Writing a comparison page that does not read as an ad

What a believable comparison contains

A comparison page that a technical buyer believes uses the same attributes for every row, fills them from sources the buyer can check, shows the rows where you lose, says plainly that you wrote it, and dates every figure. A page missing any of those reads as an advert with a table in it.

Vendors write these pages because buyers search for "X vs Y" and "X alternatives". The search brings the buyer. Whether they stay depends on whether the first table they see looks like it could have come out differently.

A table where your product wins every row could only have come out one way, so it carries no information about the products, only about who wrote it.

The checkmark column problem

The classic vendor comparison has your product in the first column with a green tick in every row and competitors with a scattering of crosses. The rows were chosen after the ticks. A buyer who has used one of the competitors can spot a wrong cross quickly, and from then on has reason to doubt every other row.

The fix is to choose the attributes before you look at the answers, and to choose the ones buyers ask about rather than the ones you win. For a GPU cloud that means card, memory, unit, commitment, regions, egress terms and minimum term. Some of those you will lose. Leaving them out is the choice a reader notices first.

A cross is also the wrong symbol when you do not know. If a competitor does not publish its egress policy, the cell should say "not published", not show a cross. Our tables hold three states for this reason: yes, no, and no answer on record.

Prices on a comparison page

A competitor price on your page is a figure you do not control and that will change without telling you. If you show one, link to the page it came from and date it, and accept that it will be stale before you next look. How far apart sellers of one card sit is itself a figure: across the 29 cards we compare, the median ratio of dearest to cheapest seller is 3.7×. A comparison that picks one competitor rate shows one point in that range and hides the rest.

A better habit is to link out. Every GPU rate we hold sits beside its source page and read date, and a comparison can point there for the market view while stating your own rate on your own page. A link to a table the vendor does not control gives the reader a way to check the comparison.

Disclosure, in the first screen

Say who wrote the page at the top, not in a footer. "We sell GPU rentals, and this compares us with other providers we compete with" is a sentence that costs nothing and changes how every row below it is read. A reader who discovers the authorship halfway down has to reread everything above it with that in mind.

If you have a commercial relationship with anyone on the page, as a reseller, a partner or an affiliate, say that too. The same rule governs our own pages: paid placement is labelled, and no money changes where a product sits in a ranked list.

Disclosure also covers how the data was gathered. "Competitor figures read from their public pricing pages on the dates shown" is one line, and it answers the question a sceptical reader asks before any other: where did you get this?

A structure that survives scepticism

Open with one sentence that answers the comparison honestly, including the case where the competitor is the better choice: "If you need a single card for a week, either works; if you need a full node reserved for a year, compare terms below." Then the attribute table, every row filled from a source or marked as not published. Then a section on where each option fits.

End with dates and sources rather than a call to action. A buyer who has read that far is deciding; a list of links to the evidence helps more than a button. The side-by-side compare view is built on the same idea, with the attributes held in common and the gaps stated.

Keeping it true after you publish

A comparison page starts going wrong the day it ships. Competitors change prices, add regions, drop a card or change their egress terms, and none of them tells you. Put a review date on the page and keep it: a named owner rereads every competitor cell against its source each quarter and updates the date only when that has been done.

When a competitor writes to say a cell is wrong, fix it quickly and say so on the page. A short note that a row was corrected on a given day, with what changed, does more for the credibility of the rest of the table than any amount of care in the first draft. It shows that the page answers to evidence.

If you cannot keep a competitor row current, remove it. A comparison with fewer rivals that are all accurate beats a long one with stale rows, because the buyer will check one row and judge the rest by it.