Comparison tables that convert — and the six that quietly do not.
The comparison table is usually the highest-converting element on a roundup and the one built with the least thought. Most are a spec dump that hands the reader a research project instead of an answer.
The short version
- Three products, five rows. Every product past three and every row past five reduces the chance a decision gets made at all.
- Delete any row that does not change a decision. Weight in grams is not a row unless weight is the decision.
- The first column after the product should be “best for”, in plain language. It is the only cell most readers actually read.
- Mark the winner visually. A table with no recommendation is a spreadsheet, and spreadsheets do not convert.
- Most of your readers are on a phone. If the table needs pinch-zooming, nothing else on this page matters.
There is a moment in every roundup where the reader stops reading and starts scanning. They have accepted that you know the subject; now they want to be told which one to buy. The comparison table is what they scan, and it is where the decision either happens or does not.
Which makes it strange that most affiliate tables are assembled by dumping every specification the plugin could find into a grid and hoping the reader sorts it out.
Why most comparison tables fail
Not because they are ugly. Because they transfer work back to the reader.
The six mistakes
| Mistake | What it costs you |
|---|---|
| Too many products | Choice paralysis is real. Past about five options the likelihood of any decision being made drops, and a reader who cannot decide does not click anything. |
| Too many columns | Each column is a thing to read across nine rows. Five is a comparison; nine is a database, and databases get skipped. |
| Specifications that decide nothing | Weight in grams, model number, box dimensions. If it does not change anybody’s choice it is padding, and padding hides the rows that matter. |
| No recommendation | The reader came to be told. A neutral grid returns the work to them, and most will leave to find someone who will answer. |
| Broken on a phone | Most affiliate traffic is mobile. A table that needs pinch-zooming or that squeezes eight columns into 360 pixels is not read at all. |
| One button at the bottom | A reader convinced by row two has to scroll past two more products to act. Every row needs its own way out. |
The rules that fix all six
Three products. Five rows. No exceptions worth making.
If your roundup covers nine products, the table does not have to. Put three in the table — the pick, the cheaper option, the upgrade — and let the article body cover the rest. That structure maps onto how people actually shop, and it means the table can be read in one glance.
Five rows is enough for: best for, price, the one specification that decides it, one differentiator, and the button. Anything else earns its place by displacing one of those.
Lead with “best for”, in plain language
This is the single highest-value cell in the table and most affiliate tables do not have it.
Not “Premium tier” or “Advanced”. Write the person: “Most people”, “Small kitchens”, “If you will upgrade in two years”, “Tight budget”. The reader is looking for themselves in the table, and this is the column where they find themselves.
The test for a good “best for” cell: a reader should be able to rule out two of the three rows without reading anything else. If every row could plausibly be them, the column is not doing its job.
Mark the winner, visually and once
A tinted row, a left border, a small “our pick” label. One product only — marking two is the same as marking none.
This is the part people resist, because committing feels risky. It is the opposite: a reader who disagrees with your pick still trusts the page more for having made one, and the alternative is a page that answers nothing.
A button in every row
The reader decides at different rows. Each needs its own exit, labelled with what happens next — “Check price on Amazon” rather than “Buy now”, which is both more honest and, since the price is what they want, more clicked.
Keep the pick’s button visually dominant and the others quieter. Three identical orange buttons is three equal choices, which undoes the recommendation you just made.
Mobile is the actual design
Most affiliate traffic is mobile, so the phone layout is not the fallback — it is the primary case, and the desktop version is the easy one.
Horizontal scroll
Keep the table shape, let it scroll sideways inside its own container, and pin the product column. Works well up to about four columns. The page itself must never scroll sideways — only the table.
Stack into cards
Below a breakpoint, each product becomes a card with its rows as label-value pairs. Better for anything wider than four columns, and the more reliable choice if you are unsure.
- Test on a real phone, not a narrow browser window. Touch targets and scroll behaviour differ.
- Buttons need about 44 pixels of height to be comfortably tappable.
- Never put a horizontal scroll inside a vertically scrolling page without an affordance — a partial cell visible at the edge tells the reader there is more.
- Check the winner marking survives the stacked layout. Tinted rows often lose their tint when they become cards.
Two technical things
Use a real table
An actual <table> with <th> headers, not a grid of divs. It is what screen readers need to announce cells with their headers, and it is what machines read as tabular data. A div grid looks identical and is meaningless to both.
Do not mark it up as a review
A comparison table is not a review of one product, and marking a roundup up as though it were misdescribes the page. Since July 2026 Google’s review-snippet guidelines have also explicitly prohibited fake and undisclosed incentivised reviews in marked-up content — and an assigned rating for a product nobody assessed sits uncomfortably close to that line. → review schema after the 2026 rules
Tables are also a common source of layout shift, because a product image loading late pushes rows down as the reader reaches for a button. Set image dimensions. → Core Web Vitals on affiliate pages
What to actually test
Table design is one of the few affiliate variables you can test cleanly, because the click is immediate and attributable.
- Table position. High on the page serves scanners; after your reasoning serves readers who want to be convinced first. Worth testing per article type — it commonly moves clicks more than any styling change.
- Three rows versus five. The most valuable test in this article.
- Button wording. “Check price on Amazon” versus “See on Amazon” versus “Buy now”.
- Whether the pick is marked. If you have never marked one, this is where to start.
Test one thing at a time, give it enough clicks to mean something, and measure clicks per visitor rather than raw clicks — traffic varies week to week and will otherwise fool you.
Part of The Amazon Affiliate Site Playbook — the whole job in six stages, from choosing a niche to keeping a mature site earning.
Tables that already follow these rules
Ama Affiliate Pro builds comparison tables from ASINs with live prices, a marked pick, a button per row and a mobile layout that stacks rather than squeezes — so the defaults are the version that converts.
Start a 14-day trialCommon questions
How many products should a comparison table have?
Three in most cases — the pick, a cheaper option and an upgrade. Five is the practical maximum. Past that, choice paralysis sets in and the likelihood of any decision drops. If your roundup covers nine products, the table still only needs three; the body covers the rest.
What columns should an affiliate comparison table include?
Product, best for, price, the one specification that actually decides it, and a button. That is five. “Best for” written in plain language — “small kitchens”, “most people” — is the highest-value column and the one most tables omit.
Should I mark a winner in the table?
Yes, and only one. A tinted row or a small “our pick” label. Readers came to be told what to buy; a neutral grid returns the work to them. Someone who disagrees with your pick still trusts the page more for having made one.
How do I make a comparison table work on mobile?
Either scroll horizontally inside the table’s own container with the product column pinned, or stack each product into a card below a breakpoint. Stacking is more reliable beyond four columns. The page itself must never scroll sideways — only the table.
Where should the table go on the page?
Near the top for scanners, or after your reasoning for readers who want convincing first. Both work and it varies by article, which makes position one of the more worthwhile things to test — it typically moves clicks more than styling does.
Should I add review schema to a comparison table?
No. A comparison is not a review of one product, and marking the page up as though it were misdescribes it. Use ItemList for the list, or no markup at all — roundups rank perfectly well without review snippets.
Do comparison tables slow down a page?
They can, mainly through product images loading without declared dimensions, which pushes rows down as the reader reaches for a button. Set width and height on every image, lazy-load anything below the fold, and avoid sortable-table scripts unless readers genuinely use them.
What should the button say?
“Check price on Amazon” tends to outperform “Buy now”. It is more honest — you are sending them to look, not to purchase — and price is what the reader wants next. Keep the recommended product’s button visually dominant and the others quieter.