows, 30 conversions is roughly what you need to know your true conversion rate to within plus-or-minus 30%, with 90% confidence. That's a reasonable standard for many decisions, but not for all of them.

Because "enough data" depends on how costly it would be to get the answer wrong, and that varies by business and by decision.

Conversion-rate specialists make the same point about website tests. CXL, a well-known conversion optimization training company, puts it bluntly: "there is no magical traffic or conversion number," and a fixed count like "100 conversions per variation" can still produce a false result if the sample isn't representative. (CXL: Stopping A/B Tests: How Many Conversions Do I Need?)

Four things move the right number up or down:

• The decision. Spotting a clearly broken keyword takes far less data than proving one ad beats another by 10%.
• How precise you need to be. Knowing your cost per conversion is "somewhere between $30 and $50" takes much less data than knowing it's "between $38 and $42."
• How sure you need to be. Being 80% sure takes less data than being 95% sure.
• How fast your customers convert. If people take two weeks to sign up after clicking, you can't judge last week's clicks yet, no matter how many there were.

Every conversion rate you measure is an estimate. If your true conversion rate is 5%, you won't see exactly 5% in any given month. Some months, by pure luck, you'll see 3%; other months 7%.

Precision is how close your estimate needs to be to the truth. "Plus-or-minus 30%" means that if you measure 5%, the true rate is somewhere between 3.5% and 6.5%.

Confidence is how sure you want to be that the truth really is inside that range. "90% confidence" means that if you repeated the measurement many times, the range would contain the true rate about 9 times out of 10.

The key idea: more conversions make the range narrower, but they don't change the rate itself. With 5 conversions, your 5% might really be anywhere from about 1% to 9%, which is too wide to act on. With 30 conversions the range tightens to roughly 3.5% to 6.5%. The conversion rate stays around 5% either way; what shrinks is the spread of plausible values around it.

This is the same reason a poll of 1,000 people is more trustworthy than a poll of 50. The textbook treatment is under "margin of error" and "sample size for a proportion." (Wikipedia: Sample size determination; Penn State STAT 200: Computing necessary sample size)

For a single campaign-level read ("what's our real conversion rate?"), a standard planning formula is:

Conversions needed ≈ z² × (1 − conversion rate) ÷ precision²

• z comes from your confidence level: 1.28 for 80% confidence, 1.645 for 90%, and 1.96 for 95%.
• Precision is written as a decimal: plus-or-minus 30% = 0.30.
• Conversion rate is your expected rate as a decimal (5% = 0.05). At typical ad conversion rates this term barely matters, which is why the answer is mostly about precision and confidence.

This is the standard sample-size formula for estimating a proportion, rearranged so the margin of error is a share of the rate rather than a fixed number of percentage points. It's an approximation for planning, not an exact guarantee.

Conversions needed at a 5% conversion rate:

Confidence	±20%	±30%	±40%	±50%
80%	39	17	10	6
90%	64	29	16	10
95%	91	41	23	15

The bold cell is where the platforms' "30 conversions" sits: plus-or-minus 30% at 90% confidence.

How to choose:

• First test ("does paid search work for us at all?"): plus-or-minus 50% at 90% confidence, about 10 conversions, is defensible. You're looking for a clear yes or no, not a precise figure.
• Deciding whether to scale spend: plus-or-minus 30% at 90%, about 30 conversions.
• Setting budgets or bid targets you'll live with for months: plus-or-minus 20% at 90 to 95%, about 64 to 91 conversions.

Once you know how many conversions you need, two more steps turn that into clicks and dollars:

Clicks needed = conversions needed ÷ conversion rate
Budget needed = clicks needed × average cost per click

Worked example: a newsletter business expects about 5% of ad clicks to subscribe, and pays about $1.90 per click.

Decision standard	Conversions	Clicks	Ad spend
±30%, 90% confidence (default)	29	~580	~$1,100
±50%, 90% confidence (first test)	10	~200	~$390
±50%, 80% confidence	6	~125	~$240

If you want the answer within 30 days, that's your monthly budget. If you can wait 60 days, halve it, but you'll wait twice as long for the answer.

Two cautions. First, use your own landing-page conversion rate if you have it, not an industry average and not a different funnel step (for example, a free-to-paid upgrade rate). Second, if you don't know your cost per click yet, don't guess low. A low guess makes the budget look smaller than it really is and loses ad auctions. Use the median cost per click of related keywords from a keyword tool such as Google Keyword Planner.

Long enough to reach your conversion target and long enough for the data to settle. Google's guidance points to three timing factors:

• Conversion delay (lag). The time between a click and the conversion it leads to. Google recommends waiting at least one conversion cycle before judging a change, and one to two cycles after significant changes "to ensure conversions are complete due to conversion delay." For a business with a two-day delay, Google's example is to look at the last 28 days while leaving out the most recent two. (Google Ads Help: How our bidding algorithms learn)
• Learning period. When you start or change an automated bid strategy, Google says "it can take up to around 50 conversion events or 3 conversion cycles for the bid strategy to calibrate." (Google Ads Help: Duration of the learning period)
• A full business cycle. Include every day of the week, and ideally a full month, so one unusual weekend doesn't skew your read. CXL recommends at least one to two business cycles for tests. (CXL)

A simple way to estimate the time to a decision: days ≈ conversions needed ÷ conversions per day + conversion lag. If you're getting 10 conversions a month (about 0.33 a day) and need 29, with a one-day lag, that's roughly 29 ÷ 0.33 + 1 ≈ 90 days.

Use a spending rule: if a keyword has spent three times your target cost per conversion with zero conversions, it's almost certainly underperforming.

Here's why. Suppose your target cost per conversion is $40. If a keyword were truly hitting that target, you'd expect about three conversions by the time it had spent $120. The chance of seeing zero conversions when three were expected is only about 5%. So pausing at that point is a decision made with about 95% confidence. At 2× target spend, the chance of zero conversions by bad luck is about 13.5%, which is roughly 86% confidence.

This matches a classic statistical shortcut called the "rule of three": if something hasn't happened in n tries, you can be about 95% confident its true rate is below 3 ÷ n. (Statology: A concise guide to the statistical rule of three) It also matches practitioner advice from Optmyzr, a PPC software company: flag a keyword that spends 2 to 3× your target cost per conversion without converting, and extend that to 4 to 5× for higher-ticket products or long sales cycles. (Optmyzr: Google Ads keywords not converting?)

To use this rule, you need a target cost per conversion. A simple way to set one: what a new customer or subscriber is worth to you, times the share of that value you're willing to spend to acquire them.

Much more than for a single read, because you're comparing two uncertain numbers instead of one. The amount depends heavily on how big the real difference is.

Conversions needed per version (5% baseline conversion rate, 95% confidence, 80% statistical power, the usual defaults for comparison tests):

How much better the winner really is	Conversions per version (approx.)
20% better	~410
30% better	~190
50% better	~75
Twice as good	~22

For most small businesses, that means only big differences can be detected reliably. Test bold changes, not tweaks. To size your own test, use a free calculator such as Evan Miller's A/B test sample size calculator. Google Ads' own experiments feature reports results using 95% confidence intervals. (Google Ads Help: The statistical methodology behind experiments)

Google's requirements differ by bid strategy:

• Target CPA: no minimum to start. Google recommends evaluating on the last 30 days with at least 30 conversions. (Google Ads Help: About Target CPA bidding)
• Target ROAS (return on ad spend): Search and Shopping campaigns need at least 15 conversions in the past 30 days. Google also recommends reporting conversion values for four weeks or one to two conversion cycles, whichever is longer, before turning it on, and leaving the most recent conversion-delay period out of your evaluation. (Google Ads Help: About Target ROAS bidding)
• Learning period: up to about 50 conversions or 3 conversion cycles to calibrate after a change. (Google Ads Help: Duration of the learning period)

If your campaign produces only a handful of conversions a month, manual bidding with clear rules, like the keyword spending rule above, is often easier to control until you have more data.

Similar. Microsoft Advertising recommends at least 30 conversions before evaluating Target CPA, Maximize Conversions or Target ROAS; allows a learning period of up to two weeks; and suggests running two to four more weeks after that before judging. Microsoft also notes that when conversions happen more than seven days after the click, it's harder for its automated bidding to bid effectively. (Microsoft Advertising: Let Microsoft Advertising manage your bids with bid strategies)

You have four honest options, and you can combine them:

1. Loosen the standard for a first test. Plus-or-minus 50% at 90% confidence needs about 10 conversions instead of about 30. That's enough to answer "is this clearly working or clearly not?"
2. Extend the window. Judge over 60 or 90 days instead of 30. Same total spend, lower monthly budget, slower answer.
3. Concentrate the budget. Run fewer keywords so each one gets enough clicks to learn from, instead of spreading a small budget thinly across many.
4. Measure an earlier step. If sales are rare, also track a more frequent action (such as a newsletter signup or demo request) as a leading indicator, and be clear it's a proxy.

What you shouldn't do is declare Google Ads "doesn't work" after a handful of conversions. At that volume, a good campaign and a bad one can look identical.

Input	Sensible default	Why it varies by business
Expected conversion rate	Your own landing-page rate	Offer, price, landing page and search intent all change it
Target cost per conversion	Value of a customer × the share you'll pay to acquire one	Depends on your margins and customer lifetime value
Precision needed	Plus-or-minus 30%	How big a bet the decision is
Confidence needed	90%	Your tolerance for being wrong
Conversion lag	1 day	Impulse purchases convert in minutes; B2B deals can take weeks
Decision window	30 days, or 2 conversion cycles if longer	Matches Google's and Microsoft's evaluation guidance
Average cost per click	Median of related keywords	Competition in your category sets the price

• Is paid search working at all? About 10 conversions (±50%, 90% confidence).
• Should we scale spend? About 30 conversions (±30%, 90% confidence). This is the platforms' "30."
• Should we lock in budgets and targets? About 64 to 91 conversions (±20%, 90 to 95% confidence).
• Is this keyword a dud? 3× target cost per conversion spent with zero conversions (about 95% confidence), or 4 to 5× for long sales cycles.
• Is A better than B? About 75 conversions per version to detect a 50% difference, and about 190 to detect a 30% difference.
• Clicks needed = conversions needed ÷ conversion rate. Budget needed = clicks × cost per click.
• Always leave out the most recent conversion-lag days, and include full weeks.