How 3-2-2 keeps testing lean and readable

What 3-2-2 actually means#

3-2-2 means three creative concepts, two primary text angles, two audiences in one lean test cell: usually one broad prospecting pool and one warm retargeting pool, or two broad themes where retargeting is thin. Concepts differ in hook, proof and format; copy angles differ in promise, not punctuation. Start at the Meta Ads hub for the family view, and read Meta own diversification research for why concept range matters.

It is a starting grid, not a religion: scale winners beyond it, shrink it where budgets demand. Consolidation keeps the structure fundable, and hook craft fills the three slots.

Building the 3-2-2 launch#

Build two ad sets, six ads each: three concepts times two copy angles per audience, with Advantage+ placements on and one optimisation event throughout. Keep ratios mixed per concept so each carries 4:5 and 9:16 variants. Budget each ad set to meaningful events in its first week, merging audiences where volume forbids two survivors. Meta covers structure choices in its learning phase guide.

Brief the three concepts for maximum difference: creator proof, studio detail, carousel logic, for example. Difference teaches; similarity merely spends. Keep one textless master per concept so reframes and translations stay cheap all year. Style splits help pick the three, and ratios keep each native.

Copy angles that pull weight#

Write two angles that argue differently: outcome promise versus mechanism proof, price logic versus risk reversal, social proof versus offer plainly. Pair each angle with every concept so the matrix reads copy and creative separately. Keep primary text front loaded, headlines under forty characters, descriptions short, all in en_GB spelling.

Log angle performance across concepts; angles that lift every concept graduate into permanent messaging. Caption lines should echo the angle without repeating it verbatim, and overlay carries one promise line only.

Reading the 3-2-2 week#

Judge after one full seven day cycle with budgets untouched: thumb stop for hooks, outbound CTR for offers, cost per new customer for verdicts. Promote concept and angle winners together, retire the bottom half, and brief replacements from winner patterns. Never judge on day two spend or platform ROAS blends that mix new with existing customers.

Hold audiences constant through the read; changing targeting mid test restarts learning per edit reset rules. A/B discipline governs any formal follow up test.

Scaling beyond 3-2-2#

Scale by adding concepts, not ad sets: graduate the winning angle into two fresh concepts, keep the consolidated ad sets funded, and raise budgets in steps the event volume supports. Move proven winners into Advantage+ sales builds via Advantage+ sales once signal allows, and expand audiences only after creative depth exists.

Retire methodically as you scale: each new concept enters beside winners, losers leave in the same session. Fatigue rotation takes over from launch testing at this point.

When 3-2-2 does not fit#

Shrink to 2-1-1 on thin budgets where two ad sets starve learning: one audience, two concepts, one angle, judged slowly. Expand to 4-2-2 or 5-3-2 where volume funds it, never by fragmenting into micro ad sets. Regulated, geo fenced or catalogue briefs adapt the grid through location rules, special categories and catalogue setup.

The grid serves learning; volume decides its size. Ask our team for a sizing check when events per ad set trail 50 a week and verdicts wobble.

Frequently asked questions#