Judging a Text Thread When There’s No Image to Hide Behind

A long text thread in an agency’s portfolio is hard to assess honestly. No colour grade to admire, no cut to praise, no photograph doing quiet work behind the words. Does the thread earn every part it takes, or was it stretched to a number picked before the argument existed?

The image is not there to save it

In most social formats, visual craft puts a floor under mediocre writing. A strong photo supplies the hook before anyone reads a word. Colour and composition signal tone in advance. A good video often carries a payoff the words never quite deliver, the last shot doing work the caption only gestured at.

A text thread has none of that. The opening line and the sentence structure do the whole job a visual normally shares. LinkedIn’s help documentation caps a standard post at 3,000 characters, room to build a real argument and room to hide a thin one.

What earned length actually looks like

First marker: every part introduces something the reader did not have a moment earlier, a fresh claim, reasoning that supports the last claim, or a complication that makes the picture harder. Not a new phrasing. New material.

Second marker: the argument breaks if you delete any single part. Where length is earned, removing part four leaves a gap you can feel, because part five rested on it. In a padded thread you can lift out two parts and the rest reads as well as before.

Third marker: the transitions do work. A good one picks up the specific claim just made and pushes against it. A weak one is a bridge phrase that would sit between any two parts of any thread. “But here’s the thing” is not a transition.

The padding tells

The patterns:

• The opening hook restated in slightly different words a third of the way in.
• One sentence split across two parts, the second opening with a clause that belongs to the first. That raises the count, it does not create a beat.
• A “let that sink in” part, which adds nothing and asks the reader to supply significance the writing failed to establish.
• A closing part that is a follow request wearing the costume of a conclusion.
• Numbered list formatting on an argument that is not a list, where items three and five overlap and the order could be shuffled.

Reading the opening line without a visual hook

With no thumbnail and no cover frame, the first line is the entire hook. A strong one promises something specific: a number the reader will be given, a named contradiction between two things they both believe, a concrete tension between what a category claims and what it visibly does. A weak one is a bare curiosity gap, withholding a subject instead of naming a payoff.

The test follows from that. If the first line promises a specific thing, judge the rest on whether it delivers that exact thing, not a nearby substitute that happens to be interesting.

Judging the shape of the argument, not the polish of each line

Sentence-level polish is the easiest thing to fake. Instead, reduce each part to one clause capturing its claim and write those clauses in order. A real progression moves: setup, complication, a turn the setup did not obviously lead to, then a resolution that accounts for the complication. A flat one restates the same clause five times. It is also how you go about reading strategy from what an agency actually published, since the shape of an argument shows how its writers think.

A thread that earns its length, part by part

Here is a generic shape to test work against, hypothetical rather than a real account.

Part one states a specific claim with a stake in it: a particular kind of campaign underperforms for a particular reason. Part two gives the reasoning, showing the mechanism instead of restating the conclusion more loudly. Part three takes on the obvious objection, the one an informed reader had from part one, and either defeats it or concedes the ground it owns. Part four says what changes if the claim holds, which is what makes it worth reading rather than merely correct. Four parts, and deleting any one collapses it.

Failures the format exposes that a caption never would

The first is unclear pronoun reference across parts. Part four opens with “this is why it fails,” and “this” could point to the claim in part two, the objection in part three, or the example between them. Under a photograph the reader’s eye supplies a referent. Across a thread, different readers pick differently, so the piece has no single meaning.

The second is an argument that implies its conclusion but never states it. A caption paired with a strong image gets away with that, because the image states the conclusion. A thread cannot. If you read every part and still cannot write the position down in one sentence, it was never on the page. One good thread also proves less than it looks like, which is part of what a portfolio leaves out.

How this differs from judging a caption

A caption is a supporting document. It can lean on its image for the hook, the payoff, or both, which is why judging a caption written to accompany a visual post means reading words and image as one unit. A thread has no partner, so it builds both in text alone. That raises the bar for the writing working on its own, and holding a thread to the caption standard will make you overrate it.

Where to take this next

If an agency’s written work stands up to the deletion test, tell us. You can nominate an agency whose threads hold together, and say which piece convinced you. Our selections are editorial and we say why, so the reasoning matters.

FAQ

Is a longer thread automatically better craft than a shorter one?

No. Length is not a craft signal by itself. A five part thread where every part earns its place is stronger than a fifteen part thread that restates its hook twice. Delete a single part and see whether the argument breaks.

What is the single biggest tell of a padded thread?

A part that could be removed without the argument losing anything, most often a “let that sink in” beat or a restatement of the opening hook partway through. Both feel like emphasis and function as filler.

Sources

• LinkedIn Help, Post and share updates, fetched 6 September 2026, for the 3,000 character limit on a LinkedIn post.