The Prescribed-Remedy Format: Findings You Can Actually Fix

Doctors figured out something a long time ago that most audit tools still haven’t: a diagnosis without a prescription is not a service. Imagine a doctor who examined you carefully, announced “your blood pressure is elevated and your ankle is broken,” charged you, and left. Every fact correct. Every fact useless. What you wanted was not a description of your state but a path out of it. Yet this is precisely the transaction most website audits offer: an inventory of what’s wrong, delivered with an air of completion, as if naming the problem were the job.

When I designed GazeSite’s report format, I made a rule that sounds small and turned out to govern everything: no finding ships without a prescribed remedy. Every finding is four parts — a score, a one-line headline, the issue, and the remedy — and the remedy is mandatory. Not “consider improving this.” An actual instruction: change this text to something like that, add alt text to these images, move this button above the fold on mobile, replace this gray with a color dark enough to read. If the system can’t produce the instruction, the finding doesn’t ship.

Here’s what I didn’t expect: this rule improves the findings themselves, not just their presentation. It’s a constraint on the auditor, and constraints on the auditor are where quality comes from. To prescribe a remedy you must understand a problem at a much deeper level than you need to merely point at it. Anyone can say “the headline is weak.” To propose a better headline you have to determine what the page is for, who arrives at it, and what they need to hear — and if you can’t determine those things, you’ve discovered that your “finding” was really just a vibe. Requiring the remedy filters out the vibes. It’s the difference between criticizing a sentence and rewriting it; only the second proves you understood it.

The four-part structure isn’t decoration either; each part carries a different mode of the reader’s attention. The score answers “how much should I care?” — it’s what lets ten findings be ranked instead of merely listed. The headline answers “what is this about?” in the half-second you give each line while scanning. The issue answers “convince me” — here’s what’s on your page, here’s why it fails, here’s who it fails for. And the remedy answers the only question that produces motion: “so what do I do?” Most reports serve the first three and abandon you at the fourth, which means they abandon you at the exact moment you were ready to act. The gap between understanding a problem and knowing its fix looks narrow from the auditor’s side. From the owner’s side it’s a canyon. The owner of a small business site can understand “your form has too many fields” instantly and still have no idea which fields are safe to cut. The remedy is the bridge, and the auditor is the one standing on the side with the knowledge, so the auditor should build it.

There’s a fair objection: a machine prescribing fixes will sometimes prescribe wrong. True, and worth being honest about. The remedy for a clarity finding is a judgment call wearing the syntax of an instruction; you might read the suggested headline and think, no, that’s not our voice. Fine. Rejecting a concrete proposal is itself fast and clarifying — you now know what you don’t want, which is more than you knew before. A vague exhortation to “strengthen your messaging” can’t even be rejected, because it never committed to anything. Concrete-but-arguable beats vague-but-unfalsifiable in any document meant to cause action. The failure mode of a specific remedy is a quick “no.” The failure mode of vague advice is a committee.

The scores matter here for an unglamorous reason: remedies compete for the same weekend. Nobody fixes everything. A report is really a proposal for how to spend a scarce afternoon, and ranking is the honest acknowledgment of that scarcity. A finding scored severe with a twenty-minute remedy is a gift; a mild finding whose remedy is a redesign can wait forever, guilt-free. Without scores, every finding arrives with equal urgency, which the reader correctly translates as no urgency.

The test I’d apply to any audit, review, or piece of feedback — from my tool, a consultant, or a well-meaning colleague — is what I’ve started calling the Monday-morning test. Read the finding and ask: does this tell me what to do Monday morning? Not what’s wrong, not what to think about, not what to “explore.” What to do. Feedback that survives the test gets acted on; feedback that fails accumulates in a document named something like “site improvements” that everyone has and no one opens.

Deep down this is a claim about what feedback is for. Description feels like the respectable part and prescription the presumptuous part, so critics of all kinds hide in description, where they can’t be wrong. But feedback that can’t be wrong also can’t be used. An audit’s product isn’t knowledge; it’s your next action. Everything before the remedy is preamble.