A shape the reader can follow twice

A written assessment is read twice: once quickly, on arrival, and once properly, some days later. Almost everything that makes a deliverable feel worth its fee happens on the second reading, and structure is what makes the second reading possible at all.

An unstructured piece of writing is read once and never opened again, regardless of how good the observations inside it are.

The default skeleton

Five parts, in this order, for a written assessment of roughly 300 to 800 words.

The summary. Three to four sentences, written last, placed first. The overall read, the single strongest thing, the single weakest thing. A buyer who reads nothing else has still received something.

The observations. The bulk of it, broken into named sections rather than paragraphs. Each section takes one dimension and says what you saw, specifically.

The scores, if you score. A short table, one row per axis, each score with a one-line justification attached. A number without a reason is a number the buyer will argue with.

What would change the assessment. One short section on what a different submission would need to show for the read to move. This is the part earners skip and buyers remember.

A closing line. One sentence, not a summary of the summary.

The whole thing has headings a reader can scan in eight seconds. That eight-second scan is the second reading, and it is why the piece gets kept. Scanning is the default, not a lapse: in Jakob Nielsen's 1997 usability research for Nielsen Norman Group, 79 percent of test users scanned any new page and only 16 percent read word by word.

Why the summary goes first and is written last

Writing the summary before the observations produces a summary you then spend the rest of the document defending. You lose the ability to notice something halfway through that changes the overall read, because you have already committed in writing.

Written last, the summary is a genuine synthesis and takes four minutes. Written first, it is a hypothesis and takes twenty, most of them spent revising it.

Front-loading is not a stylistic preference. A scanning reader gives the opening the most attention, so the most valuable sentence you have belongs there, and the buyer often decides whether the deliverable was worth the money in that window.

Named sections beat paragraphs

The difference between "three paragraphs of observations" and "three headed sections of observations" is not cosmetic.

A heading forces you to know what the section is about, which prevents the most common weakness in this work: a paragraph that drifts across two subjects and settles on neither. It also makes the piece navigable, so the buyer can find the bit they cared about without rereading.

Use the buyer's vocabulary in the headings where you can, and your own where they had none. A section headed with a word from their brief tells them, before they read a line, that you were listening.

Why the same skeleton does not flatten the content

The objection to a fixed structure is that every deliverable ends up reading identically. In practice the opposite happens, for a reason worth understanding.

A fixed skeleton removes the decisions that are not about the material - what order, how long, what goes where - so all the remaining attention goes into what you actually observed. When each job is structured from scratch, a meaningful share of your time is spent on layout, and layout is the part the buyer notices least.

The flattening risk is real but it lives in the sentences, not the shape. Reused headings are fine; reused observations are not, and the boundary is set out in the piece on templates that speed you up without flattening the work. Keeping the content specific inside a fixed frame is its own skill, and writing specifically without padding is the companion to this one.

There is also a straightforward commercial effect. A structure you can name in your listing - "four headed sections, scored, with a note on what would change the read" - is a description of what arrives, which is the thing buyers scan for when they have nothing else to go on.

Length, and where it should sit

Word count is the wrong lever, but the buyer does have a rough expectation attached to price.

These are shapes rather than rules, and the band matters less than internal consistency: a buyer who bought your standard format twice should get the same shape twice. The cost of each extra section is your time, and the effect on your rate is covered in what a second pass is worth.

Where the boundaries of the format sit

Two things stay out of a written assessment regardless of what the brief asks for.

Anything that reads as health interpretation is not a section you write. If a submission raises something that looks medical, the correct output is one neutral sentence recommending a clinician, and no assessment of it.

Anything requiring measurement precision you did not perform is likewise not yours to state. Say what you were given, not what you inferred - the conventions for doing that properly are documented on the site that owns method, and citing a standard beats improvising one.

Scoring axes are worth borrowing rather than inventing from nothing. How scores get read as signals is a useful corrective on what a number communicates once it leaves your hands, and what an automated score actually produces marks out the ground where a written human assessment is not competing - which is most of the value in your fee.

The platform's own guidance on what a judge's written deliverable should contain sits on the Rate Cock judges page, and where its house format differs from this skeleton, its format wins on its own listings.