Home › Topics › Is It Us or Is It Them? Monitoring SaaS Reachability › SaaS Dependency Inventory

Your SaaS Dependency Inventory: Knowing What You Can't Control


Somebody walks up to your desk and says "email is down." You open the mail service in your browser and it loads fine. For them it still fails. Is it their laptop? The office Wi-Fi? The internet link? The identity provider that signs everyone in? Or is the email service itself having a bad afternoon in a data center you will never see? Every one of those is a different fix, a different phone call, and a different message to the company. You cannot tell them apart until you know what "email" actually is: which hosted service, reached how, depending on what.


That is the job of a SaaS dependency inventory. SaaS (software as a service) is the general term for applications you rent and reach over the internet instead of running on your own servers. Examples include your email service and the CRM (customer relationship management system, where sales keeps its contacts). Others include the hosted file service, the chat tool, and the payroll system. A dependency inventory is a plain list of every one of them, plus the things each one silently needs in order to work. It is not glamorous. It is the single document that turns "is it us or is it them?" from a guess into a procedure. I spent years answering that question from memory, and memory is a poor place to keep a hostname.


By the end of this guide you will know what belongs on the list. You will know how to rank services by how much the business hurts without them. You will learn to record the way users actually reach each one. You will also learn which hidden shared dependencies make ten unrelated services fail at once. You will have a table template you can fill in this week and a method for keeping it from going stale.


In this guide

  • Why the inventory comes before the monitoring
  • What belongs on the list
  • Tiers: how critical is it, really?
  • Access paths: how users reach each service
  • The hidden dependencies every SaaS shares
  • The inventory table template
  • Keeping it current
  • Using the inventory to decide what to check
  • Where to go next

Why the inventory comes before the monitoring


It is tempting to skip straight to setting up checks. Resist it. A check without an inventory answers a question nobody asked. You will end up watching a hostname that turns out to be the vendor's marketing site. Meanwhile, the login endpoint your users actually hit goes unwatched. When something breaks you will not know which of your checks should have caught it.


The inventory does three things that no amount of monitoring can do on its own. First, it makes the list of things that can break finite. Second, it records the dependencies — the things a service needs that are not the service itself. Many of the SaaS outages you will see are not the vendor's application failing. They are DNS (the Domain Name System, which turns a name into an address), sign-in, certificates, or the path between you and them. Third, it tells you who to call. Every hosted service has an owner, a support channel, and an account identifier the vendor will ask for. Discovering those during an outage is miserable; discovering them on a quiet Tuesday is paperwork.


You cannot isolate what you have not listed. Everything else in this pillar assumes the list exists, so build the list.


What belongs on the list


Start with the obvious services, the ones users would name if you asked "what do you use all day?" Then keep going, because the obvious list is usually about half of the real one. A useful way to sweep is to walk through the business one function at a time and ask "where does this actually run?"


  • Communication: your email service, the chat and video-meeting tool, the hosted phone system.
  • Documents and data: the hosted file service, the office document suite, the e-signature service.
  • Line of business: the CRM, the accounting system, the HR and payroll platform, the ticketing tool, and whatever industry-specific application the company could not function without.
  • IT's own services: the identity provider, the device management platform, the hosted backup target, the password manager, the domain registrar and DNS host, and the certificate authority you buy from.
  • Customer-facing: the website host, the online store or payment processor, the support portal, the email delivery service that sends transactional messages.
  • Developer and integration: the code repository, the build service, and any API that another one of your systems calls on a schedule.

Several of these are things the business does not think of as "apps" at all. Nobody logs in to the DNS host every morning, but if it fails, every hostname you own stops resolving. The identity provider is invisible right up until it is the only thing anyone can see. Those quiet, foundational services belong on the inventory precisely because nobody would think to blame them.


For each service, capture a small, consistent set of facts. Keep the fields few so the list actually gets filled in:


  • Service name and purpose in one plain sentence.
  • Business owner: the person or team who chose it, pays for it, and can decide "we will work without it for a day." Usually not IT.
  • Technical contact: whoever holds the admin login. Often that is you; write it down anyway.
  • Tier: how badly the business hurts without it (next section).
  • Access paths: how users and systems reach it, and which hostnames each path talks to.
  • Depends on: the identity provider, DNS, other services it integrates with, and the cloud platform it runs on if the vendor publishes that.
  • Vendor support: record the support portal or number, your account or tenant identifier, and the support plan you actually pay for. A tenant is your organization's private slice of a shared service.
  • Status source: the vendor's status page and how you subscribe to it.

Tiers: how critical is it, really?


Not every service deserves the same attention, and pretending otherwise means nothing gets watched properly. A tier is a simple label for how much the business suffers when a service is unavailable. Three tiers are enough for almost any organization; more than that and people spend their time arguing about the boundaries.


The test for each service is a single question: if this is unavailable for four hours on a normal weekday, what happens? Ask the business owner, not yourself. Their answer is usually more honest than the tier you would have guessed, in both directions. The expense tool you assumed was trivial turns out to block payroll on the last day of the month. Meanwhile, the chat tool everyone loves has a perfectly good workaround called the telephone.


TierFour hours down means...Typical examplesWhat it earns
1Work stops or revenue is lost; no workaroundIdentity provider, email service, phone system, the main line-of-business app, payment processingSynthetic checks from several vantage points; immediate alerts; a communication plan
2Painful, but a workaround exists for a dayCRM, hosted file service, ticketing, video meetingsReachability checks from the office; alerts in business hours; status-feed subscription
3Annoying; nobody escalatesTraining portal, expense reporting, feedback-form tools, internal wikiStatus-feed subscription only; investigate when users report

Tiers are also an argument-ending device. "It is Tier 3, and here is the four-hour test" is a better answer than "we did not get to it." And a tier can move: a Tier 3 service becomes Tier 1 the day a critical workflow gets built on top of it. That is one more reason to review the list on a schedule.


Access paths: how users reach each service


"Can we reach it?" sounds like one question. It is usually three or four, because the same service is reached in different ways by different parts of your organization. Those paths fail independently. Each path is a separate thing to check.


The browser

The plain web application, opened in a browser on a user's machine. This path runs through the user's DNS resolver, your firewall or web filter, your internet link, and the vendor's front door. It is the easiest path to test and the one most users mean when they say "it's down."


The client application

A desktop or mobile program that talks to the service in the background: the file-sync client, the desktop email client, the softphone. These often use different hostnames than the web app. They may refuse to talk through a web proxy that inspects traffic. They may keep working on cached data for a while after the service becomes unreachable. That hides the outage until the cache goes stale. The classic ticket is "the website works but the sync client shows an error." The explanation is almost always that the two paths are not the same path.


The API from your servers

An API (application programming interface) is the door a service opens for other programs rather than people. Your own servers use it: the nightly job that pushes orders into the accounting system, the backup job that writes to hosted storage. Server-side paths have their own DNS resolver, their own outbound address, and sometimes their own firewall rules. When one breaks, no user complains; a report is simply wrong the next morning.


At Tollgate Foods the payroll platform was Tier 1, owned, watched, and on every list. What was on no list was the hosted scheduling tool its monthly run pulled approved timesheets from. A depot manager had signed up for it on his own card. It fed payroll through an API call nobody in IT had ever seen. The month the card expired the vendor suspended the tenant, and payroll ran a day late. The first IT heard of the service was the finance director asking why. The review called it "an undocumented integration". The scheduling tool now has a row, an owner, and a tier nobody expected it to earn.


The hostnames behind the path

Most hosted services are not one hostname. The login page lives on one name and the application on another. Static content lives on a CDN (content delivery network — a globally distributed cache the vendor rents). Uploads live on yet another name. Vendors usually publish the hostnames and ports you must allow through your firewall; that document is a map of what your checks should watch. Record the names your users actually hit; a CDN hostname that five of your Tier 1 services all use is itself a dependency worth a row.


The hidden dependencies every SaaS shares


Every hosted service you use, no matter who makes it, stands on the same handful of foundations. When one of them wobbles, a dozen "unrelated" services fail at the same moment, and the pattern is the clue. Put these on the inventory as services in their own right, and mark every other row that depends on them.


DNS

Nothing reaches a hosted service without first turning its name into an address. So every row on your list depends on your DNS resolvers working and on the vendor's DNS host working. A resolver outage looks like "the whole internet is down." A vendor-side DNS failure looks like one service vanishing while the rest of the world is fine. Because DNS gets blamed for everything, it is worth being able to prove quickly whether it is at fault; Proving It Isn't DNS shows how.


The identity provider

Your services may use single sign-on (SSO — one login that every application trusts). If they do, the identity provider (IdP — the service that performs that login) is the most important row on the entire inventory. One IdP outage makes every application look down at once, even though all of them are running perfectly. Users report it as "I can't get into anything," and the natural instinct is to open five vendor status pages, all of which are green. (They are telling the truth. None of them is the thing that broke.) Mark the IdP as Tier 1, mark every SSO-connected service as depending on it, and test it first when several things fail together. For walking the sign-in chain itself step by step, see Auth Path Troubleshooting.


TLS and certificates

TLS (Transport Layer Security) is the encryption behind the padlock. It depends on certificates that expire and on both ends agreeing to trust each other. A vendor can let a certificate lapse. A firewall that inspects encrypted traffic can present its own certificate that a client application refuses. A device with stale trusted roots can reject a valid new certificate. A machine with a wrong clock can decide every certificate is invalid. TLS problems tend to hit client applications harder than browsers, one more reason to record access paths separately.


The internet path

Between your users and the vendor sit several things. These include your LAN (local area network), your firewall and any web filter, and your ISP (internet service provider). Then comes a chain of networks you do not own. It is not one thing: a branch office can lose its link while headquarters is fine. A single upstream network can have trouble reaching one vendor while every other destination works. Record which sites reach the internet through which links, so that "everyone at the branch" immediately points somewhere.


Vendor-to-vendor dependencies

Your SaaS runs on someone else's cloud. Most hosted services are built on one of a small number of large cloud platforms, use a CDN they do not own, and rely on a hosted DNS service. When a big platform has a regional incident, several services on your list fail together even though none of them can do anything about it. You cannot always learn a vendor's platform, but many publish it. Where you know it, write it down. A column that reads "runs on Platform A, region East" across six rows explains a lot on the day those six rows go red together.


The inventory table template


Keep the inventory somewhere that survives an outage of the thing it describes. A spreadsheet in your hosted file service is a fine working copy. But keep an exported copy on a local share or a printed page in the server room too. An inventory you cannot open because the file service is down is a joke that stops being funny at 9 AM.


ServiceOwner / tech contactTierAccess pathsHostnamesDepends onVendor support / tenant IDStatus source
Identity providerIT / you1Browser (login page), every SSO appLogin name, tenant portal nameDNS, internet path, vendor's cloud platformPortal; business plan; tenant ABC-123Feed + email
Email serviceOffice manager / you1Browser, desktop client, mobile app, inbound mail via MXWebmail name, autodiscover name, MX hostsIdP, DNS (MX and autodiscover records), TLSPortal; tenant XYZFeed
CRMSales director / CRM admin2Browser, mobile app, nightly API sync from the accounting serverApp name, API name, CDN nameIdP, DNS, email service (notifications)Portal; standard planFeed
Hosted file serviceOperations / you2Browser, sync client on every laptopWeb name, sync API name, download CDNIdP, DNS, TLS (sync client is strict)Portal; tenant XYZFeed

If you prefer plain text — it compares well between versions and lives happily in a repository — the same template as a comma-separated file looks like this:


service,owner,tech_contact,tier,access_paths,hostnames,depends_on,support,tenant_id,status_source
identity provider,IT,you,1,"browser;sso",login.idp.example.com;portal.idp.example.com,"dns;internet;cloud platform A",portal,ABC-123,feed+email
crm,sales director,crm admin,2,"browser;mobile;api from acct-srv-01",app.crm.example.com;api.crm.example.com,"idp;dns;email",portal,,feed

Whatever the format, the columns matter more than the tool. Owner and tier drive who you call and how loudly you alert. Access paths and hostnames drive what you check. Depends-on drives the order you check things in.


Tip: fill in the "depends on" column from the service's point of view, not yours. Ask "what has to work before this service can work?" and list DNS, the identity provider, and the internet path explicitly on every single row. The repetition is the point: when nine rows share the same dependency, the inventory itself tells you where to look first.


Keeping it current


An inventory decays the moment you stop tending it. New services arrive constantly, usually without telling IT, because signing up for a hosted tool takes a credit card and a lunch break. That is shadow IT: services in use that IT never heard about. It is not a crime; it is a row that needs an owner. These are the places it turns up.


  • Procurement. Ask for one question to be added to the purchasing process: "Is this a hosted service, and who in IT knows about it?" One question, early, prevents most surprises.
  • Expense reports. Finance sees every subscription you do not. A quarterly list of recurring software charges, however rough, is the most reliable shadow-IT detector in the building.
  • The identity provider's app list. If applications are connected through SSO, the IdP already maintains a partial inventory. Anything it knows about that your list does not is a row to add.
  • Your own logs. DNS query logs, firewall logs, and web-filter reports show which hosted destinations your network actually talks to. A monthly "top external destinations" report reveals the file-sharing tool a department adopted without telling anyone. And it is the one detector that does not depend on anybody remembering to tell you.
  • Offboarding. When a service is cancelled, remove the row and its checks; alerts about a service nobody uses are pure noise.

Review the whole inventory with its owners on a fixed schedule: quarterly is realistic, twice a year is the minimum. The review is short: is the tier still right, is the owner still here, has the access path changed, is there a new dependency? Skip it for a year and the owner column names people who have left, and the tier column describes workflows that have since moved somewhere else.


Using the inventory to decide what to check


Read the finished inventory top-down and it tells you what to monitor, in what order, and how hard.


Start with the shared dependencies, because one check there covers many rows. Check that your DNS resolvers answer and that the identity provider's login page loads. Check that your internet link is up and reasonably fast. Between them, these checks will explain most "everything is down" moments. They are also cheap: three checks that protect thirty services.


Then work through the rows by tier. Tier 1 services get full synthetic checks. These are scripted tests that behave like a user. They run on a schedule from the places users sit. They cover each recorded access path, from more than one vantage point, and alert immediately. Tier 2 gets a reachability check from the office and a subscription to the vendor's status feed. Tier 3 gets the status feed and nothing else, until a user reports something. The synthetic checks guide explains how to build those tests level by level. The isolation playbook is what you run, inventory in hand, when one of them goes red.


Finally, use the access-path column to avoid the most common blind spot: checking only the browser path. If the file-sync client and the nightly API job are on the inventory, they deserve their own checks. The browser test will cheerfully pass while both of them are broken.


Where to go next


The inventory is the map: what the business runs on, who owns it, how it is reached, and what it stands on. Every other guide in Is It Us or Is It Them? Monitoring SaaS Reachability assumes that map exists. The checks, the status-page reading, the isolation sequence, and the outage communications all point back to it. Build the list first. Everything after it gets easier. The next time somebody walks up to your desk and says "email is down", you will know what email is. You will know which service, reached how, depending on what.


Put this into practice with SysPeep Monitor

The inventory is a document, and no software will write it for you. But once you know which hosted services matter, SysPeep Monitor can watch your side of the road to them. That means the DNS servers, the internet link, and the firewall. It uses a central Windows service plus lightweight probes at each branch. Its dependency-aware alerting tells you when one upstream failure explains ten red rows. The free trial installs in minutes.

Download Free TrialLearn About SysPeep