When to start paid ads as an early-stage startup

Start paid ads once people who sign up keep coming back and you can track a sign-up from an ad click. Before that, ads mostly show you faster that the product or the page isn’t ready.

When you are ready, size the first test from your own numbers: the sign-ups you want, divided by your sign-up rate, times your cost per click. This post walks through the checks, the sizing, which platform to try first, how long to leave a test alone and how to decide what happens next. If you’re pricing a whole app marketing budget instead, see our guide to app marketing cost in 2026.

The short version

• Wait until people stay. In a GTMnow guide to paid media, Rex Gelb puts the line here: “If your week-4 retention is below 20%, ads will amplify a leaky bucket.”
• Track one sign-up event, from ad click to new account, before you spend anything.
• First test budget = target sign-ups ÷ sign-up rate × cost per click. Use your own numbers as soon as you have them.
• Leave the test alone while the platform learns. Meta says an ad set usually leaves its learning phase “after about 50 results in the week after the ad set’s last significant edit.” Google says to judge Smart Bidding over 2 conversion cycles or 2 weeks, whichever is longer.
• Scale only when the CAC payback works for your business. The CAC payback calculator does the sum.

Three checks before you spend

1. People who sign up keep coming back

Ads buy attention. They don’t make anyone stay. If most new users leave in the first few weeks, paid traffic brings in more people who leave, and you pay for each one.

Rex Gelb, in the 0 to 1 guide to paid media he wrote with GTMnow, says it plainly: “Ads are not the tool that creates product-market fit. They amplify it.” He also points out a second cost. Without people who stay, “the platform learns to find more users who look like your churning users.” His 20% week-4 line is one experienced buyer’s rule of thumb, not a study, but it’s a useful bar to check yourself against.

He allows one exception, a small test run to learn rather than to grow: “Small campaigns ($2K-5K) to recruit beta users, test messaging, or validate positioning are fine.” If that’s your goal, write it down before you start, and judge the test on what you learned, not on cost per customer.

2. A sign-up is tracked from the ad click

Before the first dollar goes out, click one of your own ad links, sign up with a fresh email and check that the sign-up shows up once, with the right source, in your analytics and in the ad platform.

This matters more than it looks. Ad platforms aim at whatever conversions you send back, so a missing or doubled event teaches them the wrong thing. Google also warns that “missing UTM parameters will result in (not set) values being present in reporting,” which leaves you guessing which ad did the work. Tag every ad link the same way with the free UTM builder. Our guide to early-stage startup metrics covers which events to log and how to save the source on each new user.

3. The landing page converts the people who arrive

Every paid click lands on a page. If the page doesn’t say what you do, or the form asks too much, you pay for clicks that leave. Unbounce’s 2024 report puts the median SaaS landing page at 3.8%. Know your own rate before you buy traffic, because it sets the size of your test. Our signup conversion rate guide has the formula and a 30-minute check.

How to size a first test

Here is the formula:

First test budget = target sign-ups ÷ sign-up rate × cost per click

• Target sign-ups. Enough to judge. Pick the number before you start. 50 is a sensible default, because that’s roughly where Meta says an ad set usually leaves its learning phase in a week.
• Sign-up rate. The share of ad clicks that become a new account. Use your landing page’s own rate if you have it.
• Cost per click. Your own, once you’ve run any ads. If you haven’t, use a published median as a placeholder.

Here is a worked example. It multiplies published medians, so treat it as an example, not a budget. WordStream’s 2025 Google Ads benchmarks give Business Services a median cost per click of $5.58 and a conversion rate of 5.14%.

• 50 sign-ups ÷ 5.14% = about 973 clicks.
• 973 clicks × $5.58 = about $5,430.
• Over a two-week test, that’s about $388 a day.

Now swap in a page that converts at Unbounce’s 3.8% SaaS median. 50 ÷ 3.8% is about 1,316 clicks, and 1,316 × $5.58 is about $7,340. The page alone moves the test by roughly $1,900, which is why check 3 comes before any spend.

WordStream’s conversion rate covers conversions of any kind across US search campaigns, not only sign-ups, and it has no software or SaaS row. So once you have a week of your own clicks, rerun the formula with your own cost per click and sign-up rate, and throw the placeholder away.

If the number is far more than you can spend, lower the target and accept a rougher read, or narrow the test to the few searches or audiences closest to buying. Don’t spread a small budget thinner.

The published figures

Last checked on 25 October 2026.

WordStream also found that cost per click went up for 87% of industries year on year, so expect your own number to drift and recheck it every month.

Which platform first

Pick one platform for the first test, so the whole budget sits behind one result.

• Google Search if people already search for the problem you solve. You’re catching demand that exists. Gelb’s advice is to “always start with Search ads, not display or YouTube.”
• LinkedIn if your buyer is defined by job title, seniority or company. He calls it “usually the first paid channel B2B teams test,” because the targeting is built for professional audiences.
• Meta if your buyer is defined by interests or habits rather than job title, and your budget can buy around 50 results a week.

If you can’t say which fits, start with Search on the few phrases a buyer would type just before signing up.

Don’t judge the test in its first week

Both big platforms tell you to wait.

Meta says that during the learning phase “performance is less stable,” and that editing an ad, ad set or campaign during it will “reset learning.” It also asks you to avoid frequent budget changes, which can push an ad set back into learning.

Google says that while you may see a “Learning” status “for up to 7 days, your ads continue to run and serve as normal.” After that, its advice is to “evaluate performance over 2 conversion cycles or a 2 week window (whichever is longer).”

So set the run length before you launch, write down the one number you’ll judge it on, and don’t touch it until the window ends. Gelb’s advice for teams that are further along is the same in spirit: “Resist the urge to quit after four weeks.”

When to stop, change one thing, or scale

At the end of the window, look at where people dropped out.

• Clicks, but almost no sign-ups. The ad promised something the page doesn’t show, or the page itself leaks. Change one thing, the page or the ad message, and run it again.
• Sign-ups, but nobody comes back or pays. The ads are finding the wrong people, or the product isn’t ready for them. Stop the spend and fix that first.
• Sign-ups, and some of them pay. Now work out your CAC payback: how many months a new customer takes to earn back what they cost to win.

CAC is all the spend that won those customers, divided by the number of new customers. Payback is CAC divided by monthly revenue per customer times gross margin. Put your numbers into the CAC payback calculator. It runs in your browser and compares your result with a sourced benchmark.

If the payback fits what your cash can carry, raise the budget in steps and give each step time to settle. If it doesn’t, change one thing and test again before you spend more.

Mistakes to avoid

• Starting ads before people stay, then deciding paid doesn’t work.
• Judging a campaign in its first few days, or editing it while it’s still learning.
• Splitting a small budget across three platforms, so none gets enough results to learn from.
• Counting clicks or leads as the result, instead of sign-ups and paying customers.
• Keeping the benchmark cost per click after you have your own.
• Sending ads to untagged links, so the sign-ups land in your reports as “(not set)”.

FAQ

When should a startup start running paid ads?

Once people who sign up keep coming back and you can track a sign-up from an ad click to a new account. Rex Gelb’s rule of thumb in GTMnow is that week-4 retention below 20% means ads will amplify a leaky bucket. A small test to learn about messaging can come earlier.

How much should a startup spend on its first ad test?

Use target sign-ups ÷ sign-up rate × cost per click. With WordStream’s 2025 Business Services medians, 50 sign-ups at a 5.14% conversion rate and $5.58 a click comes to about $5,430. That’s an example, not a budget, so rerun it with your own numbers.

Should a pre-seed startup run Google Ads or Meta ads first?

Google Search if people already search for what you solve. Meta if your buyers are defined by interests rather than searches and your budget can buy around 50 results a week. For B2B buyers defined by job title, LinkedIn is the usual first test.

How long should I run an ad test before judging it?

Meta says an ad set usually leaves its learning phase after about 50 results in the week after its last significant edit. Google says to judge Smart Bidding over 2 conversion cycles or 2 weeks, whichever is longer. Set the window before you launch and don’t edit during it.

Why do my ads get clicks but no sign-ups?

Usually the ad promises something the page doesn’t show, or the page itself leaks, with an unclear headline or a long form. Check that the sign-up event fires, then compare your page’s rate with Unbounce’s 3.8% SaaS median and fix the page before you buy more clicks.

Want a channel that isn’t priced per click?

Ignura is an AI-native agency that automates your GEO and SEO. You approve what goes out. We don’t run ads. We work on the pages people find when they search on Google and in AI answers: AI search, technical SEO and content. Plans are on the pricing page.

Book your free Growth Check