How long should I test a creative before I judge it?

Long enough for enough signal to accumulate – not by gut feeling. As a rough guide: sufficient impressions and ideally several purchases before you decide. Switching off early after a few hours isn't testing, it's guessing. An attribution window that's too tight can make winners falsely look like losers.

What's the difference between exploratory and iterative testing?

Exploratory testing looks for new angles and messages – this is where real breakthroughs happen. Iterative testing takes a winner and varies it (hook, edit, format) to extend and exhaust it. You need both: exploration for leaps, iteration for stability.

Testing · Guide

Why Systematic Testing Decides Your Success

Without structure, creative testing is burning money. How to build a testing agenda that reliably finds winners – exploratory vs. iterative testing.

By Denys Lichtenstein Prefer us on Google

Cover image: Why Systematic Testing Decides Your Success

Most accounts don’t test – they tinker. They switch on a few creatives, look at them after two days and turn off whatever happens to look bad. That’s not testing, that’s noise. Systematic testing is the difference between “we once had a good creative” and “we reliably find new winners”.

Two kinds of tests

Exploration – new angles, messages, formats. This is where the leaps happen. Most exploration creatives lose; the few winners carry the account.
Iteration – a winner is varied: a different hook, a different edit, a different format. That’s how you extend its lifespan and exhaust an angle.

Whoever only iterates optimises themselves into a dead end. Whoever only explores leaves winners on the table. You need both – deliberately managed.

The testing agenda

Instead of uploading creatives at random, you work with a prioritised list of hypotheses: “We believe angle X works for audience Y, because Z.” Each week, the strongest hypotheses move into the test. This makes testing plannable and the results readable – you know why something won, not just that it did.

A clean setup: one variable at a time

If you test hook and format and audience at the same time, in the end you don’t know what worked. Isolate the variable you want to learn about. And give the test enough budget and time, so that the numbers are meaningful – a result based on three purchases is a gut feeling, not a decision.

Clear rules for winners and losers

Define in advance what a winner is (e.g. target CPA reached with sufficient signal) and what a loser is. Then act on it – without ego:

Scale and iterate on winners.
Stop losers cleanly.
Retest borderline cases rather than letting them run endlessly.

Rhythm beats frantic action

A weekly loop – new tests in, results evaluated, winners scaled – is more powerful than hectic intervention in between. Consistency is the real framework.

The foundation: data

Testing is only as honest as your measurement. A misattributed setup declares winners to be losers. That’s why clean tracking and the view via MER are an essential part of it – more on this under Data & Scaling and in the article MER instead of ROAS.

We sharpened exactly this framework on our own brand and applied it across more than 100 accounts. A free account check shows you where your testing is currently leaving money on the table.