Business Technology & SaaS · Definitions
What is SaaS, measured against the standard rather than the marketing
Software as a service has an official definition, published by the US National Institute of Standards and Technology in 2011 and running to seven pages. Not one of the ten pages ranking for this question cites it. Here it is, alongside the boundaries it draws and the ones it refuses to draw.
Reviewed August 2026 · The Insight Journal Editorial Team
In short
The definition
The definition almost nobody quotes
In short
Here is the definition in full, because paraphrasing it is where most of the confusion starts.
"The capability provided to the consumer is to use the provider's applications running on a cloud infrastructure. The applications are accessible from various client devices through either a thin client interface, such as a web browser (e.g., web-based email), or a program interface. The consumer does not manage or control the underlying cloud infrastructure including network, servers, operating systems, storage, or even individual application capabilities, with the possible exception of limited user-specific application configuration settings."
Read it twice and the last clause does most of the work. You get user-level settings, and nothing underneath them. That is the whole bargain, stated in a single sentence by a federal standards body, and it sits at the base of the wider business software landscape that most companies now buy.
The omission
What the standard leaves out, and why it matters
In short
This is not a pedantic point. If subscription were definitional, a tool billed per transaction would fall outside SaaS, and plainly it does not. The billing model is a commercial decision layered on top of an architectural one.
NIST's own companion document is specific about how the money actually works. Fees are charged, it says, "based on the number of users, the time in use, per-execution, per-record-processed, network bandwidth consumed, and quantity/duration of data stored".
Six fee bases, one category
- 1. Number of users
- 2. Time in use
- 3. Per execution
- 4. Per record processed
- 5. Network bandwidth consumed
- 6. Quantity or duration of data stored
Source: NIST SP 800-146, section 5.
The test
The five-part test hiding inside the definition
In short
| Characteristic | What NIST says it means | The sign it fails |
|---|---|---|
| On-demand self-service | You can provision it yourself, "without requiring human interaction with each service provider" | You had to email an account manager to add a user |
| Broad network access | Reachable over the network by "heterogeneous thin or thick client platforms" | It only runs on one managed desktop image |
| Resource pooling | Provider resources "are pooled to serve multiple consumers using a multi-tenant model" | A dedicated box was racked for you alone |
| Rapid elasticity | Capacity scales "rapidly outward and inward commensurate with demand" | Growth means a migration project |
| Measured service | Usage is metered, "monitored, controlled, and reported" | Nobody can tell you what you consumed last month |
A product that fails two or three of these is a hosted application. That is not an insult, and for some workloads it is the better arrangement. It is simply a different thing from what the standard calls SaaS, and knowing which one you are buying changes what you ask the vendor.
The three service models
SaaS, PaaS and IaaS, drawn on who controls what
In short
| Layer | IaaS | PaaS | SaaS |
|---|---|---|---|
| Physical servers and network | Provider | Provider | Provider |
| Operating system | You | Provider | Provider |
| Runtime and hosting environment | You | Partly you | Provider |
| The application itself | You | You | Provider |
| What you get to change | Anything you install | Your own deployed apps | User-level settings |
The older bargain
How this differs from software you used to buy outright
In short
What you stop carrying
- No install, no patch cycle, no hardware to size.
- New capability arrives without an IT rollout.
- Adding a person is an administrative act, not a project.
What you hand over
- Limited user-specific settings, and no control beneath them.
- The vendor's roadmap becomes your roadmap, including removals.
- Stop paying and access typically stops with it.
There is a fourth item most guides omit, and NIST files it under network dependence. Availability rests on a network neither side owns: in the public case, reliability "cannot be guaranteed either by the cloud consumer or by the cloud provider because the Internet is not under the control of either one".
An operator feels the difference twice. On day one it feels like relief. Somewhere in year three it feels like the morning a feature you built a process around is retired in a release note.
The architecture you are buying
Multi-tenancy, and what it actually costs you
In short
Favouring isolation
A separate copy for each client
The provider runs a separate application instance per client, with a separate database behind it, and the operating system provides the tenant isolation. NIST is blunt about the economics: the overhead of a separate copy and a separate database is incurred for every active client.
Favouring efficiency
One application, one combined database
The provider reengineers the application to serve many clients at once and stores everything in a combined database. NIST's verdict on the trade is a single parenthesis: this approach "lowers costs for the provider (but at an increased security risk to consumers)".
Where your data sits
The same characteristic settles the residency question. Under resource pooling, the customer "generally has no control or knowledge over the exact location of the provided resources but may be able to specify location at a higher level of abstraction (e.g., country, state, or datacenter)".
Country-level choice is often available. Street-level certainty is not, so data residency is a contractual question rather than a technical one.
Why leaving is harder than arriving
NIST lists portability between SaaS clouds as a named concern. Export and import formats "may not be fully compatible", and customised workflow, business rules, interface settings, support scripts, data extensions and add-ons "can also be provider specific and not easily transferable".
The records usually come out. Three years of accumulated configuration usually does not, and that gap is what vendor lock-in means in practice.
Named examples
What SaaS looks like across four categories
The SaaS examples people reach for span consumer and business tools, and the delivery model is the common thread rather than the industry. We name these as recognisable members of a category and make no claim about any product's pricing, features or architecture.
Collaboration and email
Google Workspace and Slack sit closest to the definition NIST wrote: hosted centrally, reached through a browser, provisioned by an administrator in minutes.
The customer record
Salesforce and HubSpot are the business-facing archetype. The record lives on the provider side, which is exactly the shift the standard is describing.
Finance and commerce
QuickBooks Online, Xero and Shopify moved categories that were once shipped on disks into the same delivery model, one company at a time.
Consumer services
Netflix and Dropbox are the examples people reach for first, and they are fair ones. The delivery model is what makes them SaaS, not the industry.
Two questions usually follow this list, and both have their own home. If you want the argument about whether large language models displace any of it, read our analysis of whether AI is actually replacing SaaS. If you are already running a handful of these and need the vocabulary that comes with them, start there instead.
Boundary problems
Where the word gets used loosely
In short
-
A login page is not a service model
Any web application with accounts now gets called SaaS. The standard asks a harder question: are the provider’s resources pooled, is capacity elastic, is usage metered. Fail those and you have a hosted application, which is a perfectly good thing to be.
-
Private cloud is a deployment model, not a rival to SaaS
NIST defines four deployment models, private cloud, community cloud, public cloud and hybrid cloud, and they sit at right angles to the three service models. A private-cloud SaaS product is coherent. "We are private cloud, not SaaS" is a category error.
-
One company can sell you both SaaS and PaaS
The same underlying technology often ships twice: as a finished application you log into, and as an interface developers build on. The first behaves like SaaS and the second like PaaS. The vendor name tells you nothing.
-
"Cloud" used where "hosted" is meant
NIST SP 800-146 quotes the loose working definition, "Software deployed as a hosted service and accessed over the Internet", and notes in the same paragraph that only the most advanced SaaS architectures appear to satisfy the NIST cloud definition. The gap is acknowledged inside the primary source.
None of this decides what you should buy. That question starts one level up, with the job the software is meant to do, which is where our guide to choosing business software picks the thread up.
The evidence for writing this at all
A settled term that people still cannot pin down
Search-demand and SERP measurements taken on 18 August 2026. They describe how many people ask the question and what they are shown. They are not market size.
27,100
US monthly searches for "what is SaaS", a term the industry treats as settled
DataForSEO, August 2026
0 of 10
Page-one results that cite NIST, or any standards body, for the definition they give
Live SERP analysis by The Insight Journal, 18 August 2026
7
Pages in SP 800-145, the document the whole cloud vocabulary rests on
NIST, 2011
Method
How we checked this
Read, not remembered
Both NIST publications were downloaded and read as text in full on 18 August 2026. Every quotation on this page was copied from those documents, not from a summary of them.
Measured, not assumed
The search volume and the page-one analysis come from a live SERP pull on the same date, United States, desktop. Four of the ten results are published by companies that sell cloud services.
What we leave out
No market size, growth rate or adoption percentage appears here, because none could be traced to a primary source we could check. No vendor price appears either. Our research and sourcing policy explains the rule.
Questions