I had breakfast the other day with a software entrepreneur. When I asked if his company was on a subscription or perpetual model he said: “we should kill the guy who invented the perpetual license — I’m on the perpetual money model, subscription all the way.”
Having worked largely in perpetual license firms, I admit there are many downsides to the perpetual model. Companies on perpetual models typically:
- Have more volatile revenue performance due to a relatively smaller annuity “keel” on the business (in the form of maintenance renewals).
- Are more exposed to end-of-quarter shocks driven by backend-loaded sales. (Most software companies get 70%+ of their orders in the last month of the quarter and most of those in the last week.)
- End up with “drive-by sales” cultures because sales reps are paid only on license sales and not on maintenance renewals.
- Have less customer-success-focused cultures because sales reps care about customer success only to the extent they see potential follow-on license business in the short term.
That said, there are many ways to mitigate each of the above points and all of the world’s largest software companies, such as Oracle and SAP, still do most of their business on a perpetual license model.
Over the past decade companies like Salesforce, NetSuite, and SuccessFactors have pushed the software as a service (SaaS) model where the vendor both runs the software and bills on an annual subscription basis to use it. While the SaaS model cut its teeth in applications like sales force automation, vendors are increasingly selling platform as a service (PaaS) offerings as well, such as Amazon Web Services, Google AppEngine, or Force.com.
Clearly SaaS interest and hype remain strong. Salesforce is trading at 100x FY11 earnings. Bankers have told me that the IPO bar for SaaS companies is $75 to $100M in revenue, while for perpetual companies it might be 1.5 times higher than that. A recent Software Equity Group report pegs the median enterprise value (EV) of of SaaS companies at 4.9x revenues, almost double the 2.7x revenues for perpetual companies. On an EV/EBITDA basis, it’s even more dramatic with SaaS companies at 44x and perpetual ones at 13.6x.
Given all this, I thought it would be fun to make an Excel model that concretely demonstrates some of the differences between perpetual and SaaS software companies. To do so, I’ll first model a fictitious, red-hot software startup on a perpetual basis. Then I’ll remodel the same company on a SaaS basis. Then we’ll play around with the models and see what we find. (For Excel geeks, my model is here; you’ll need to download it.)
To make my model, I started with bookings for the perpetual company and hard coded $5M in the first year on a reasonable ramp. Then I made a set of reasonable assumptions (for a hot startup) that drove the rest of the model: 100% license bookings growth, a 20% maintenance rate, a 90% maintenance renewal rate, a 50% rate of professional services organization (PSO) services bookings relative to license, and a bookings-to-revenue conversion rate of 85% for PSO in the subsequent quarter. To keep things simple, I didn’t model months, I didn’t model cash, I assume all bookings happen on the last day of the quarter, and I assume all license revenue is immediately recognizable.
Then I remodeled the company on a SaaS basis. The most important assumption to make here is labeled “subscript as % of license” – i.e., if someone was ready to pay 100 units for a perpetual license to use something, presumably they want to pay some fraction of that for a one-year subscription to use it. (I’ll call this F for fraction.) For the initial model, I assumed F=50% which is arguably aggressive. I kept the renewal rate at 90%. I assumed that configuring a SaaS system requires less PSO than customizing a perpetual one, so I assumed a 50% PSO bookings rate relative to the subscription (or 25% of the total PSO required from the perpetual vendor). I assumed subscriptions were one year and revenue was recognized ratably over the year and that all orders were received the last day of the quarter.
When you make these two models, here is what you find:
In year 4,
- The perpetual company is 2.2 times larger than the SaaS company at $62M vs. $28M
- The perpetual company is growing at 103% and the SaaS one at 115%
- The perpetual company has an 8% “annuity keel” in the form of maintenance renewal bookings while the SaaS company has a 33% annuity keel in subscription renewal bookings. (You can’t see this in the picture, but it’s in the model.)
Valuation and The Fallacy of Equivalence
Using the standard multiples above, let’s see what each of our companies is worth:
- The $62M perpetual company is worth 2.7 x $62M = $167M
- The $28M SaaS company is worth 4.9 x $28M = $137M
Simply put: the stock market works. With only a 20% difference in valuation between what ostensibly seem like two very different companies you can see that higher EV/R multiple for SaaS companies is almost completely offset by the increased difficulty of building a SaaS revenue stream. Wall Street “sees through” the differences in the models and values the companies roughly equivalently. Put differently, SaaS companies fetch 1.8x the revenue multiple of perpetual companies because they are worth 1.8x the revenue multiple of perpetual companies.
During the past few years I have spoken with several CEOs who transitioned their companies from perpetual to SaaS. The standard word is that it takes 3 years to make the transition and the transition must be a top-three company goal for that entire period. While there are many good reasons for perpetual companies to consider moving to SaaS models, valuation isn’t one of them. Yes, you get roughly twice the EV/R multiple, but building the R (revenue) stream is just about twice as hard.
Max Schireson calls this the fallacy of equivalence. If gold is worth twice silver and assume we have an equal amount of gold as we had silver then we are worth twice as much. The fallacy is that gold is twice as hard to come by as silver so you can’t assume equal amounts — see the huge revenue delta which is largely driven by the SaaS company’s need to spread revenue over 4 quarters.
Taking a Bad Quarter
Let’s look at how each company takes a bad quarter by assuming that we hit 70% of our bookings target in 3Q13 — doing only $4M in perpetual license bookings (cell P8) and only $2.25M in new subscriptions (cell P27).
- In the perpetual company 3Q11 revenue drops from $8.7M to $6.7M, the year/year growth rate drops from 105% to 58%, the stock is presumably crushed by 80%, and the CEO summarily fired.
- In the SaaS company 3Q11 revenue is unchanged. (Recall I modeled all bookings on the last day of the quarter.) 4Q11 revenue drops from $4.5M to $4.0M, 1Q12 drops from $5.8M to $5.6M, and the following two quarters also take ~$100K to $200K hits. The stock drops 20% because 4Q11 guidance is dropped but the company appears in control of its business and no one is fired.
Hitting The Flat Part of the Market
Now let’s examine both companies assuming that the market goes flat in 2014 (i.e., that 2014 license bookings / new subscriptions do not grow over 2013, cells S8-V8 and S27-V27).
- Our perpetual company sees 2014 revenue growth slow from 106% in 2013 to 17% in 2014. Revenue drops from the plan of $62M to $35.9M. The CEO is fired for flying the company off a cliff.
- Our SaaS company sees 2014 revenue growth slow from 141% in 2013 to 76% in 2014. Revenue drops from the plan of $27.9 to $22.9M. The CEO is commended for successfully managing the company through a tough transition.
What going on here is simple: volatility is being damped — for better and for worse — by the SaaS company’s need to spread revenue over the four quarters following the booking. That makes it harder to grow the revenue stream quickly. It also makes it harder to change once established.
One tricky issue in the SaaS model is sales compensation. In a typical perpetual company total sales commissions (at all levels) add up to around 10%. So, for 100 units of revenue, you pay 10 units in commissions. Sales reps are usually not paid on the 20 unit annuity stream of maintenance renewals.
In SaaS model, we have a conflict. If you assume the annual subscription fetches 50 units (i.e., if F=50%):
- The company wants to pay 10% of 50 = 5 units in year 1 and then pay little or nothing on the renewals.
- Sales want to argue either that  the deal is worth 150 units over three years and compensation should be 15 units or  (if they’re good at math) 300 units if you look at the stream’s terminal value (factored by renewal rates and discounted by 8%) and thus sales compensation should be 30 units.
So what do you pay: 5, 15, or 30 units? I believe that most SaaS companies end up splitting the difference in the some way, perhaps paying on a declining scale over the first 3 years. If you have good examples here, please share them in the comments.
While I didn’t model cash in the spreadsheets, one huge issue is the timing of commission payments. For example, if a company were to adopt the 3-year 15-unit commission argument and foolishly pay those three years up front, it would have a big cash consumption issue because effective year 1 commission rates would be 15/50 = 30%, three times the industry norm of 10%.
I think the best answer is to pay commissions on an declining scale and timed close to the receipt of cash from the customer (e.g., on booking the annual renewal).
What if F>=1?
Recall earlier that we talked about the fraction, which I called F, that represented the fraction you would be willing to pay to use something for a year as opposed to license it forever. Because of the big difference between “forever” and “1 year,” I led you easily to the assumption that F should be less than 1.
But should it be? When you look at total cost of ownership, it’s not obvious. In the perpetual model you need to license the software, pay annual maintenance, pay typically 4x the license payment in total deployment costs, and buy the hardware on which the system will run.
In the SaaS model, you have the subscription cost each year and some modest year 1 costs to configure the application. See this simple model:
With F at 50% the SaaS TCO is $200K vs. $610K for the perpetual model. With F at 100% the SaaS TCO is $400K. Even with F at 150% the SaaS TCO is $600K — still less expensive than the perpetual TCO at $610K.
And this, by the way, isn’t theory. A friend who worked at Siebel told me that a typical Siebel sales perpetual license seat sold for about $1,500 back in the day. A friend’s company recently renewed Salesforce at roughly $100/seat/month, that is $1,200/seat/year — not quite F=1, but in the same order of magnitude.
Let’s finish the post by seeing what happens to our model when we assume that F=1, i.e., that the SaaS vendor can get an annual subscription equivalent to the license fee a perpetual vendor would have charged.
In year 4, our our SaaS company is now $55.8M or 90% of our perpetual company, but with all the added benefits of being on a SaaS model. In terms of valuation it is now worth $274M vs. $167M for the perpetual company. This is clearly SaaS panacea. The implicit assumption that an annual subscription to use a service should cost less than equivalent perpetual license is both invalid from a customer TCO viewpoint and suboptimal from a SaaS vendor viewpoint.
While this would seem to suggest that every software vendor should switch to a SaaS model, it is important to remember that many customers don’t want to buy — particularly development platforms — on a SaaS basis. Why? Some of it is about ownership and control. But much of it is because many customers think on time horizons much longer than a 3-year TCO. With F=100% in our TCO model (and ignoring TVM effects), the SaaS system becomes more expensive after year 6.
If you like playing with financial models, I encourage you to download the model spreadsheet that I built for this analysis, play with the assumptions, and share your own conclusions. My plan is to do some open source analysis by setting F=35% and the license fee to zero.