165 design references

Clarify the campaign before generating variations

@sociyell on Instagram

Set the campaign direction, the voice rules, and the one message before you make posts. When you test an ad, change one thing and leave the look and the message alone.

What to look at

Nine panels, and the ground changes each time. A cream carved mask on a black hood, against blue, reads Stop using Calude to write. Start using it to structure. A black page says the brief was vague and names Claude at 9am, before Figma. Build Content Sytems (Not Posts) sits over a map, a paper bag, and a compass. A man in a white shirt on a red sofa sits above three chips: campaign direction, five angles, eight weeks. A paper page lists what you never say, then claims revisions drop by half. One black page says change one variable, keep the message locked, and shows an open padlock. The last page asks for a comment.

Change one variable, and keep identity, the visual system, and the message locked. The padlock on the slide is open.
Change one variable, and keep identity, the visual system, and the message locked. The padlock on the slide is open.
The brief is supposed to become a direction, five angles, and an eight-week flow before the posts exist.
The brief is supposed to become a direction, five angles, and an eight-week flow before the posts exist.

Why it works

Nielsen Norman Group says a live test is readable when the new version differs "only in one design element alone," such as a button, an image, or a description. That finding is about websites and products, not about this post. The same article says a test that changes several things at once does not show which change did the work, and that this kind of test is a poor fit for a page with little traffic. Their multivariate write-up says you can try combinations, but every added combination splits the visitors into smaller groups and needs more traffic. A separate article, also about screen design, says items that share a color, a shape, or a size are read as related. That is why a locked visual system keeps variants looking like one brand.

Try it

  1. Write the campaign sentence, the lines you will never say, and the eight-week list before anyone opens a design file.
  2. For a test, change only the hook, or only the angle, or only the framing. Leave the logo, the colors, and the offer sentence as they are.
  3. Name the one result you hope will move, and a second result that must not get worse, such as refunds or replies that say the page was confusing.
  4. Leave the test up for at least one to two weeks of normal traffic. Do not pick a winner from one afternoon of generated options.
  5. If the page gets few visits, rewrite from questions customers already asked, and skip the split test.

What it costs

These slides do not keep one visual system. The blue mask, the black type, the red sofa, the paper, the sneakers, and the grass are different worlds, and the padlock beside Keep locked is open. Listing a new hook, a new angle, and a new framing together is three changes, and the website research says you then cannot tell which one did the work. A stack of generated lines is not a live test. Sociyell says revisions drop by half, that a meeting is avoided, and that the rewrite loop is hard past 10 clients. Those are lines on the slides. They were not checked. The cover spells the tool Calude, and a later title spells systems as Sytems.

Words worth knowing

Sources