Section Unavailable: What It Means When We Can't Grade a Page

Occasionally a GazeSite report comes back with a section marked unavailable. One of the six reviewers — say, readability, or SEO — has nothing to say, because the evidence it needed wasn’t there to review. People sometimes read this as the product failing. I want to explain why it’s the product working, and why that empty section is often the most important thing in the report.

Start with what has to happen before any grading occurs. You give us a URL. A real browser fetches it, waits for it to load, takes screenshots at desktop and mobile sizes, and saves the HTML. Only then do the reviewers go to work, each grading the page from its own angle and returning findings with scores and remedies. Every judgment downstream depends on the capture upstream. If the capture is compromised — the server didn’t respond, the page returned an error, a bot-protection wall served us a challenge instead of your content, a script crashed and left the page blank, the load timed out halfway — then some or all of the reviewers are looking at something that isn’t really your page.

At that point there are two options. Grade it anyway, or say so and decline. Grading it anyway is tempting, commercially. Customers paid for six sections; give them six sections. A reviewer confronted with a blank page could dutifully produce findings: no headline, no navigation, poor content structure, score of zero. Every word technically true and every word garbage, because the page it describes is an artifact of a failed load, not the thing you built. Worse than useless, actually — actively misleading, because it sends you off to fix problems that don’t exist. The moment a report knowingly includes even one section of confident nonsense, you have to wonder about all the other sections too. So we take the other option. When the evidence isn’t good enough to support a judgment, the section says unavailable, and it tells you why. I think of it the way I think of null in a database. Null doesn’t mean zero. It means we don’t know, and the entire value of the convention collapses if you ever write zero when you mean null.

Here’s the part that reframes the empty section from disappointment into finding. Ask why a robot with a real browser, decent patience, and no ad-blocker couldn’t get a usable copy of your page — and then ask who else can’t. Search engines fetch your page with automated browsers; if we got a bot-challenge wall, there’s a fair chance some crawlers sometimes do too, and a page a crawler can’t read is a page that can’t rank. Link previews in chat apps and social feeds are built by robots fetching your page; if the fetch fails, your links unfurl into gray boxes. Screen readers depend on the HTML actually arriving and rendering. Visitors on slow connections experience your twenty-second load as a blank tab and a back button. The population of “clients that fail to load your page” is not one audit robot. It’s a whole demographic, mostly silent, that simply never becomes your audience. The audit is just the one member of that demographic that files a report on the way out.

So when a section is unavailable, the prescribed remedy is implicit but urgent: find out what your server does when a stranger’s browser shows up, and fix that before worrying about contrast ratios. Sometimes it’s overzealous bot protection that’s catching far more than bots. Sometimes it’s a fragile single-page app that renders nothing when one script fails — and shipping a page whose entire content depends on every script succeeding is a bet you’re placing on every visitor’s network. Sometimes the site was just down, which is worth knowing immediately rather than whenever a customer bothers to email. In each case the unavailable marker is the earliest, cheapest signal you’ll get. Your human visitors won’t send you an error report. They’ll just leave.

There’s a broader principle here that I care about beyond this product. Automated judgment is spreading into everything, and the systems doing the judging face a constant temptation to always produce an answer, because an answer is what the customer expects and confidence is what sells. But a system that always answers is a system whose answers you can’t trust, since you can never tell which ones were forced. The most useful thing a judgment system can know is the boundary of its own competence, and the most useful thing it can say at that boundary is: I couldn’t evaluate this, and here’s why. Calibration is worth more than coverage. I’d rather ship you five sections I’d stand behind and one honest gap than six sections where an unknown one is fiction.

The happy ending, most of the time, is mundane. You fix whatever blocked the load — loosen the bot rule, repair the crashing script, restore the server — run the audit again, and the section fills in with ordinary findings about headlines and alt text. The gap closes and the report becomes boring again. But it closed because something real got fixed, usually something that was quietly costing you visitors the whole time. That’s what unavailable means. Not “we have nothing to tell you.” Rather: the most important thing we have to tell you is that we couldn’t see the page — and you should find out who else can’t.