Yesterday’s report looked fine. Today’s contains a warning. Your supplier says, “We haven’t changed anything.” It is tempting to decide that either the tool or the supplier must be wrong. But both statements could be accurate.
A monitoring service repeatedly collects information and applies rules to it. A different result can therefore come from a different website, a different response, new outside information or a changed rule. Understanding which of these happened is more useful than watching the colour of the result.
To see why, follow one warning through two reports. Suppose the first report says an expected page instruction was present, while the second says it could not be inspected. These are invented report entries used to explain the comparison, not measurements of a provider. The tempting story is “the supplier removed protection.” But that story skips how the answer was collected.
A browser is the program used to open web pages. It sends a request, a message asking the website’s receiving program for a page. That program sends a response back. The response can include the displayed content and accompanying information called headers; some headers instruct the browser how to handle the page. A reporting tool may obtain that response too. This is the material its rules examine. 9
A useful comparison has two layers. First compare the material obtained: was it the intended page, the same destination and the same kind of visit? Then compare what the rules did with that material. If you start with the final score, those two layers have already been mixed together. A colour can tell you where to look; it cannot tell you which layer changed.
The same address does not settle the comparison
Think of comparing photographs of a shop. A photograph taken while its shutters are down tells you something different from one taken during opening hours. The building has not necessarily changed. Something similar can happen when a tool receives an access challenge instead of the page it expected.
For example, suppose an earlier report inspected the supplier’s product page, but a later attempt reached only a message asking the visitor to prove they are human. The second attempt has less information about the product page. Calling it an improvement or deterioration would miss the first issue: the observations are different.
Start by comparing the complete page addresses and any destinations reached through redirects. Then check whether the tool was signed in and whether the same checks completed. A public page and a logged-in page should not be treated as the same observation just because they share a logo.
Sometimes you receive a saved copy
A browser does not always need to fetch a newly produced copy of everything on a page. It can keep a response and reuse it under rules set for that purpose. This is called caching. Other services can also keep copies and deliver them to multiple visitors. The intention is to avoid repeating unnecessary work, not to make every visitor see an old page. 7
For example, the supplier corrects an instruction on its public page. An intermediary service still holds an earlier response that its reuse rules permit it to send. The next visitor’s request may receive that saved answer rather than a newly produced answer from the supplier’s own web service. A check of the saved answer describes what was delivered on that visit, even though the supplier’s underlying setting has changed. This is how caching can create an apparent disagreement without either person deliberately misrepresenting the change. 7
Ask someone able to examine delivery to explain whether this happened in your case. The explanation should connect the earlier answer, the correction and the later answer. “It is probably the cache” is only a suggestion. Likewise, “the address is the same” does not rule it out. A saved copy is not automatically bad or insecure; its suitability depends on what it contains and how it is managed. The lesson here is comparability, not a recommendation to disable all saved copies.
The practical lesson is simple: ask to compare what the two checks received. Repeating a check until the preferred result appears does not explain the difference. If the collection conditions remain different, keep that visible rather than combining the results into a smooth trend.
The judgement can change while the page stays still
Time matters even when nobody edits the website. Grove’s methodology names new information about software weaknesses, the end of a certificate’s validity period and new detection rules as possible reasons for changed results. These are explanations given by the vendor, not causes established by testing its product for this article. 1
A certificate is a digital identifying document used when establishing a protected connection. A checker can therefore raise a new validity question even if the page’s text has not changed. The document’s role in the connection is separate from the page’s content. 11
Take the new-information case. A report may have identified a particular software product. Later, an advisory describes a weakness in that product. The tool now has a reason to raise a question it did not raise before. The supplier still needs to establish whether the identification is right and whether the actual installation is affected; an external version label alone does not settle those details.
In that software example, distinguish a product name from an established weakness in this particular installation. A visible label can lead the tool to a product and a published advisory. It does not show every relevant setting or correction applied by the operator. A person with knowledge of the installation must connect the advisory to what is actually running. Grove describes version-based vulnerability matching; the report is a lead for that conversation, not an independent reproduction of an attack. 1
This gives monitoring a useful purpose even when the website has not changed. It can bring newly available information to someone who needs to assess it. You are not paying for the fiction that yesterday’s reassurance lasts forever. You are arranging to notice when the reasons for that reassurance may need revisiting.
A rule change is different again. Suppose the report author adds a check that was not previously included. The new warning may reveal an existing condition, rather than something that has just gone wrong. Preserve the earlier report with its original rules. If old information is judged again using new rules, describe that as a new interpretation of old information—not a fresh observation.
The reverse matters too. A warning disappearing might mean the supplier fixed the problem. It might instead mean the check was disabled, could not complete or was reclassified. A useful improvement note says what happened, not merely that the warning vanished.
Monitoring needs a person who can explain the change
A monitoring subscription is only part of the arrangement. Decide who receives a material alert and who contacts the supplier. If that person is absent, somebody else needs to know what to do. You do not need an elaborate process; you need an alert to reach someone able to investigate it.
Ask for a short explanation that connects the report with the supplier’s own information. NIST (the US National Institute of Standards and Technology)‘s Cybersecurity Framework includes combining information from several sources when analysing potentially adverse events. For a small business, that means not deciding the whole story from a single outside warning. 2
There are different useful outcomes. The supplier may confirm a defect and correct it. The report author may explain a collection failure. Or the finding may remain uncertain and need specialist review. A potentially serious but uncertain finding deserves prompt attention, not automatic dismissal or an instant accusation.
Finally, keep website monitoring separate from checking that invoices actually move through your service. A healthy product page does not show that a submission from your accounting software completed. If you need a demonstration of that journey, arrange it with the supplier using test accounts and made-up invoices. A notification gives you a reason to ask; it does not give you permission to experiment on live customer systems.