Bruno Fernandez-Ruiz

I remain skeptic about any attempts to break out SaaS value/price in a pure TCO, like in this article by Barry Rosenberg and Craig Wright on techweb. If nothing else, simply because of the (human) random nature of choosing between intangibles such as security, mobility, usability, … which are all perceptions.

Whereas I agree with the authors that the key challenge in comparing licensed to hosted software is establishing a common measuring tape to avoid comparing apples and pears, I don’t agree that it’s possible to quantify such measuring tape as a TCO model, simply because of the random nature of perception.

The choice process in deciding between licensed software and SaaS is not very different from the one we go through to choose between a Mac and a PC, between buying a house and a flat, between choosing to drive to work or taking a train, etc. The decision maker is forced in these cases to make a choice based on things that are simply not rationally comparable. In the case of SaaS, we must compare licensing costs to convenience of upgrades; in-house security-tight infrastructure to SLAs and trust models on hosted services; ubiquitous presence to speed of use; and a long list of other incomparable properties. While it is true that each decision maker, each CIO, will have different a measuring tape, no CIO will be able to put down the choice of SaaS vs licensed software as a TCO decision.

The choice of software hosting model (in-house vs off-house/SaaS) is a random utility-based decision, which can be modeled using random utility models (RUM). RUMs are not only necessary for SaaS vendors to be able to adequately price their services, they are also a good modeling technique for decision makers to be able to compare between otherwise incomparable choices.

Google Apps is already grabbing corporate attention. It’s not the Google apps’ themselves, but the APIs to allow ISVs and other power software houses can use to write applications that extend or integrate with Google Apps. Whereas security was already a known concern about Google Apps, and Postini is certainly betting on the right horse, Avaya is being innovative and extending its IVR portfolio into the SaaS space. The full article is here.

It seems like the SaaS media relationships departments have been busy since last night. Microsoft and BT are talking about BT Application Marketplace, Salesforce is hinting about its 25,000 user customer and Google is in fanfarre-mode with Apps Premier Edition. It’s a busy today for Software-as-a-Service (SaaS), but I am not sure it’s a coincidence but a follow up on last night’s rumour, now confirmed, about Google’s launch of Google Apps Premier Edition.

Besides the news fanfare, I am afraid there isn’t much new technology onto Google’s announcement, but a rebranding of the existing Google Apps for your Domain (GMail, Google Calendar, and Google Talk) joined with Google Spreadsheet and Docs.

But there is a significant point to take into consideration with Google’s announcement, and that is price, with the Premier Edition account being offered at $50/year subscription for 10Gb and a 99.9% SLA for email.

As we discussed back in August when looking at the SaaS pricing models, the SaaS model addresses the challenges small and medium businesses IT departments face to provide cost-competitive secure and reliable infrastructure to the business. As such, the low-cost, enhanced mobility, enforcable SLA, and improved security present in SaaS are very attractive factors to the business, to the extent of possibly threatening IT departments with closure.

Taking cost into account, Google is pricing Apps Premier at $50/year/user account, which is just over $4/month, significantly under what other SaaS are currently charging. As we discussed then, the current pricing model of other providers, resulted in a magic number of roughly 500 employees upto which SaaS is more attractive than inhouse hosted software. But Google’s pricing model highly undermines that, and pushes that boundary all the way up to almost 15,000 employees. It’s this aggressive pricing which is likely to make this offering succesful, and where Google will leverage economies of scale to make a good profit.

We know Google Apps is competitive against inhouse software for businesses upto 15,000 employees, but how does it leave its competitors?

Let’s first look at the storage costs, by comparing it with the best online secure storage infrastructure, Amazon S3:

• $0.15 per GB-Month of storage used.
• $0.20 per GB of data transferred.

Typically in a hosting business you oversell, and you don’t actually need all the storage you sell since most users never consume their quota. I don’t expect users of Google Apps to be any different. Assuming a utilisation of 50%, which means an actual use of 5Gb per month per user, and a data transfer of 1Gb/month/user, the total yearly cost with S3 would be $11.4/year/user, or $11,400 for 1,000 employees.

That’s for storage alone. In terms of servers, using this time Amazon’s EC3 virtual hosting, at $0.10 per instance-hour consumed, the CPU cost of a 24×7 server would be $876/year. EC3 VPS is equivalent to an 1.7Ghz x86 processor, and assuming we can host 10 users on each server, an organisation of 1,000 employees should be able to run on a $87,600 budget.

Considering only infrastructure costs, and taking application costs aside, a 1,000 employee organisation would spend almost $100,000 on an outsourced hosted model, versus $50,000 with Google, but this time application included. It’s hard to arguee that Google’s is not a very interesting proposition thanks to its extremely aggressive pricing.

Eweek reports about Microsoft’s architecture guidance on SaaS (Software as a Service). There is code, a video and a screencast released on MSDN which I highly recommend watching. Well done, Microsoft.

As we discussed originally while looking at Web 2.0 software-as-a-service business models, we saw how hosted software is not a competitive offering for mid- to large companies over 500 employees. New research by Quocirca and Forrester now comes to a similar conclusion, and they add that there is a grey zone between 250 and 500 employees where it’s not clear the value in hosted services. Quocirca concludes saying that hosted services are rarely cheaper than in-house services, overseeing the 5.7 million businesses in the US under 500 employees. I am still to read both research reports to understand the full details.

Additionally, these studies seem to use current pricing models, such as the one from salesforce.com, but miss some of the points raised by Mark: the utility of hosted software goes well beyond cost (the focus of Quocirca and Forrester research) and includes less tangible things such as training time, added security, and productivity (as already commented by Steve Garnett, of salesforce.com, when asked about these two research reports). Also, as we discussed while comparing the Web 2.0 to the “old economy”, it will become necessary for hosted software providers that want to remain in the market to start providing free services, adding premium services, leveraging information lock-in and enhancing the value of the intangible benefits such as usability, ease of upgrade, automatic security/patch management, etc.

There seems to be at least 17 startups taking on the A-Team of the desktop applications, and possibly another hundred thousand teenagers creating their own little Web 2.0 application-du-jour in communities like entrepreneur extraordinaire Mark Andressen’s Ning.

Since the software-as-a-service business model is mainly only attractive for smaller to medium companies, why are there so many web 2.0 applications popping out everywhere?

I believe the catalyst has been a generational change. Web 2.0 is not anymore an emerging phenomena only found among geeks playing with XML, and asynchronous HTTP requests. The generation of young adults in their early- to mid-twenties have grown using the internet as a routine. They learned how to collaborate online using Yahoo!, how to use Google to find the solution to their assignment or homework, how to download MP3s using eMule, etc. and most importantly how to skim through the rubbish to get the very simple: content. The internet is just one more thing they use in their lifes, just like a phone. Given that they have grown on it, doing homework and making friends on the web, it is the collaborative aspect of the internet what possibly makes software-as-a-service interesting. It’s Wiki-Extreme if you let me put it that way.

Okay, so you have a market. You have entrepreneurs. How do they meet each other? Most of the startups doing Web 2.0 services will fail (I hope this is not a surprise!) and only those that are able to see beyond the collaborative, minimalistic aspects of the software-as-service business model will be able to survive. Simple usability, Apple-like design and aesthetics, and a beta-always badge are requirements to be in the Web 2.0, but fulfilling functional requirements are at the end of the day what makes an application work. Collaboration is fine, but not all business processes are suited for collaboration. So, what’s the new old?

• Information lock-in is the biggest simple asset these startups should leverage. They will need to lock in their customers into their proprietary document/workflow formats, to avoid switching. It’s a market with almost no barriers of entry, and the only protection they will have is information lock-in.
• Web 2.0 software-as-a-service should offer the products for free right now (yes!). The marginal cost for you to give the application to an additional user is close to zero, so your only short-term objective should be to create a large customer base, bigger than your competitors. As the market develops, the number of players will be reduced, but you will have your customers locked-in.
• As the market matures, you will need to start making money (uh?). Seriously, don’t expect to make a business out of AdSense! So how do you make money? Your best returns will come from premium services targeted those companies to who software-as-a-service is borderline to not being competitive anymore, and would be tempted to go for traditional solutions ala Microsoft Office. For these customers you will need to offer the extra mileage: better backups, better training, consultancy, etc.
• Price accordingly: free for most, pay for a few. The more free customers you have, the more the few will be willing to pay to ensure format compatibility.

These are nothing but the good old economical principles that apply to the web 2.0 information economy. Not surprising.

The recent advent of Web 2.0 no-software start-ups like 37signals, salesforce.com, Writely, etc. is getting plenty of attention by the media and VCs alike, but I have not seem much about the pricing model of these no-software business models.

The basic idea, repeated all over the place in Web 2.0 start-ups, is that one can develop very rich user interfaces using thin client technology. Instead of installing a local copy of Microsoft Word, you use Writely. Instead of installing Siebel in your enterprise as a CRM solution, you use Salesforce.com. And so forth. Even Microsoft’s Bill Gates, highly concerned about the future of Microsoft in a world of “applications built from the grassroots” is putting forward an online Microsoft Live CRM service.

The advantage of using online solutions are allegedly savings in infrastructure and support costs. But is this really true? Will individuals and business of all sizes alike chose online services versus an offline dedicated solutions? Obviously not, so who is then the target customer? Or rather, when will a consumer chose online vs offline applications?

An informed consumer (be it a home user, a web designer, or a CIO) will make a choice based on the expected utility he gets as a consumer from the software. He will compare the utilities and chose the one that seems most likely to suit his needs. The expected utility for each option, online or offline, will surely include quantifiable factors, such as software price, downtime costs, support costs, infrastructure costs, etc. as well as perceptions such as privacy, reputation, etc.

I prepared a highly simplified utility model that I could use to bring some light. In this model the only factors I considered were:

• Software price, for an offline solution, expressed as a monthly depreciation cost as a function of the number of employees. For solutions that require systems integration and professional services, I have diluted the cost of consulting in the total price of the solution, which then gets expressed as part of the monthly depreciation costs per employee.
• Service price, as a monthly recurring charge by the provider as a function of headcount.
• Infrastructure costs, for the offline solution, usually comprising storage, servers, backup, etc.
• Service support costs for the offline solution, to support the local infrastructure costs.
• Cost of downtime. I have assumed local class 4 availability in each domain. This translated into 0.01% downtime for offline software, 0.01% to local network, 0.01% to the local ISP, 0.01% to the provider ISP, and 0.01% to the service itself.

I have run the model with existing data for both “consumer applications”, using 37signals Basecamp and Microsoft Project for pricing data, and for “enterprise applications”, using Siebel and Salesforce.com for pricing.

The model yields interesting results:

• For both consumer and enterprise applications, the break-even point where the software-as-a-service is not competitive anymore is somewhere around 500 employees.
• The key differentiation between the utilities derives from infrastructure and support costs. For small to medium businesses, the cost of a dedicated and under-utilized infrastructure and service support personnel is prohibitive, and only large companies are able to realize economies of scale to make offline support viable.

We can clearly see the validity then of these start-ups, since in 2003, there were 5.7 million businesses in the US with less than 500 employees, at an average of 10 employees per firm; and 16,926 businesses with more than 500 employees, at an average of 3,300 employees per firm.

These 57 million businesses with 10 employees cannot afford the costs of dedicated infrastructure and service support, and are better off to chose an online application. For an average of 10 employees per firm, the model yields a cost of approximately $5,000/month for an offline solution, versus a $150/month for an online solution. $4,850/month of infrastructure and support savings surely justify the consumers choice.

As a side effect, the model yields a potential market size of approximately $10 billion for the Office applications, possibly close to $50 billion for the standard applications used currently in small and medium-sized businesses.

To conclude, it’s interesting to note that the online software-as-a-service business model is a side-effect of an underlying consumer pain: lack of secure shared infrastructure.