How to write a comparison post

The structure that works

Step by step

1. 01

   Pick a comparison with live demand

   Compare what buyers already shortlist: the pairs raised on your sales calls, the “X vs Y” searches with autocomplete behind them, the competitor your churned customers actually left for. If you sell one of the options, the useful comparison is against the rival you lose real deals to — a matchup staged against a weak opponent persuades nobody and ranks for nothing.

   Hold the format to two options, three at most. At four, the verdicts blur and the choose-if logic stops being writable; that market wants a category listicle plus separate head-to-head pages for the two or three pairs people actually search.

2. 02

   Use both products for real work

   Run the same real project through each: import your actual data, build the same workflow, invite the same teammates, hit the same limits. Three to four hours per product exposes onboarding truth; two to three weeks exposes the deal-breakers — the export that mangles data, the limit that only appears at scale. Screenshot as you go; the screenshots are your proof of presence.

   Keep a running friction log — every moment of confusion or surprise, with timestamps. Those notes become the observations readers quote, because they cannot be derived from either pricing page. The difference between a comparison and a features rewrite is exactly this log.

3. 03

   Choose the 5-7 criteria buyers decide on

   Source criteria from decisions, not feature lists: what buyers ask on sales calls, why switchers say they switched, what the migration threads complain about. Pricing at realistic usage is nearly always criterion one — compute a real scenario, say 10 seats plus the add-ons the workflow needs, because from-prices are marketing rather than cost.

   Refuse the union of both feature matrices. Forty rows of checkmarks compare everything and decide nothing. Each criterion buys its place with one property: it could change the reader’s decision. A row that could not swing a purchase is trivia.

4. 04

   Build the verdict table before the prose

   Rows are criteria, columns are products, every row gets a marked winner and one deciding fact per cell — “$104/month for 10 seats”, not “affordable”. Building the table first forces every judgment while changing your mind is still cheap, and the finished table becomes the article’s outline: one section per row.

   Where a row is genuinely hard to call, split the criterion until it decides — “reporting” becomes “built-in reports” and “exports” — or mark the winner honestly as narrow. The table then sits in the post’s first screen with a one-sentence verdict beside it, because half the readers came only for this.

5. 05

   Write each criteria section to a verdict

   Fixed pattern, 150 to 300 words per section: what you tested, what happened in each product — with the screenshot or the number — and the winner, for whom. The observational register does the persuading: “A’s importer mapped 14 of our 16 fields; B mapped all 16 but silently dropped 200 rows” outweighs any adjective either vendor ships.

   Ban the tie. “Both are fine here” means the criterion is too broad to decide anything — split it or replace it. Every section must move the reader toward one column of the choose-if block; a section that ends neutral is word count spent avoiding the format’s job.

6. 06

   Draft the choose-X-if block and the verdict

   Write 3 to 5 choose-X-if bullets per product, each starting from the reader’s situation rather than the product’s feature: “You invoice in three currencies”, “Your team lives in Slack”, “IT must approve every vendor”. Test the block against real people — every customer and lost deal you can remember should land in exactly one column.

   Then the verdict: the majority pick with its boundary quantified (“under 20 seats, A”), the runner-up’s win conditions, and one explicit skip-it line per product. If you sell one of them, concede a real loss here in plain words. That concession is the honesty premium — the sentence that makes the recommendation quotable instead of dismissible.

7. 07

   Date it and schedule the re-check

   Comparisons rot faster than any other format: prices change, tiers get renamed, the missing feature ships. Put a visible “last verified” date near the verdict table and re-verify quarterly — immediately when either product announces pricing. One stale price a reader catches retroactively invalidates every claim they cannot check themselves.

   Make the re-check cheap: keep a 30-minute verification list at the bottom of your draft — prices at your scenario, plan limits, the screenshots that age — and run it each quarter. Maintained comparison pages hold rankings and citations for years; unmaintained ones lose both to whoever verified more recently.

Before and after: what good looks like

The before is the verdict readers dread: 81 words of “it depends” that hand the decision back — the try-both-trials advice outsources exactly the work the post promised to do. The after draws the boundary with numbers (15 seats, $8 versus $19, self-sufficient by day two), names each product’s deal-breaker, handles the overlap zone, and closes with a disclosure that doubles as proof of use. Note that it concedes real failures on both sides — the missing reports, the manual exports — which is precisely what makes its recommendation quotable.

Mistakes that sink a comparison post