How many creatives per ad set? Four to six genuinely different concepts, launched together with equal budget and left alone for at least 72 hours. Genuinely different means a different angle or a different format, not a different crop or button color, because delivery systems now cluster near-identical variants and effectively treat them as one ad.
That number is a testing number rather than a ceiling. Once a concept wins, keep one or two challenger creatives running against it at all times, iterate the winner with new hooks and new formats, and rotate direct response creative every one to two weeks so a fatigued winner never leaves a gap. Ship the next batch weekly if budget allows; slower is fine as long as it is regular.
This guide covers how to plan, make, and test ad creative in 2026, now that creative is the biggest lever left in paid advertising. It is written for a founder running their own ads without a creative team, a media buyer, or much patience.
Creative is the targeting now
The big platform shift of the last two years: delivery systems read your creative and decide who sees it. Meta's Andromeda retrieval overhaul analyzes the visual elements, format, copy, and concept of each ad, and uses that as the primary signal for matching ads to people. Practitioner analyses of the update describe Meta assigning similarity-based identifiers to ads, so ten near-identical variants of the same product shot reach the same audience ten times, while a genuinely different concept unlocks new people. See Webtopia's breakdown of Andromeda for the mechanics.
Do this now: open your ad account and cluster your live ads by concept, not by campaign. If every ad is the same image with a different text overlay, you effectively have one ad. Your job for the rest of this guide is to fix that.
Pick your formats: static, video, UGC
You do not need all formats on day one, but you do need more than one. Format diversity is one of the cheapest forms of creative diversification, and Webtopia cites data showing a varied format mix reducing cost per acquisition by 7.3 percent.
- Static images are the fastest to produce and still convert, especially for clear offers, strong before-and-after visuals, and retargeting. A screenshot of a tweet-style text box, a product on a plain background with one benefit line, or a review pulled into a graphic all count.
- Video earns its production cost when something needs demonstrating: the product in use, a transformation, a founder explaining the origin story to camera.
- UGC-style video is creator-shot, native-looking footage: someone talking to their phone camera about a problem and how they solved it. It works because it does not pattern-match to "ad" in the first second. You can commission it from creators or film it yourself; polish is optional and sometimes harmful.
Do this now: for your next batch, take one concept and produce it as a static, a talking-head video, and a text-heavy "ugly ad". Same message, three formats.
Hooks: win the first three seconds
Most video ads die at the open. The practical measure is your 3-second view rate: what share of impressions watched at least three seconds. Motion's analysis of DTC accounts treats this as the core measure of hook effectiveness, and notes that most ecommerce video ads fail in those first seconds because they look like ads.
Hook types that keep working: a direct question aimed at the persona, a bold or contrarian claim, a "stop doing X" negative hook, a before-and-after teased upfront, a pattern interrupt (unexpected visual, odd prop), and a POV or story open. Do this now: write ten hooks for one concept before you film anything. Film the body once, then shoot the ten hooks separately and cut ten versions. That is ten ads for the cost of one afternoon.
Map personas to awareness levels
Angles are not decoration, they are targeting. The classic tool is Eugene Schwartz's five stages of awareness: unaware, problem aware, solution aware, product aware, and most aware, with messaging matched to what the prospect already knows. This walkthrough of the five stages is a good refresher.
Build a simple grid: personas down the side, awareness levels across the top, one angle per cell. Example for a bookkeeping app:
- Solo founder, problem aware: "You have revenue but no idea if you are profitable."
- Solo founder, solution aware: "Spreadsheets versus software: what four hours a month actually costs you."
- Solo founder, product aware: handle the objection, "Switching took me 20 minutes, here is the migration."
- Solo founder, most aware: the offer itself, plain and urgent.
Do this now: two or three personas, three awareness levels each, gives you six to nine distinct angles. That is a quarter of creative concepts, planned in an hour.