How much should I spend to test whether ads work for my app?

Start from what a customer is worth

Before choosing a budget, work out the most you could pay for one customer and still be fine. A rough version: what an average customer pays you a month, times the number of months they usually stay. If they pay $20 a month and stay about eight months, a customer is worth around $160, and spending $80 to win one leaves room for everything else.

That number, your maximum cost per customer, is what the test measures against.

Size the test to buy a few customers

A test that can’t buy a single customer at your maximum price can’t tell you anything. Size it to buy three to five:

• Maximum cost per customer: $80 (from the example above).
• Budget: $80 × 4 = $320.
• Time: two to four weeks, so the people who click on day one have time to try and pay.

Write the budget, the dates and the stopping rule down before the first ad runs. Deciding afterwards makes every result look promising.

Count customers, not clicks

The ad platform shows clicks, cost per click and its own idea of conversions. None of those is the number you need. Tag each ad’s link with its own campaign (?utm_source=reddit&utm_medium=paid&utm_campaign=ad-1), keep that source on every signup, and match payments to signups by email or user id. Then:

• Cost per customer = what the ad cost ÷ paying customers it brought.
• Return on spend = revenue from those customers ÷ what the ad cost.

How to read the result

• Customers at or under your maximum cost: keep the ad, raise the budget in steps, and watch whether the cost holds.
• Signups but no customers: the ad finds interested people, but not buyers. Change who it targets before you change the page.
• Clicks but no signups: the ad and the landing page promise different things. Make them say the same thing.
• Nothing: stop on the date you set. Try a different channel before a different ad.

One ad at a time

Change one thing per test: the audience, the message or the page. Several changes at once leave you with a result and no idea what caused it.

With Nordva Launch

Every visit and signup keeps its UTM tags, and with Stripe connected (read-only, every plan) paying customers are traced to their first visit. On Indie, enter what each channel cost over a period and Launch shows cost per customer, return on spend and LTV:CAC for each referring site. Grouping paying customers by UTM campaign, to compare single ads, needs Builder.