The 3-2-2 Facebook Ad Testing Framework Explained

Most people test Facebook ads like they're firing a shotgun in the dark. They throw fifteen creatives into one ad set, wait three days, panic, pause everything, then ask their friend on Twitter why nothing is working.

The 3-2-2 framework is the antidote. It's a structured testing approach used by performance media buyers running serious ecommerce accounts, and it scales down beautifully to small budgets too. Once you internalise it, you'll never test creatives randomly again.

What the numbers mean

• 3 concepts — three fundamentally different angles or offers
• 2 hooks — two distinct opening lines or visual hooks per concept
• 2 formats — two ad formats (e.g., static image and video) per hook

Multiply it out and you get 12 ad variations per testing cycle. That sounds like a lot, but it's structured chaos rather than random chaos. Every variation exists for a reason and tells you something specific about what's working.

Why concepts matter more than formats

Most beginners obsess over format — "should I run a carousel or a video?" — when concept is doing 80% of the heavy lifting. A great concept on a mediocre format will outperform a mediocre concept on a polished video almost every time.

A concept is the underlying message. Examples for a sleep supplement:

1. Concept A: "Fall asleep in under 15 minutes" (speed/efficiency angle)
2. Concept B: "Wake up without an alarm" (lifestyle/aspirational angle)
3. Concept C: "Stop tossing and turning at 3am" (problem/agitation angle)

Three completely different ways into the same product. Now you know which message resonates, regardless of how it's delivered.

Why two hooks per concept

A hook is the first 1-3 seconds of a video or the headline + opening line of a static. It's what stops the scroll. Even with the same underlying concept, different hooks land with different people.

For Concept A ("Fall asleep in under 15 minutes"), your two hooks might be:

• Hook 1: "I haven't taken longer than 15 minutes to fall asleep in 6 months"
• Hook 2: "Counting sheep is for amateurs. Here's what actually works."

Same concept, completely different opening. You'll often find one hook outperforms the other by 2-3x even with identical messaging beneath it.

Why two formats per hook

Finally, you split each hook across two formats — typically video and static, or short-form vertical and feed-style 1:1. Format affects placement performance massively. A great Reels video might bomb in the feed; a beautiful static might die on Stories.

By testing format last, you're isolating the variable that matters least but still capturing it in your data.

The maths: 3 × 2 × 2 = 12 ads

A full 3-2-2 cycle gives you 12 ads. At £30-£50/day for a week, that's enough budget to get statistically meaningful signal on which concept, hook and format combination wins.

Don't run all 12 in separate ad sets. Put them all in one ad set and let the algorithm distribute spend. You're looking for the cream to rise.

How to run a 3-2-2 test step by step

1. Brief the concepts. Write down your three angles in one sentence each. Make them as different as possible.
2. Write hooks. Two hooks per concept = six hooks total. Write 12-15 and pick the best six.
3. Produce formats. For each hook, create both a video and a static version. Keep production scrappy — you're testing concept, not Cannes.
4. Build the campaign. One campaign, one ad set, 12 ads. Use a Sales or Leads objective, broad targeting, CBO off (use ABO so you control budget).
5. Set budget. £40-£60/day for a 7-day test. That's enough to spend £300-£420 across the test.
6. Don't touch it for 5 days. This is where 90% of testers fail. Leave it alone.
7. Read the results on day 7. Look at CTR, CPM, CPC, and (ideally) CPA per ad. Identify the winning concept, then the winning hook within it, then the winning format.

What to do with the winners

The winning ad becomes your control. The winning concept tells you which angle to mine for the next testing cycle. The winning hook tells you what kind of opening converts your audience. The winning format tells you what to lean into next.

Then you run a fresh 3-2-2 with iterations on what you learned. New hooks within the winning concept. A new third concept inspired by what you saw. A new format twist.

It's a flywheel. Every test makes the next one better.

Common mistakes to avoid

• Testing too many things at once. If your three concepts are all variations of the same message, you're learning nothing.
• Killing ads on day 2. Statistical significance needs spend. £40 across 12 ads on day 2 is £3.30 per ad. That's not a verdict, that's a feeling.
• Not documenting hypotheses. Write down what you expect each ad to do before launch. You'll be wrong half the time, and that's where the learning happens.
• Testing on a brand-new pixel. If your account has no conversion history, the algorithm has nothing to optimise toward. Get baseline data first.

When to graduate from 3-2-2

Once you're spending £500+/day and producing 5+ creatives a week, you can move to a more sophisticated cadence — usually 5-3-2 or rolling daily creative refreshes. But 3-2-2 is the perfect baseline for accounts spending £30-£300/day, and you can run it for years without it getting stale.

Letting AI run the test cycles

Manually running a 3-2-2 every two weeks is doable but it's easily an hour a day in Ads Manager. Pix-Vu automates the whole loop for £99/month — it sets up the testing structure, monitors statistical significance, kills losers and promotes winners without you having to log in. Fewer late nights staring at the breakdown view.