KorvaJoin the waitlist
Startups

Ten Page Exits Should Not Outrank Two Good Interviews

Counts show how far a problem reaches, not how much it matters or why it happens.

AI image
aicontent.marketing · From Content Signals to Product Roadmap: A Practical Customer-Insight Loop - AI Content Marketing

Maya Ellis, who wrote a November guide on turning search, support, sales, and content clues into product ideas, had walked readers through how to label evidence when she came to the temptation to count.

Her warning was blunt. “A cluster with ten low-context page exits should not outrank two well-documented interviews simply because its count is larger.”

By low-context, she means exits with no explanation attached. They show that someone left, but not whether the person was confused, missed a feature, or simply got an answer and moved on.

Ellis argues that not all evidence deserves the same weight. A customer saying, “I cannot export this report,” is a direct report of a problem, while “a reader leaving a page after 20 seconds,” in Ellis’s words, is only behavior with several possible explanations.

That is why she argues a high count shows how widespread a behavior was, not how much it mattered or what drove it. Ten exits can say how many people left, not how serious the snag was.

To avoid the trap, Ellis urges teams to group clues “around the customer problem rather than the proposed solution.” “Add a CSV export” is one person’s proposed fix, while “teams cannot move report data into their monthly review” describes the job and leaves room for several solutions.

She also asks teams to keep evidence that cuts against the theory. If users search for “bulk edit” but still complete the task another way, she writes, that counter-evidence matters.

Her fix is human review and honest labels. Keep the original source attached, note what the data cannot prove, and check that a summary has not erased qualifiers or turned a guess into a quote.

Questions people ask

What does a useful handoff to product look like?

A useful handoff is one short card per problem, with the task that looks difficult, the evidence and sources, and an honest note on confidence and limits. That format makes uncertainty visible and gives product a question to investigate rather than a feature to build.

Who decides what gets built?

Content teams can surface and explain evidence, but product teams retain decision rights. They decide whether a signal merits discovery, how it compares with other opportunities, and what solution, if any, belongs on the roadmap.

When can modest evidence still deserve a closer look?

When the possible harm is large and the issue fits company goals, even modest evidence can merit a research test. A frequent but low-impact issue, by contrast, may be better solved by updating help content.

More in Startups