compare-table

compare-table

Build a small table when the reader is choosing between things. Prose buries differences; a row-by-row scan reveals them.

Audience

Someone deciding. They want to glance down a column, glance down the next, and feel the contrast. They will not read a paragraph that says "X does A and Y does B and X also does C while Y…"

When to use

Trigger on any of:

• "compare" / "what's the difference" / "X vs Y" / "side by side"
• "should I use A or B" / "which should I pick"
• "table this" / "make a table"
• The explanation is three or more sentences of the shape "X does [thing], whereas Y does [other thing]" — that's prose smuggling a table

Do not trigger when:

• There is only one option. A one-column table is a list.
• The differences are nuance-heavy and qualified ("it depends on…"). Tables flatten. Use prose with caveats instead.
• The items aren't actually comparable on the same axes (apples vs oranges). Don't force a table; explain the category mismatch.

Pick the shape first

Two layouts. Pick before you draw.

Shape	Use when	Header row	First column
Options-as-columns	2–4 options, 4+ criteria	option names	criteria
Options-as-rows	many options, 2–4 criteria	criteria	option names

Default to options-as-columns. It's how readers expect to compare. Switch to options-as-rows only when you have more than four options.

Rules

1. Four options max. Five columns of comparison is a feature matrix, not a decision aid. If you need five, you're not comparing — you're cataloguing.
2. The first column is the axis of comparison, not an option. The axis names the criterion ("Startup cost", "Latency", "Maintenance"), not the answer.
3. Cells: five words max. A cell with a full sentence is prose in disguise. If a cell needs a sentence, the axis is too vague — split it into two narrower axes.
4. Empty cells use —, not "N/A" or "none". — reads instantly; "N/A" stops the eye.
5. No "winner" row at the bottom. Readers decide. A "winner" row is you selling, and you lose trust the moment they disagree with one criterion.
6. Headline above the table: the takeaway, not the topic. Not "Database comparison" but "Postgres wins on flexibility; SQLite wins on ops cost."
7. Prose below the table, 2–3 sentences max. Cover the nuance the table flattened — the "it depends on" and "but only if".

Format

**Takeaway:** [one sentence — what the reader should conclude]

| Criterion | Option A | Option B | Option C |
|---|---|---|---|
| [axis] | [≤5 words] | [≤5 words] | [≤5 words] |
| [axis] | [≤5 words] | [≤5 words] | [≤5 words] |
| [axis] | [≤5 words] | [≤5 words] | [≤5 words] |

[2-3 sentences: the caveat, the "it depends", the thing the table
cannot say.]

Example

Bad (prose smuggling a table):

Postgres is a full relational database with strong concurrency and rich SQL features but you have to run it as a service. SQLite is a single-file embedded database that's trivial to deploy but has weaker concurrent-write performance. DynamoDB is fully managed and scales horizontally but locks you into AWS and a key-value model.

Good (takeaway + table + nuance):

Takeaway: Postgres for flexibility, SQLite for simplicity, DynamoDB for scale-without-ops.

Criterion	Postgres	SQLite	DynamoDB
Deployment	Run as service	Single file	Fully managed
Concurrent writes	Strong	Weak	Strong
Query model	Full SQL	Full SQL	Key-value
Vendor lock-in	None	None	AWS only
Ops cost	Medium	None	Low

If you're prototyping and may grow, start with Postgres — the cost of switching from SQLite later is real. DynamoDB only makes sense if you're already on AWS and know your access patterns.

What you are not doing

• You are not making a feature matrix for marketing. Marketing matrices have 30 rows and a checkmark column. Decision tables have five rows and force-rank the real differences.
• You are not selling. The reader is allowed to pick the option you wouldn't.
• You are not flattening real complexity. If the choice genuinely depends on something the table can't show, say so in the prose below — don't lie with a clean grid.