What makes one Web applications “Software as a Service” (SaaS) and another a “plain old Web application” (POWA)? Or is there no such distinction?
Wouldn’t it be convenient if we had an answer that has some functional relevance? Here are the different axis on which I (unsuccessfully) tried to project Web applications to sort them between SaaS and POWA:
1) Amount of data
Hypothesis: If the Web application stores a lot of your data, it’s SaaS, otherwise it’s a POWA. For example, Webmails like GMail and Yahoo Mail have a lot. So do hosted CRM systems. They are all SaaS. Zillow and Expedia do not, so they are POWA.
2) Criticality of data
Hypothesis: Flickr and YouTube may have many Gigabytes of your data, but it’s not critical data (and you most likely have a local copy), so they are POWA. An on-line password manager, or a federated identification service, stores little information about you but it’s important information so they are SaaS.
Hypothesis: If you’ve invested time to customize the Web application (like Salesforce.com) then it’s SaaS. If, on the other hand, users all get pretty much the same interface, give or take some minor branding (e.g. YouTube or GMail), then it’s a POWA.
Hypothesis: If it takes many man-years to build it it’s SaaS, otherwise it’s a POWA. So Google Search is SaaS, but twitpics.com is a POWA.
5) Organization mandate
Hypothesis: If all your employees use the same Web application for a given task, that’s SaaS. For example, some companies mandate that all travel reservations go through American Express or Carlson Wagonlit so they are SaaS. If your company lets you use Expedia or any travel site of your choice and you just expense the bill, then it’s a POWA.
Hypothesis: If your company has its own administrator(s) to manage the use of the Web application by its employees, then it’s SaaS. Otherwise it’s a POWA. For example, Google Apps is SaaS, but Facebook is a POWA.
Hypothesis: If you pay for using the Web application, its’ SaaS. In that respect, LotusLive (and porn sites) are SaaS. Facebook and Twitter are POWA.
Hypothesis: If you have guarantees of service, and they are backed by some compensation, then it’s SaaS. For example Aplicor and AWS S3 are SaaS, but (last I heard) SalesForce.com is not.
Hypothesis: If the Web application conforms to your compliance needs, then it’s SaaS. By this metric, pretty much nothing is SaaS if you have any serious compliance requirement, it’s all POWA.
10) On-premise alternative
Hypothesis: If the Web application replaces something that you could do on-premise, then it’s SaaS. If it intrinsically involves a third-party provider, then it’s a POWA. Zoho (which replaces Office+Sharepoint) is SaaS, while Amazon (the store) and EBay are POWA. You couldn’t run your own version of them, at least not as a replacement for the public versions (you could run an internal auction site, but that’s a different use case from what you use EBay for).
11) On demand
Hypothesis: If you can initiate service via just a Web request it’s SaaS. If it requires you to interact with a human it’s a POWA. Or is it the other way around? Most silly little Web apps out there would fit the SaaS bill by this definition, while LotusLive-type deployments (or HP Cloud Assure) would not. So are we now saying that Cloud requires *more* friction?
Hypothesis: If you can quickly ramp up your usage (in terms of load per user and/or number of users) without having to warn the provider and without noticing performance degradations then it’s SaaS, otherwise it’s POWA. The Amazon storefront and GMail are SaaS, Twitter is a POWA.
13) Pay per use
Hypothesis: If your usage of the Web application is metered and your bill is based on it then it’s SaaS. Otherwise it’s a POWA. iTunes is SaaS. GMail is a POWA.
13) Primarily computational
Hypothesis: If the service provided is primarily computational (storage/processing of data) it’s SaaS, otherwise it’s a POWA. GMail and Salesforce are SaaS. A Web application for pizza delivery or travel reservation is a POWA. Banking services are SaaS (let’s chew on that one for a second).
14) Programmatic usage
Hypothesis: If the web application is primarily used programmatically (e.g. SOA) it’s SaaS, if it’s primarily used by humans through the Web then it’s a POWA. There aren’t that many SaaS applications by that measure. Maybe some financial/trading/information services? Or AWS S3 (which has the amazing property of belonging to IaaS, PaaS and SaaS depending on who you ask). Or iTunes?
15) Standard interface
Hypothesis: If there is a standard interface to the Web application, then it’s SaaS, otherwise it’s just a POWA. Webmails (at least those that support IMAP) are SaaS. Most other Web applications are POWA, by virtue of there being very few standard application-level interfaces (remember RosettaNet?).
16) Gets the CIO on magazine covers
Hypothesis: If adoption of a Web application gets the CIO on the cover of trade magazines, then it’s SaaS. Otherwise it’s a POWA. The problem with this definition is that it’s a moving line. The first CIO to use external email or CRM probably landed on a few cover pages. Soon nothing short of putting your pacemaker firmware in the Cloud will.
17) Open Source
Hypothesis: If the Web application is built using Open Source software it’s SaaS. Otherwise it’s a POWA. WordPress.com and the various Zimbra hosted providers are therefore SaaS but SalesForce.com is not. Yes, that’s a silly criteria to even consider. But someone was going to bring it up, so let’s kill it now.
18) Enterprise usage
Hypothesis: Web applications used for business are SaaS. Those used by consumers are POWA. For example GMail (as part of Google Apps) is SaaS while the regular GMail is… a POWA.
Which of these sorting criteria work for you? I don’t find any of them convincing. Many feel very artificial, as the example for the last one illustrates. Several (like the first three) seem to test how tied to the Web application you are (with the hypothesis that commitment means true SaaS while a one-night-stand is just casual SaaS, aka POWA). But this goes against the common notion that “Cloud” means more flexibility, not less.
I could list more question, but this list should suffice to illustrate the difficulty of establishing a Litmus test that separates SaaS from POWA. Even if we give up on a clear definition, I don’t even feel like I can say that “I know SaaS when I see it”. At least not in a way that I can expect most others to agree with.
Are there actually useful characteristics that am I forgetting to test against? Is it a carefully weighted combination of those above? In the end, is there a reasonable way to tell SaaS apart from POWA? If we can’t find an answer that people eventually agree on and that has functional relevance (by which I mean that there is a meaningful difference between how you consume/manage SaaS vs. POWA) then I’ll have to conclude that the term SaaS only exists for some of the following (bad) reasons:
- A two level-stack (IaaS and Paas) looks silly,
- adding SaaS to IaaS and PaaS instantly inflates the size, adoption and relevance of the “Cloud” umbrella , by co-opting pre-existing and already-thriving businesses,
- conversely, riding the coattail of the hot “Cloud” buzzword helps providers of such applications. SaaS sounds much better than the worn-out “ASP” appellation, especially with ASP.Net cheapening it.
After throwing stones, here is my proposal.
There is a definition I would like to use to qualify which Web applications qualify for the SaaS appellation: it’s any Web application which satisfies all (or at least several) of the layer 3 benefits in this “Taxonomy of Cloud Computing Benefits”. The first test is to get authentication and authorization right. The second is to offer proper programmatic remote interfaces. Etc (there are 5, remember to scroll down to “layer 3” to find them).
But if we take that list as a definition, rather than just as an aspiration, then let’s face the truth: right now there isn’t a single SaaS offering in the market.