Winning the comparison questions

The problem

Fernway’s buyers were asking AI engines to compare it with two larger employee-onboarding platforms. The answers usually favored the rival, sometimes using integration lists and product descriptions that were years out of date.

Fernway already had comparison pages, but they read like sales pitches. They asserted that the product was better without giving a reader much evidence to inspect. The answers we tracked leaned on older third-party reviews instead. Fernway won just 3 of the 40 head-to-head prompts in the starting set.

The brief was to improve the quality of the comparison record. We wanted the product to win when it fit the buyer’s situation, with the reasons visible in the answer.

Agreeing what a win meant

We defined a win as an answer recommending Fernway for the situation described in the prompt. A passing mention did not qualify. Nor did a recommendation that depended on a feature the product did not actually have.

The 40 prompts included the company sizes and requirements that made the comparisons meaningful. Keeping those constraints in the questions prevented the exercise from becoming a contest to see which product could attract the broadest praise.

What we did

Weeks 1–2. We tracked the head-to-head prompts and inspected the reasons each answer gave. We separated incorrect facts from legitimate differences between the products. The former belonged in a correction queue; the latter belonged in a fair explanation of fit.

This distinction changed the content brief. A useful comparison page needed to help a buyer make a decision, including situations where the competing product would be the more sensible option.

Weeks 3–8. We rewrote six comparison pages as plain tables with supporting sources. The copy made the relevant differences easier to inspect: integrations, reporting, setup and the kinds of team the product served. Claims were tied to evidence rather than a repeated “best for everyone” conclusion.

We also corrected integration lists on review sites. Updating Fernway’s own pages would have left the comparison record inconsistent if the same old limitations remained visible in the sources already being cited.

Months 2–5. We invited happy customers to leave honest reviews on the three sites appearing most often in the tracked answers. Customers were asked to describe their own use of the product. The goal was specific experience a buyer could evaluate, rather than a prescribed phrase or guaranteed positive rating.

What moved

Fernway went from winning 3 of 40 comparison prompts to 24 of 40 over five months. The recommendations became more closely connected to the buyer’s stated requirements and the current product facts.

The pages started being quoted once they were more candid about trade-offs. That observation informed the direction of the work, but it does not isolate a single cause: comparison copy, review-site corrections and customer reviews were all changing during the engagement.

We did not treat the remaining losses as automatic defects. An answer recommending a rival for a requirement Fernway did not meet was a reasonable outcome, not something to conceal from the report.

What we’d do differently

We would be fairer sooner. Admitting where a competitor was stronger made the comparison pages more useful and gave the team a clearer way to explain its own strengths.

For another software team, the starting exercise is simple: read a comparison page as a buyer with a specific constraint. If it never explains who should choose the other product, it probably has not explained the decision well enough yet.