You open a report about your invoicing supplier. There are coloured boxes, unfamiliar abbreviations and a reassuring overall result. Where should you start? Not with the biggest number. Start with the small line that says which address was checked.

A report is rather like a vehicle inspection sheet: the result is useful only when you know which vehicle was inspected and what the inspection covered. A website report needs the same reading discipline. The supplier’s public homepage is not automatically the service through which your accountant sends invoices.

Before reading the warning, picture how the tool obtained anything to read. Your browser is the program in which you open websites. It asks another computer’s web-serving program for a page; that answering side is called the server. The request identifies what is wanted. The response carries a result and information about it, sometimes including the page itself. A reporting tool can collect such an answer too. It does not learn everything about the business simply by knowing its web address. 9

For our example, keep three moments separate. The tool obtains an answer. A checking rule looks for something in that answer. The report turns the rule’s result into a warning and perhaps a score. If the first moment did not produce the intended page, or the second rule was not run, the final box cannot reasonably be read as if the complete inspection happened. This sequence is the useful thread through a confusing report.

First find the page behind the company name

For example, a report might concern the supplier’s product page, while customers sign in at a different address. The product page explains prices. The customer portal handles accounts and documents. The software connection from your accounting program may be a third service. A finding on one does not automatically describe the others.

Look for the complete page address, the observation time and whether the tool signed in. If the address led somewhere else, the destination matters too. These details explain what the report actually looked at. Without them, you cannot tell whether a warning concerns your daily work or a separate public page.

Grove’s published methodology describes several approaches: reading public responses, running pages in a browser, consulting outside databases and optionally sending test requests deliberately designed to investigate how the service behaves. These are the vendor’s descriptions, not capabilities independently tested for this article. Nor does a list of available checks tell you which ones ran in your report. 1

Read one finding from beginning to end

Here is an illustrative finding, not a real scan result: “An expected browser instruction was not found in the response from the product page.” Websites can send instructions alongside a page; these are called response headers. Some concern how a browser should handle that page. Grove describes checks for such instructions in its methodology. 1

A response header is not necessarily something you see printed above the page. It is information supplied for the receiving program to interpret. The displayed page and these accompanying instructions are different parts of the answer. This explains how a page can look perfectly ordinary to a visitor while the reporting tool raises a question about its instructions. It also explains why the tool needs the response, rather than a screenshot of the supplier’s logo, to examine them. 9

To read our illustrative line, imagine the report showing the address of the product page, the time it was fetched, the relevant instruction it looked for and the answer it obtained. You can then trace the conclusion: this rule expected that instruction; this collected answer did not contain it. That still leaves the rule’s suitability and the business consequence to explain. “No instruction found” is a narrower statement than “someone can read customers’ invoices.” The report has not followed an invoice or investigated who may open it.

Make the instruction concrete: a browser can display one page inside another. A page can tell the browser not to embed it that way, using the header named X-Frame-Options. A rule might warn when it finds no such header. However, the page can send a different browser instruction that restricts embedding instead. The missing header therefore need not mean missing protection. Why restrict embedding? A misleading surrounding page could disguise where someone is clicking and induce an unintended action on the embedded page. Preventing unauthorized embedding helps address that danger; it is a different protection from deciding which invoices an account may read. 10

The useful part of our example is the observation: an instruction was not found in a particular response. The warning label is the tool’s interpretation of that observation. The business question comes next: what protection was expected here, and does its absence matter on this page? The answer may call for a correction, or it may reveal that the rule was not suitable for the situation.

Now change one detail in the example. Instead of the product page, the tool received a “prove you are human” challenge. It has not inspected the response it intended to inspect. “Could not check” is different from “checked and missing.” Grove’s methodology expressly describes unreadable security-header checks on blocked responses as unverifiable rather than absent. 1

For the same reason, an empty list of warnings is only reassuring if the relevant checks actually completed. A disabled check, an unavailable database or an interrupted request cannot tell you that the thing being checked is in good order.

Colours help sort the work, not explain it

Reports may cover page speed, search visibility, accessibility and security. These are worthwhile but different subjects; Grove describes multiple assessment categories. A better page description for a search engine does not show that customers’ invoices are better protected. Read the category before interpreting an overall result. 1

A severity label is also a starting point. It says how seriously the tool classifies something under its rules. You still need to connect that observation to the service. A warning about the public brochure page and a problem with signing in need different conversations, even if both have the same colour.

A repeated, unexplained website defect might make you question how carefully maintenance is managed. That is a reasonable question to investigate, not proof that the invoice service has the same defect. Keep the distinction when forwarding the report to a colleague.

If the report instead describes looking up information in an outside database, the chain is different. The tool asks that database for a record and places the returned information in the report. You need to know what the database recorded and when; the report’s creation time is not automatically the time of the original observation. Grove declares outside lookups among its methods. It would be misleading to read every reported line as a fresh direct inspection of the supplier. 1

The useful response from a supplier is therefore specific. In our header example it might explain the public-page setting and a planned correction, or explain why the collected response was not the ordinary page. Neither answer settles separate questions about the invoice portal. Those need their own explanation. You can keep investigating a troubling public defect without pretending that every service using the company’s name shares its software or maintenance team.

Turn the finding into an understandable question

Ask the supplier: “This report says it received this response from this page. What does that mean for our use of your service?” A useful answer explains the finding in ordinary language, identifies any correction and says how the result will be checked again. “The score is better now” leaves the mechanism unexplained.

Also ask how the information was collected. Looking in someone else’s published database is different from contacting the website. Loading a page normally is different from deliberately changing requests to investigate its behaviour. The word “passive” should not hide those differences, and a public website does not give you permission to probe customer accounts.

For questions about access to invoices, arrange a separate supplier-authorized demonstration in a separate test environment using made-up documents and temporary test accounts. A website report alone cannot answer them. This outcome-first approach is consistent with NIST (the US National Institute of Standards and Technology)‘s Cybersecurity Framework, which describes desired security outcomes without prescribing one implementation. 2

Before your next meeting, choose one finding and explain it without its score: which page was checked, what was seen, what could not be checked and why it matters to your business. If that explanation is clear, the report has become useful—even if some questions still need answers.

Sources