This is part 1 of a multi-part series on the virtualization revolutions no one is really talking about. If you enjoy my longer blarticles, make a cup of coffee; you won't be disappointed by this one.
Why Enterprise Software is Not Dead
I’d like to show you something I just drew on the back of a napkin:
“You know.. for VCs!” (imagine me doing my best Tim-Robbins-ala-Hudsucker-Proxy voice here).
drawing is actually something I have been predicting for a while now. And recently
I seem to be observing it happening at an increasing rate. Enough so that I
believe its going to radically change the way we think about the software
All that in two lines and one virtual napkin – how can you not read on?
New Kids on the Block
Basically the scribble represents an argument of old versus new. For a few years now, VCs (and many entrepreneurs) have been increasingly repeating the new software investment mantra “We don’t do enterprise software; we only do SaaS.” So much so that you’d have a harder time finding a classic enterprise software investment (CES) on Sand Hill than you would a WGA member at the Golden Globes. These days, most VCs politely decline your request for a meeting when you fess up and tell them you’re coming over to drink their coffee and pitch them a CES investment. It’s the equivalent of trying to pitch “tools” in 1999. So this is the debate: Which is better, SaaS or CES? And no, the title of this blarticle actually doesn’t give the answer away. Nice try!
Before I wade into it let me say that I think I have an interesting perspective in which to enter into this debate. The most recent software company I founded, Newmerix is a CES company by all accounts. It says “Change management for packaged applications (SAP, Oracle EBS, and PeopleSoft)” somewhere on the very first slide in our VC deck. That’s like bold-facing and 24-point-fonting the “classic” in CES. Before Newmerix though I co-founded, built, and sold a company called Service Metrics. Back in the day we would have called that business an ASP (Application Service Provider) but now-a-days the kids are calling it SaaS. So I’ve played for both teams and I watch the current tides turning towards SaaS only investments with intent fascination.
Proof of Life
The great thing about shifting any investment philosophy is that only a long enough history will tell you if you’ve headed the wagons in the right direction. So far along our little SaaS Oregon Trail there have been a lot of notable successes that have shown us proof of life and put some wind in the investment sales. Salesforce.com is obviously the poster child here, NetSuite just had a successful IPO, and big companies like SAP have finally figured out the subtle difference between what “SaaS” and “Hosted” means.
said, things change quickly in the technology (and thus venture) world. I’m
here to argue that SaaS may not be as good an investment as everyone thinks,
enterprise software is definitely not dead, and in 5 years it won’t make a
difference one way or the other.
If You Don’t Have Something Nice to Say..
me start off by playing my own devil’s advocate and state why I think VCs are
so gung ho for SaaS investments. The reality is that SaaS has exposed (and
perhaps solved) a lot of the problems with CES investments. If we set aside
some of the penguins-on-the-iceberg mentality about why people invest in SaaS,
I think most VCs would probably say one or more of the following things about
their affection for these types of investments right now:
1) An annuity model is more attractive to the company and the buyer than the perpetual licensing model
2) When we build software in the SaaS model we can get to market faster than when we build it in the traditional CES way (ship it on a CD)
3) We can sell to a departmental buyer and circumvent the need to include IT in the buying syndicate (and that’s a good thing)
4) It costs less to build a SaaS company than a classic enterprise software company
There is no doubt I have missed a few key points. I am sure some folks are just in it for those cool “No Software” buttons that Beniof likes to hand out. For now, let’s at least start with the building blocks I listed. And to continue my devils advocacy, let’s fire up the old SaaS versus CES scoreboard and see who the winner really is in each of these arguments:
Saas – 0, Enterprise Software - 0
Round 1: Annuity Models
My fellow Irregulars have publicly debated the shift to annuity based software licensing models from the buyer’s perspective countless times. In general I think the answer is the buyer likes it. You don’t have big up front investments, you don’t have to worry about capitalizing software over long periods of time and generally you pay as you consume. The lesser asked question is what is better for a startup software company: annuity based sales models or classic enterprise sales models? In my best legal approach to the answer I’d say, “It depends”.
would argue that they can predict and model the CES approach of selling the
first thirty $100K deals directly to IT (assuming you have a compelling product
to get that high of a price in your first year). Some would gladly take on the
alternative of trying to missionary sell 100 deals at $30K a year arguing there
is less risk to the buyer which is nice for a startup.
While the benefit of annuity models is that (once sold) annuities keep on giving (as opposed to your standard 18% renewal maintenance on CES as a startup), you have to be able to show one or more of the following three things very quickly about the SaaS sales process:
1) You can sell using almost all inside sales (flying people on planes for $30K deals is a non-starter sales model)
2) You can use a direct sales force because you’re average contract length multiplied by your annuity price approaches the same number that CES deals end up at (e.g. >$100K minimum is required for direct selling models or you go bankrupt while your sales force racks up frequent flier miles)
3) You can do broader up front deals with more seat licenses and bigger annuity value so your contract length doesn’t matter as much (e.g. $250k site licenses).
The last thing you need to prove is that your product is sticky. If it’s not (in either the SaaS or CES model) you’re in trouble. Clearly the annuity model just isn’t interesting if no one renews. In CES, you might get away with a less sticky product for a while because you get bigger upfront checks, but you can’t run from that issue forever either. In both cases, a non sticky product is a recipe for failure. Slowly due to the market’s awareness of poor quality or shelf-ware, your average sales price or annuity contract lengths will shrink to unprofitable sizes.
So, on our SaaS versus CES score board I think this one is actually a wash when it comes to what is good for a startup. While it’s hard to predict that you can sell a CES piece of software for more than $100k, it’s also pretty damn hard to ensure one of the three required selling dynamics needed for a SaaS model will take place. Rather than score zeros, let’s feel good and give each side a point here.
Saas – 1, Enterprise Software - 1
Round 2: Getting to Market Faster
So why is it so much faster to get to market with a SaaS model than it is in a traditional CES model? One word: testing. Was that not the word you were expecting? Well, when it comes down to it, testing really is the determining factor. Let me explain why.
assume you actually strive to put out high quality software (in either model you
choose). And let’s also assume you do the appropriate amount of testing to
ensure high quality. No matter what development methodology you use (XP, Scrum,
Waterfall, “something those guys in engineering do”), whenever you GA a
software release you have to test it. It’s this process of testing that
differentiates SaaS from CES and is the fundamental Achilles’ heel of the CES
In the CES model you end up spending a good 30-50% of your total GA cycle time dedicated to testing. Note that I did not say cost (e.g. you don’t usually have a 1:1 software engineer to tester ratio), but very specifically time. The time problem comes from the reason that in a CES model, testing complexity grown exponentially. To test a piece of software (in either model) you have to figure out every realistic combination of environments that your software will be deployed into and make sure you run at least some basic tests in that environment before shipping it. When talking about the possible environmental variables that might affect the functioning of your software, this could be the operating system, OS patch level, database type, database version, database client adapter version, internet browser version, application integration API versions, .NET framework versions, web server versions, etc.. etc.. (these are actually only a subset of the variables we consider at Newmerix). The resulting matrix is n-dimensional and starts to quickly look like Brian Green trying to draw a Calabi-Yau space with a crayon. To make matters worse, the more successful your software becomes, the quicker your matrix grows as you have esoteric customer situations you must support. There is always at least one customer who still has Oracle 6 running somewhere in their IT department and would just “really love it if you could support that” for them. In addition, the sales team is going to expand this matrix by being opportunistic, especially in the early days. No matter how much you double underline and highlight that you only support Oracle databases in your requirements document, as soon as a $1MM SQL Server deal comes walking through the door, guess what, you’re supporting SQL Server too! And in a final crushing blow, whenever you add something to the matrix, for the most part you can never get rid of it.
those of you who have not been through this virtuous startup sales cycle
before, you may be shaking your head in disbelief. Come on you say, you should
have more restraint than that. Focus on the revenue you want, not just any old
revenue in the pipeline. Well, as a business grows, I 100% agree that you
should strive for this focus. But let’s put it this way. Say your spouse is
knee deep in cooking for a dinner party you’re having that night. They send you
to the store for 5 limes they forgot to pick up for the prize entrée of the
night. Like a well trained spouse, you brave the freezing night weather and
drive around to every grocer in town only to find there is no one open that
late (it’s a Sunday). In despair you end up at the liquor store, ready to drown
your sorrows with your new friend, the tequila worm. As you’re checking out, you
notice a basket of fruit. Limes! But there are only 4 janky limes and some
lemons left. What do you do?
Well, I guarantee you’re gonna bring home the limes no matter what they look like. And depending on how well you know your spouse, you’re probably going to pick up 5 lemons too. It can’t really taste that different with lemons instead of limes, right? They are all citrus fruits after all; just a little switch to the recipe and off we go. Clearly in this little story the VCs are your spouse, tonight is the end of the quarter, you’re the CEO and the recipe is the engineering team. It’s really not much different in a startup. Well okay, maybe there is more tequila in a startup.
To cap of this discussion, I included for your viewing pleasure a little example from a big company. Below is a small slice of a spreadsheet I created which shows PeopleSoft’s PeopleTools version support for the various customer database types and release versions they test. Every time Oracle releases a new version of PeopleTools they need to run a battery of tests against every supported database type and version in this matrix. And this is only 2 variables in the complete set of considerations when testing PeopleSoft. You get the point (and I’ve cropped even this picture dramatically in both dimensions!).
I’m going to bastardize a Winston Churchill quote here and say “Quality is just one damn test after another”.
And if you’re thinking to yourself, “Why not just add more testers, bare a little more cost and get on with solving another problem already?” then please read the Mythical Man Month. To summarize it, it just doesn’t work that way. 9 women just can’t have a baby in 1 month. The one viable solution is to invest heavily in testing automation. While it’s not a complete panacea for this problem, it definitely lets you scale in a way that can’t testing just manually. While I have a lot to say about automated testing (its one the core products Newmerix makes and sells to packaged application owners), I won’t go into that long digression right now. You’ll thank me later.
to the plot. So along comes SaaS with its sexy web front ends, its new-fangled
ajax, and its smug little annuity model. But underneath the gloss and glam, the
best thing about SaaS is that they control their own infrastructure. Engineering
has only one customer – themselves. There is no n-dimensional testing matrix. In
fact there is only one cell in the testing environments spreadsheet – it’s
whatever the hell the SaaS company wants it to be. If you’re building a SaaS
app and you want to run Oracle 10.0.0.3 on Windows 2000 SP1 with a WebLogic 6.7
application server on powder-blue Power Macs running Parallels then fine,
that’s all you have to test! Calabi and Yau would be behind the counter at
Starbucks foaming you up a double caramel mocha latte if it was a SaaS-only
universe out there. And to make things even rosier for SaaS, you don’t get
crushed under the weight of your success (at least from the testing aspect).
There is no ever-growing QA lab full of customer environment testing machines.
It’s one environment all the time, totally predictable, totally testable.
This singular architectural benefit allows SaaS companies to move infinitely more quickly that CES companies. Big releases come out faster, bugs get fixed and deployed quickly, and features can be rolled into the GA product in a much more incremental fashion than is realistic in CES. In a SaaS model you can literally roll out a change every week or two if you’re agile development methodology allows this and your testing is well automated. This is simply an impossibility in CES as it stands today. In addition, CES customers don’t want to upgrade their applications on a weekly basis. In the SaaS model it all happens behind the scenes.
and more incremental time to market means less of a gamble on getting the whole
thing right in one big bang and more feedback from the market as you try and
define the optimal feature sets for your customers. This, in my view is the
biggest strength of the SaaS model.
So let’s score 1 for SaaS here, by far.
Saas – 2, Enterprise Software - 1
Round 3: The Departmental Buyer
I won’t belabor this vastly discussed point (read any article on SaaS and you’ll see a discussion of why SaaS allows non-IT departmental buying). The truth is that if you can cut anyone out of the sales process whether it be IT or anyone else, this just makes things easier. The downside is that most non-IT departmental budgets simply don’t have large carve outs for huge software expenditures. So, one could argue it’s a wash (or even argue it’s a big challenge for SaaS if you go back to my earlier discussion of price points versus sales structure). But, since Salesforce.com seems to have proven there is a viable business model selling SaaS to sales and marketing I won’t deny the reality that it can work. Score one more for SaaS.
Saas – 3, Enterprise Software - 1
Round 4: It Costs Less to Build a SaaS Company
go back to my napkin drawing for a second. The scribble represents two things.
The bottom line is the revenue curve of most CES companies over time. The top
line is the revenue curve of most SaaS companies over time. Notice two things. The
first thing is that the SaaS line gets to revenue much quicker than the CES
line. This would be consistent with witnessed phenomenon from the VCs I know.
The second and more important part of the diagram that I will have you notice
is that the lines converge at some point. I’ll argue that point is really the
average break even point for most software companies (about $10-12MM run rate).
But more subtly, at some point in time, something about the SaaS model starts
to make it more inefficient than the CES model. Or to say it in a more positive
way, something about the CES model starts to kick into gear if you get far
enough along the curve. While this scribble is clearly not going to win any HBS
awards for discrete financial modeling, there is one observed phenomenon and
one technical reason that make me think this will become a common perception very
Let’s go back to my napkin drawing for a second. The scribble represents two things. The bottom line is the revenue curve of most CES companies over time. The top line is the revenue curve of most SaaS companies over time. Notice two things. The first thing is that the SaaS line gets to revenue much quicker than the CES line. This would be consistent with witnessed phenomenon from the VCs I know. The second and more important part of the diagram that I will have you notice is that the lines converge at some point. I’ll argue that point is really the average break even point for most software companies (about $10-12MM run rate). But more subtly, at some point in time, something about the SaaS model starts to make it more inefficient than the CES model. Or to say it in a more positive way, something about the CES model starts to kick into gear if you get far enough along the curve. While this scribble is clearly not going to win any HBS awards for discrete financial modeling, there is one observed phenomenon and one technical reason that make me think this will become a common perception very soon.
In talking with many of my VC friends, I am seeing that the break even point (money invested versus cashflow break even) is actually in the end pretty much the same for SaaS and CES. As a rule of thumb, $10M is a pretty good target for break even in a software business. Guess what, it takes about $25M of real cash to build a solid, sustainable growing software business no matter how you do it! Having built a few, my gut tells me this is true for a number of reasons. Simplistically I believe that buyers (of either 2000 SaaS seats or 250K of CES software) have the same perspective of total value in what they are buying. It does not matter if their software is delivered through the web or sits in their own facility. Couple this with the fact that developers write the same number of lines of code to create that value every day whether it is delivered in a SaaS model or a CES model. Testing may take longer in one model due to the matrix but developers still crank out their 600-1000 lines of code a day to add perceived value to the product.
The technical reason why these two lines catch up with each other is that in the SaaS model you trade off deployment complexity (e.g. what environments do I need to deploy my software in) with deployment scalability. Its one thing for a developer to design a system to manage the transaction or data volume of the largest enterprise customer they can think of, its another to design a system capable of handling the aggregate transaction volume of 400,000 customers of all sizes. Along with that comes the DBAs, the actual hardware, the power costs in the data center, the bandwith, etc.. In CES, Brian Arthur’s classic increasing returns philosophy is right. Spending $50MM to make the first Windows disk and then $3 for each copy after that is a pretty good model. In SaaS it doesn’t really work that way. The first system might cost you $20MM to build but your incremental delivery cost is much higher. Unfortunately scale issues in software are non-linear – they too are an S curve, but an inverted S curve. You really don’t want to get on the wrong side of that S. Trust me – it’s worse than messing with Texas. In the CES model the ISV pays to handle deployment complexity (testing) for their customer and they don’t worry about deployment scalability. In the SaaS model the opposite it true, there is little deployment complexity but the deployment scalability is massive. This is why the two curves catch up with each other sooner or later. You’re really just moving the total cost of delivering the software from pre-GA to post-GA.
am sure on both sides of the equation everyone can name aberrant data points (e.g.
PeopleSoft went cash flow break even for $6MM – I am not sure what that is
inflation adjusted from 1985 – maybe something like $17MM). I am sure there is
a SaaS company or two that got break even for less than $25MM invested. But
again, while its emergent phenomena, I think VCs are going to have to come to
terms with it in the next few years.
I’m going to score on for CES here because I think this dynamic is vastly underestimated in the investment community right now.
Saas – 3, Enterprise Software – 2
Okay so we’re sitting here looking at a scoreboard that clearly has SaaS in the lead. Hit the gas you say, lets go find some SaaS companies to fund. Well hold on. I said in the next 5 years the dynamic would shift again. All it would take is 1 point to move from one side to the other and CES looks like a better investment again. Well keep reading and I’m going to give you more than 1 point to consider.
Why (for the most part) None of This Will Matter in 5 Years
At a high level I fundamentally believe the advantage of SaaS and even the core difference between SaaS and CES will start to dissipate in the next 5 years. This will be due to one thing: virtualization. Let me give you context to this thought by explaining what I think are the relevant technical differences between SaaS and CES applications at it stands today:
1) SaaS is browser based (as opposed to client based)
2) SaaS is web based (e.g. a hosted application)
3) SaaS is multi-tenant (this actually doesn’t matter to the customer other than from a data protection and security standpoint but it’s an architecturally relevant note in terms of the SaaS application design)
4) SaaS is hosted by the ISV and there is no software on the customer premise
5) SaaS runs on one infrastructure stack dictated by the ISV (as described above)
While these might seem like big differences, virtualization has the capability to completely normalize them all. Let’s go through the list:
Being Browser Based
There has been an immense amount of time and energy spent making browser-based applications work a lot more like a classic windows client. With the arrival of desktop remoting and application virtualization such as Microsoft’s Softricity products, the pendulum will swing back again and some people will simply favor using a client. The interface is more familiar, the functionality can be designed more directly to the application needs, and the actual user’s desktop OS becomes less relevant (running a Windows client remotely on a Mac is becoming common place with Parallels).
change in score here but an important shift to note.
Being Web Based
again with remoting of desktops, there is no need to install software on each user’s
desktop. Whether its 3rd party software we’re talking about or
in-enterprise applications the trend will be to remote everything to the user’s
desktop anyway. I have another post on how virtualization is changing the face
of consumer and enterprise desktops coming. So many posts so little time.
While this may seem the same as point #4 above note that customers do install CES web based applications in their own environment now. Lotus Notes would be a good example and Newmerix has a web-based product in it’s suite called Automate!Control. Thus no change in score here but again an important shift to note.
Vishal Sikka, SAP’s CTO made a great point to me when I spoke to him last (I have commented on this in other blog posts as well as I think it’s so relevant). Multi-tenancy has become inaccurately synonymous with the database layer in the last few years. Over time though, people will shift where multi-tenancy lives in the application stack. It might live in the database (e.g. serve all customers from one app but filter the data they see by their customer id), application server, application itself, hardware cluster (shared compute cloud), or simply through security policy. Assuming licensing structures catch up with the flexibility of virtualization, there is no reason why as a CES ISV you can’t simply replicate the complete application stack in a VM image and use utility computing to handle the scaling issues on a per customer basis. I think that Salesforce.com may find itself in a position of having invested a huge amount of money in a multi-tenant database structure only to find out that it simply doesn’t matter anymore. It’s even possible this ground breaking design approach might become an impediment to the speed at which they can move in the future.
is actually a shift in favor of both CES and SaaS players. Either type of
player can change where multi-tenancy lives to their advantage. So no change in
score here either.
utility computing and virtualization any classic ISV will have the option to
install their software on a VM and run it from any one of a number of
virtualized utility computing infrastructures, essentially providing their
application as a hosted service. All the complexity that SFdC deals with
related to managing a proprietary multi-tenant architecture will essentially
disappear into the cloud and hosting VMs will become a matter of policy and
security management and little else.
This capability will drive CES ISV pricing options to include an annuity option (as opposed to a perpetual license). CES vendors will simply install and host their own applications on VMs. The multi-tenancy will actually come from the utility compute grid. This VM hosting will also allow CES ISVs to target a departmental sale as the customer does not need to manage or upgrade any of the underlying application infrastructure. Classic CES vendors will be able to cut IT out of the buying loop just like SaaS players.
score 1 point back for CES here.
a CES ISV moves towards a hosted virtualized platform, they all of a sudden
gain the immense benefits of having only one reference infrastructure. As
discussed above, this massively reduces testing time, is a dynamic that does
not change with increased customer adoption, and drives the ability to get to
market in a much quicker fashion. To some extent, this advantage of virtualization
alone will end most of SaaS’s advantage. Once CES ISVs can deploy hosted VMs
running their own selected software stack, the playing field becomes instantly
leveled and disruption happens all over again in reverse.
Again score a point back for CES, or at least consider it a wash.
I do the quick tally, we’re back to SaaS – 2, Enterprise Software 3 (erring on
the side of being conservative). Oh how quickly the tides shall turn!
Disruption vs Innovation
number of times in this article I have mentioned disruption. I fundamentally
believe that the success of SaaS as a generalization is more about taking
advantage of market disruption than really driving new conceptual invention. I
think this strength might become its ultimate weakness as a sole investment
There has been massive consolidation in the CES market in the last 5 years. This has put the large ISVs (Oracle, SAP, IBM, BEA, etc..) on the back foot in terms of their ability to react quickly to dynamics in the market. One need only look at Oracle’s constantly elongating Fusion roadmap timeline for proof that good CES just takes a long time. And because it takes a long time the players must devise strategies that are worth the effort. Hence these big companies have broad reaching, deep-infrastructure, transformational strategies for their next big thing. Most of the time it’s just about swinging the pendulum back in the opposite direction on the consolidate/modularize spectrum from whichever way it has been going for the last few years.
CES today is all about SOA, merging packaged applications and custom development into composite applications, building out application platforms on top of existing packaged application deployments, BI and centralization of data. The first blarticle I ever wrote was about this very trend. The SaaS players that have been successful basically do what the large software companies do but faster, in a new model (annuity versus perpetual license), in more modular fashion from a sales perspective, and with a divide and conquer approach to the buyer. Essentially they identified the monolithic ISV-IT juggernaut and built a speedy little boat to zigzag right around the massive aircraft carriers that most large ISVs have become. There are always new winners in a market when some kind of disruptive force leaves open opportunities for new players to shake up the status quo. Sometimes this comes in the form of architectural shifts, sometimes it’s pricing, sometimes it’s core technology changes, and sometimes it’s a mixture of these things. To date that’s what the most successful SaaS companies have been about. Lets be honest, there was already more than enough CRM platforms in the world when Salesforce came along. While I am sure many of you will disagree, I think SaaS’s successes have not been about inventing new things, but doing existing things better.
As a side note, I think a great analog here is the open source market. As the CTO of Digital Persona once noted to me, the best open source applications are those that simply copy an existing well known product and replicate it with a different development model. We’re seeing some big wins in this space. The classic example would be Red Hat but you also have JBOSS (now part of Red Hat), Zimbra (now part of Yahoo), SleepyCat (acquired by Oracle), Firefox, Xen (acquired by Cirtix), the new Cobia (full disclosure I am an investor in StillSecure), and the growing footprint of OpenOffice. It’s hard to name a number of incredibly successful open source products that were completely new innovations. The same could be argued for SaaS.
From an investors perspective market disruptions have well defined windows. It’s usually 7-10 years in total length with heightened activity and value creation just past the halfway point. That’s what we’re seeing now with SFdC, NetSuite, and other going-public SaaS plays. But this disruptive situation will taper off in the next 3-5 years as large ISVs right their ships, buy strategic innovation from leading SaaS companies and learn how to compete in the new market. It was already happening when Siebel bought UpShot and now it’s happening in spades with SAPs BBD platform.
an investment thesis perspective, I think it’s very useful to look at the
timeline of this current disruption. How long does it really takes for a SaaS
company to gain critical mass and how much longer is the disruption going to
last? If the window is really shrinking to less than 5 as I HAVE argued, that’s
a pretty tight window for any new software company to build a product, enter
the market and declare some kind of victory. Perhaps its time to start shifting
from investments in market disruption and go back to investments in conceptual
innovation. In the same way that investing in a social network if probably a
bad idea right now, my guess is by 2009 there will be few new companies that
get started that actually make any real money for investors in the SaaS space
doing the same thing an existing CES player is doing.
If I Was King for Just One Day
so it’s easy to sit around in a nice glass houses and throw lots of stones. Lest
I be accused of doing so, let me outline a few software investment theses I
would have if I was a software VC making investments in the market right now.
As mentioned, for the next 2 to maybe 3 years, simply disrupting the existing
software incumbents with web-based, annuity based recreations of the heavy
software they already sell is still going to be a smart investment. After that
(and even right now one could argue) you’ll need a change in strategy. I should
say that in the end this thesis is not about SaaS or CES at all. It’s really about
new ideas that can get to market first in whatever model makes most sense.
Massive amounts of IT budget goes into solutions that address changing regulatory requirements. The most recent example is the emergence of the compliance market. Whether it was to manage SOX, J-SOX, CLERP 9, PCI, BASEL, GLB, or HIPAA, a number of companies emerged that provided true value to IT owners in dealing with these issues. Virsa was bought by SAP for $400MM, a number of security vendors were bought by the majors to handle security based compliance, and even companies like Newmerix have benefited broadly from the compliance backdrop of IT departments.
The biggest new trend in IT is ITIL. For those not familiar with ITIL it’s is a service based approach to managing IT resources and doing impact analysis of changes to the IT environment. In general, ITIL (and COSO and COBIT) or manifestations of the fact that IT departments simply must find a better way to manage the rate of change across the IT stack. We’ve seen companies in the hardware stack (Troux), the OS stack (BigFix, HP/OpsWare), and the application stack (Serena, MERQ/HP, Newmerix) all benefit from this new challenge in IT. It’s not necessarily sexy but it’s fundamental stuff.
As we have learned over time, technology has a long tail in IT departments. How many companies still run massive transaction processing systems on mainframes? How many companies still have COBOL and FORTRAN running applications in their IT department? How many companies are stuck with packaged application infrastructure they bought 9 years ago to solve the Y2K problem? These things never go away – they have 20 year life spans. Many a company has been immensely successfully managing these core (I am avoiding the term legacy here) systems? Simple examples would be Candle, BMC, Newmerix (so far so good), and the new rollup Rocket Software.
Drive to Efficiency/Drive to Scale Pendulum
companies flip flop in their macro strategy between driving to efficiency
versus driving to scale. These are long cycles (7-10 years) and repeat
themselves over and over. Examples of companies that made money off helping new
technologies scale are F5 (and all the load balancers), my last company Service
Metrics that measured the performance of web sites as people tried to deliver
scaled web sites to them, a host of players in the SAN/NAS space, and companies
like Akamai which help companies scale their web delivery infrastructure. On
the flip side, virtualization is absolutely and incredible example of the shift
back to efficiency. If you own a data center its all about consolidation,
maximization of resources, and modularization of infrastructure for efficiency.
I just got some frontline statistics on the average utilization of hardware
resources in major enterprise customers. Less than 10% is normal. That’s LESS
than 10%. Tons of money (we’ve already seen this with VMWare and Xen) will be
made helping people make better use of what they have.
Behind the Firewall or Integration Plays
thing that SaaS currently struggles with is integration into existing IT
infrastructure. While SFdC has made strides in its APIs and Web Services
interfaces, it’s really not an enterprise level integration platform. In
addition, it’s still a pain to tunnel from outside the firewall back into it
for general software infrastructure. This may change over time but CES vendors
have the jump on SaaS players here. Note for example SAPs MDM or any BI product
or the recent purchase of FAST by Microsoft for enterprise search. I believe
there are huge opportunities to understand the existing application
interdependencies and build products to help integrate, transform, scale and
manage these interactions behind the firewall.
That said, we’re starting to see SaaS integration plays that are doing well. Eloqua is a good example. I guess when you have enough important data across multiple systems sitting anywhere (inside or outside the firewall) there is a need to integrate it in some shape or form.
The Reports of My Death Have Been Greatly Exaggerated
the last few months I’ve been talking to a lot of VCs about investments in CES.
Partially this is due to my involvement with Newmerix, and partially because
it’s a hot topic on the Irregulars mailing list. In the end, everyone has there
theses and they need to stick to them. That said, venture investment strategies
are long term strategies. Most funds have a 7-10 year realization timeframe in
their investment structure and need to consider macro shifts that are both
opportunities and dangers for their theses. I hope I’ve done a little bit to
shine the light on what I might call “increasing irrational exuberance” about the
SaaS market. While many of you are bound to disagree (all comments welcome), I’ll
put my virtual 20 bucks on the table that time will bare out my arguments here.
And hey, if I’m wrong and enterprise software really is dead, you’ll all be so knee
deep in SaaS S1’s that you won’t care either way.