Home › Topics › Is It Us or Is It Them? Monitoring SaaS Reachability › Status Page Skepticism

Status Pages: Trust, but Verify


Every hosted service has a status page: a public web page with a row of green dots. It says, in effect, "everything is fine, it must be you." On a good day it is a helpful courtesy. On a bad day it is the most frustrating page on the internet. The email service has been failing for twenty minutes, your users are lining up at your desk, and the dot is still green. You refresh. Still green. Eventually a yellow "investigating" appears, roughly when the outage is already old news in your building.


None of that means status pages are useless or dishonest. They are what they are: a human-written summary of a vendor's view of its own service, published on the vendor's schedule, with the vendor's incentives. Once you understand why they lag and understate, you can use them for what they are good at. You can stop expecting what they cannot give and fill the gap with your own measurements. I learned to read them this way after more afternoons of refreshing than I would like to count.


By the end of this guide you will know why status pages behave the way they do and why you should subscribe to them anyway. You will learn to read one like an engineer rather than a customer. You will also learn to correlate it with your own synthetic checks to reach a verdict. Finally, you will learn to build the one status page your users actually need: what is slow or down from here.


In this guide

  • Why status pages lag and understate
  • Subscribe anyway, and pipe it into context
  • Reading a status page like an engineer
  • Correlating their page with your checks
  • Building your own "our view" status page
  • The honest limits of a status page
  • Where to go next

Why status pages lag and understate


A status page looks like a monitoring dashboard. It is not one. Understanding the difference explains almost everything about how it behaves.


It is updated by humans

A monitoring dashboard changes when a check fails. A status page changes when a person decides it should. Inside the vendor, that means an alert fires and an engineer confirms it is real. Someone decides it is customer-visible, someone writes a sentence, and sometimes someone approves the sentence. Each step is reasonable; together they add up to a delay that is measured in many minutes at best. Your users do not experience that delay. They experience the outage from its first second.


Green is the default

A status page is green until somebody changes it. That means a green page is not evidence of health; it is evidence that nobody has updated the page. A truly stale page, one nobody is watching at 3 AM on a holiday, is indistinguishable from a healthy one. Read "operational" as "no declared incident," not as "we checked and it works."


Outages are usually partial

Large hosted services run in many regions, on many clusters, behind many load balancers (the machines that spread traffic across many servers). Most incidents hit one slice: one region, one data center, one feature, one group of tenants. A tenant is one customer's private slice of the shared service. From the vendor's seat, the service is mostly fine, and the page says so. From your seat, if your tenant lives in the affected slice, the service is entirely down. A single row of dots cannot express "fine for most people, gone for you," so it rounds toward fine.


The vocabulary is soft on purpose

"Degraded performance." "Some users may experience intermittent issues." "Elevated error rates." These are not lies, but they are chosen words. A vendor's service level agreement (SLA) is the contractual promise of how available the service will be, with credits owed if it is missed. It is usually tied to declared outage minutes. Declaring a "major outage" starts a clock that costs money; declaring "degraded performance" often does not. Add a communications team that would rather not alarm prospects reading the page, and the gravitational pull is always toward the mildest true description.


Their monitoring runs from their side

The vendor's own checks run from inside or next to their platform. They will catch a crashed application quickly. They will not catch a broken route between your ISP (internet service provider) and their region. Nor will they catch a DNS problem affecting one part of the world. That is a failure of the lookup that turns a name into an address. They also will not catch a content-delivery node near you (the vendor's cache server closest to your users) serving errors. Those are exactly the failures that look, to the vendor, like nothing is wrong.


Put those together and the status page's behavior is not a mystery. It is late because people write it, green because green is the default, and calm because calm is cheaper. It is blind to your path because it never looks down your path. Now you can use it properly.


Subscribe anyway, and pipe it into context


A late, understated update is still an update. When it arrives it carries information you cannot get anywhere else: the vendor's official acknowledgment and the components involved. Eventually, it carries the timeline you will need for the follow-up conversation. So subscribe to every status page on your dependency inventory. Nearly all offer some combination of an RSS or Atom feed (a machine-readable stream of posts your tools can poll), email notifications, and text messages. Sometimes they offer a webhook: an address of yours that the vendor sends a message to the instant something changes. It posts updates into a chat channel or ticketing system.


The trick is where the updates land. A vendor email in a mailbox nobody reads during an incident is worthless. Instead, route status updates to wherever your own alerts already appear. That could be the operations chat channel, a note on the open ticket, or an annotation on the monitoring timeline. The goal is simple. When your own check for the CRM (the customer relationship management system) goes red, the vendor's "investigating elevated errors" post sits right next to it. The person looking at the alert sees both at once. That is context: information that helps a human decide what an alert means. A status update is context, not an alert on its own; treat it that way and it will never become noise.


  • Tier 1 services: subscribe by feed and email, filtered to the components and regions you use, delivered into the operations channel. Consider a text message for the identity provider, because when it fails the channel may be unreachable too.
  • Tier 2 services: feed into the channel; email to the service's technical contact.
  • Tier 3 services: feed only, into a low-priority channel nobody is required to watch.
  • Scheduled maintenance notices: route these to a calendar or a separate channel, so that a planned window does not look like an incident at a glance.

One practical warning: some vendors host their status page on the same platform as the service, so it goes down with the service. Others host it with a third party specifically so that it survives. A status page that goes down with the service is at least consistent. Either way, bookmark the page's address and subscribe by a channel that does not require you to visit it.


Reading a status page like an engineer


A customer looks at the dots. An engineer reads the structure. Three questions turn a status page from decoration into data.


Component, region, or global?

Most pages break the service into components (login, web app, the API that other programs call, mobile, email delivery, file sync). Many break it further by region (a geographic area served by a particular set of data centers). Find the components your users actually depend on and the region your tenant lives in. A red API in a region you do not use is trivia; a yellow login component in your region is an emergency. If the page does not show regions, check your admin console. Many vendors publish a private, tenant-specific status view inside the product that is more accurate than the public page.


Which phase of the incident?

Incident posts move through a standard sequence, and the phase is more informative than the color. Read the last two phases with care: "resolved" means the vendor considers it over, and your tenant may disagree for a while. The vocabulary varies slightly between vendors, but it usually looks like this:


PhaseWhat it actually meansWhat to do
InvestigatingThey have confirmed something is wrong but not what. Expect no estimate.Confirm your own impact; tell users it is acknowledged.
IdentifiedThey know the cause. A fix is being worked on. Sometimes an estimate appears.Relay the estimate with a caveat; keep your checks running.
MonitoringA fix is deployed and they are watching for recurrence. Your region may still be catching up.Verify with your own checks before telling users it is fixed.
ResolvedThey consider it over. Your tenant may still be affected.Run the post-resolution checks; note the timestamps for the record.
Post-incident reportA written explanation published days later, describing cause and prevention.Read it. It feeds your review with the service's business owner.

What does the history say?

Nearly every status page keeps an incident history. Spend ten minutes with it before you ever need it. How often does this service declare incidents, and how long do they last? How quickly does the first post appear after the incident start time they later admit to? Do they publish post-incident reports? A vendor whose history shows honest, prompt posts earns more trust during the next incident. That beats one whose page has been solid green for years while your ticket history says otherwise. Also note the time zone the page uses; a surprising number of confused timelines come from comparing a vendor's timestamps with your own without converting.


Correlating their page with your checks


Now combine the two sources. Your synthetic checks say whether the service works from where your users sit, and from an outside point if you have one. The status page says what the vendor is willing to admit. Put them side by side and four situations fall out.


Your checksTheir pageMost likely meaningNext move
FailingGreenEither their page is late, or the problem is on your side of the internet. The vantage points decide: if the outside point fails too, they are late; if only your sites fail, it is you.Compare vantage points before assuming either. Open a vendor ticket with your evidence if the outside point agrees with you.
FailingRedConfirmed vendor incident that affects you.Stop troubleshooting the network. Communicate to users, track the incident, keep measuring.
PassingRedYour tenant is in a lucky region, the affected component is one you do not use, or the vendor is being cautious.Warn users lightly that the vendor has a declared incident; watch your checks more closely for the next few hours.
PassingGreenFine — or your checks do not exercise the feature that broke. A user report that contradicts both is a signal to look at what your checks miss.Ask the reporting user exactly what fails; consider adding a check for it.

The first row is the one that matters, and it is where the rest of this pillar comes in. "Failing here, green there" is an argument only if you have a second vantage point outside your own network. Without one, you and the vendor are simply two parties each insisting it is the other. With one, the argument ends: the outside point failing means the problem is bigger than your building. The outside point passing means the road from your building is the problem. The synthetic checks guide explains how to set up those vantage points. The isolation playbook walks the comparison in order.


Whatever the row, write down the timestamps. Note when your first check failed, when the first user reported, and when the status page first changed. Add when it declared resolution and when your checks actually recovered. The gap between your first failure and their first post is the single most useful number you will bring to the review afterward. It is only available if you record it while it happens.


At Harbourview Clinic the appointment system stopped loading at 14:05 on a Thursday, and the vendor's page was green. The two administrators spent the next forty minutes on the firewall, the web filter, and the DNS forwarders. To them, green meant "it must be you," and nobody had an outside vantage point to say otherwise. The page stayed green until the service came back at 15:20. The next morning it showed an amber incident, "degraded performance for a subset of customers", with a start time of 14:00. The clinic now has an outside probe, and a rule that forty minutes on the firewall requires some evidence the firewall is involved.


Tip: when your checks fail and the page is green, resist the urge to refresh the page every thirty seconds. Refreshing does not make it update faster; it just makes you feel busy. Spend those minutes comparing vantage points instead. By the time the yellow dot appears, you can already have a verdict and a user message written.


Building your own "our view" status page


Here is a question your users ask constantly and no vendor page can answer: "Is anything broken for us right now?" The vendor knows only about the vendor. Your monitoring knows about every service on the inventory, from every site. It also knows the shared pieces nobody else sees: the internet link, the DNS resolvers, the identity provider path. That makes you the only party who can publish the view your users want. Call it the internal status page, and keep it small.


The content is a short list of the Tier 1 and Tier 2 services from your inventory. Give each a status from your own checks and the site it applies to if you have several. Add a one-line plain-language note and the time of the last update. Something like this:


IT service status — from our network — updated 09:41

Email service          OK        all sites
Identity provider      OK        all sites
CRM                    SLOW      main office only: pages taking 6-10 s since 09:12
                                 branch and remote users unaffected; we are investigating our link
Hosted file service    DOWN      all sites: sync and web both failing since 09:20
                                 vendor has declared an incident; next update from us at 10:00
Phone system           OK        all sites
Internet link (HQ)     OK
Internet link (branch) OK

Notice the language. "SLOW, main office only" and "DOWN, all sites, vendor has declared an incident" say more than any vendor's row of dots. They answer the question from the user's chair. They also quietly separate "us" from "them" without blaming anyone. The CRM line says we are looking at our link; the file service line says the vendor is on it. (Neither line mentions a subset of customers.) That framing is the subject of the outage communication guide; the status page is where the framing lives between messages.


A few rules keep it useful:


  • Generate it from checks, not from memory. The status column should come straight from your synthetic checks, updated automatically. A page you must remember to edit will be wrong precisely when you are busiest. Many monitoring tools can publish a simple read-only dashboard that serves this purpose directly. A small script that turns check results into a static HTML page is the do-it-yourself version.
  • Host it where the outage cannot reach it. Not in the hosted file service, not behind the identity provider it reports on. A plain page on a local web server, mirrored to a pinned message in the chat tool, is a good combination. If the chat tool is on the inventory too, the local page is the one that survives.
  • Show "unknown" honestly. If a check has not run or a probe is unreachable, say so. A page that only ever shows green or red will be believed less than one that admits uncertainty.
  • Keep the human note human. The automatic column says OK, SLOW, or DOWN. The note is written by you and says what it means and what happens next. Update the note when the situation changes, and put a time on it.
  • Tell people it exists. The page only reduces tickets if users check it before opening one. Link it from the help desk, the intranet home, and the auto-reply of the support mailbox.

The honest limits of a status page


Even read carefully and combined with your own data, a vendor's status page has hard limits. Be clear about them, so you never lean on it further than it can hold.


It cannot see your path. Nothing on the page will ever tell you that your ISP has a broken route to their region. Nor will it tell you that your web filter started blocking a new hostname or that your DNS forwarder is timing out. Those are yours to detect. It is coarse: one dot for a component used by many thousands of tenants cannot express your tenant's experience. It is late, for the human reasons above, and the delay is largest at the start of an incident, exactly when you need it most. It is biased toward calm, not because anyone is dishonest but because the incentives all point the same way. And it knows nothing about the dependencies between vendors. If your CRM is down because its identity provider or cloud platform is down, the CRM's page may say very little. In that case, the real story is on a page you were not watching.


So use the status page as a witness, not a judge. It can confirm what you have already measured, supply the vendor's timeline and vocabulary for your communications, and tell you when they believe they are done. It cannot tell you whether the service works for your users. Only your own checks can do that, which is why they come first and the page comes last.


Where to go next


A green dot is not evidence. You now know why, and how to subscribe so vendor updates land next to your own alerts. You know how to read components, regions, and phases. You also know how to combine their page with your checks into a verdict and publish the view your users actually need. The other guides in Is It Us or Is It Them? Monitoring SaaS Reachability supply the pieces this one depends on. These include the inventory of what to watch and the checks that produce your side of the evidence. They also include the ten-minute isolation sequence and the messages you send while it is all happening. The dot will go on being green until somebody changes it. The difference is that you will no longer be waiting for them to.


Put this into practice with SysPeep Monitor

A vendor's status page is their view of their service; the dashboard built from your own checks is yours. SysPeep Monitor shows what is reachable from the office and from each remote probe on live dashboards and maps in a single web console. Its alert history gives you the timestamps to lay next to the vendor's timeline when the review comes. The free trial installs on any Windows machine in minutes.

Download Free TrialLearn About SysPeep