True which is why one of the more mature (timeline) SaaS companies, Dropbox, decided to in-house their infrastructure to bring costs down. While this was a massive undertaking that took many years, they were able to achieve it and are reporting various levels of profitability (free cash flow). Also if you dig in further to most SaaS companies P&L you'll see that infrastructure by and large isn't the biggest driver of losses, but sales & marketing is.

Personally I think companies still continue to overspend in sales in marketing and with SaaS they use a blended average of CAC which includes their organically acquired customers so that the blended CAC looks extremely low. But the reality is you are paying 10x more for "paid" customers than organic and adding an only incremental amount of growth (5-10%). I think that's where the real profitability of SaaS companies lies.

I've worked in finance for several SaaS companies and you're 100% right. Infrastructure is never the biggest cost, it's sales and marketing and most of that spend is a mystery in terms of attribution, so blended CAC is not terribly useful.

The more customers you get, the more customers you get from "organic" sources. Blended CAC makes sense from this perspective.

The amount of cash Snapchat is throwing at GCloud/AWS according to their recent IPO filings is just insane.

They are the exception for sure, but their platform also supports a tremendous amount of video and that's the real killer for them. This is the same problem that youtube had, video just consumes many times more bandwidth than doing a similar service in text (twitter), or images (instagram). So that's not surprising.

But one exception doesn't make a rule.

How would you avoid blending the cost to acquire paying, non-paying, etc.?

It’s not easy to get an exact figure but you can fairly easily split out customers who arrived via an ad vs those that didn’t. Likewise it’s easy to split out those customers who arrived via cold selling

For starters I would separate them out.

Paid I would consider traditional forms of advertising and marketing such as TV, FB, Google, ads, etc.

Organic or community driven advertising has it's costs as well but I would separate that out to a different section. This would include budgets like meetups, some conferences (though many I would actually lump into traditional ad spend), t-shirts, stickers, credits to customers, community expenses and so forth.

Then for non-tracked customers I would assume that percentages work well. Meaning if 25% of my customers come from community, then 25% of my non-tracked organic customers are from community. If 5% are from paid, I would say 5% of my organic non tracked are paid.

If you do the break out this way you will see that old-gaurd paid advertising is 80% of the cost of the entire sales and marketing function, but brings in an overall 5-10% of net new sales.

Now if you are profitable, or near profitability go for it, but if you are losing a sizable portion of yearly revenue, say 50% or more due to sales and marketing and you see it broken out this way you will immediately realize how false this spend is.

Just because something is done a certain way and accepted, doesn't mean that it's correct.

Most of the public SaaS companies around today spend a significant part of their S&M expenses on sales reps, and their associated selling commissions, not advertising and marketing. There are exceptions like Atlassian however, where they have managed to make the economics work at scale without a sales rep or any implementation assistance.

> maintaining a large and expensive ops team and dealing with ever-growing AWS bill

Set it up on real hardware with 2-3 good admins and you'll save 6x per year. AWS has its place, but it's unlikely that place is anywhere near 100% of everything. This truth has practically become a heresy, which means you can make good money by having the gumption to take advantage of it.

> Some are basically in the business of being an intermediary mailbox for transferring money from their customers to AWS.

Does this mean that you can win just by hiring one of the few people who can accomplish the terrifying task of, GULP, setting up a server without dragging around little blocks in a web GUI? :P

I don't disagree on a lot of this but I'm willing to give businesses like Workday, Twilio and Salesforce more time to develop their business before saying the market has decided.

I had to look it up, but Twilio gross margins are healthy -- 55%. AWS expenses are not really a problem for a company like Twilio, their well publicized arrangement of relying on other large tech companies is a bigger threat.

Sure, the gross margins are healthy for many SaaS companies. Take a look at the other expenses. They eat that margin, and then some, quarter after quarter, year after year.

Where does it end? Will these companies ever be profitable? I'm invested in several public SaaS companies, just following the trend...

I think that's because too many SaaS CEOs think something else mentioned in the article: "Margins, to a first approximation, don’t matter...If they’re quickly growing, the company can ignore every expense that doesn’t scale directly with the number of customers"

It makes sense in theory, but businesses don't grow to infinite size so fixed costs do matter. Ignorance of this fact drives the lower than anticipated gross margins of many SaaS businesses.

> Ignorance of this fact drives the lower than anticipated gross margins of many SaaS businesses.

It's intentional. Great margins don't matter if you don't have customers, and no VC is going to get excited about a few extra margin points on OpEx when should be focusing on growing.

Selling recurring subscriptions for software requires continual updates, and lots of customers are still put off by that model. Yes, there are a few megaliths with untouchably huge products that are sold as regular software; but how many of them is there room for in this world?

As someone building a company with the model you speak of: do you think communicating to potential purchasers via website copy that others in my industry have essentially abandoned their platforms because of lack of incentive to continue updating post-purchase would perhaps soothe some of that?

We also use this model (software subscription). One thing that is very important is that never, under no circumstance be tempted to sell perpetual licenses even if there is demand for them. If you do subscriptions you should do only subscriptions.

Thanks, that's advice you're probably giving for a good reason you've experienced, I'm assuming!

Any other insight into your experiences/troubles? How you decided on pricing, whether it had any impact on sales/sales efforts, what you had to do to support such a model?

Thanks very much for your time!

We transitioned from freeware to subscription (it's B2B), a subscription model shares a lot with freeware logistics, for example most users are in the last version. Another advantage has been to be able to add features any time it was fit, no need to save them for the big upgrade. We decided a price at the beginning (EUR 99/year) that we thought would allow us to expand our offering under the same subscription umbrella. We run several promotions to experiment with pricing (in particular a Dutch action which was very successful at EUR 29/year, that brought a big revenue spike and also many new paying customers). But even after experimenting with good results we always stuck with our EUR 99/year price point. Due to the rather low price pint, we chose not do any type of customer segmentation (all customers pay the same per user, regardless of quantity of users). We do not do promotions anymore to avoid confusing users with pricing (since it is a subscription, we want to avoid that users wait for the next promotion to renew). Curiously very few potential customers ever asked for a permanent license. We have a extremely generous trial 30 days of actual use, not calendar days. Conversion at the end of trial is high, about 30%, probably because by the end of trial, users probably found a few use cases that somehow justify the price.

No, they just won't care.

I was on a non-software forum last week where a dev was gathering pain-point information because he intended to produce software for that niche. The main feedback I remember was "just let me buy it. No f'ing subscriptions."

Any chance you can share what niche this was?

Machined-object Contract Manufacturing. e.g., https://www.milacron.com/services/contract-manufacturing/

What does the software for that industry track?

Is this really a problem?

A SAAS client of mine's operational costs are like 1% of our revenue. Ok, they're at Series A/B funding sort of size. they've started to go international and we could also quite easily do an instance per country if we wanted, though I don't know what sort of pain that might cause.

Running ASP.Net/C#/Azure. Azure's a bit of a nightmare, would be running much better/cheaper on a dedicated server in my professional opinion (shock, horror, yes, I said it, cloud is a bit of a con).

I've spent 2-3 weeks per year "scaling" the code, which was written by other people who in some ways didn't know what they were doing and were trying to make it too much like a SPA which resulted in lots of ridiculous, excessive dupe calls for data and over-complicated code. After my last round of optimisation I doubt they'll need to scale up for a couple of years now, the CPU was basically sitting less than 20% and DB was about 40% even in the crazy Xmas rush, and we were on standard tiers (compared to the year I started where they had to turn up to premium a couple of times with 10x less users).

Isn't this more a problem of billing than of SaaS in general?

But yes, I recently read that games make more money in general than even movies, books and music and most of them don't have subscriptions but one time payments.

It has nothing to do with billing. I was just pointing out that the second 'S' in 'SaaS' means "service" and once you get into services your margins often go down, that's why I (and other folks I know who've done SaaS at scale) disagree with Patrick on "economics of SaaS are impossibly attractive". I am willing to bet that Splunk's margins are way better than Loggly's.

Moreover, if you're smart & lucky and you manage to find a large niche with good margins running SaaS on AWS, Amazon will launch it as a service. Mr. Bezos is around & watching you :)

[EDIT] Besos -> Bezos :)

Not to be a jerk by pointing at a innocuous typo... but "Mr. Besos" put a huge smile on my face. Especially since "besos" means "kisses" in Spanish.

Mr. Kisses sounds even more ominous!

Let's just go with Besos from now on.

All AAA games have either micro-transactions, additional purchasable content, or both

I seem to recall seeing a report recently that Activision-Blizzard made more money from micro transactions than non-MTX revenue last year.

The payment model used in Starcraft 2 - a ancient game at this point - is sophisticated. A good illustration of the old game-economics model fused with the new. Oldest single player content is free; multiplayer is free; newer content and cosmetics are purchased in-game a la cart.

Instead of paying $45 one time, you might spend $120 over the course of a few years.

I suspect that will include game time tokens for WoW that can be then sold for ingame currency or used for gametime.

They also basically converted a big chunk of their subscribers to pure mtx players with that move further muddying the signal.

They also own... King was it? One of the big mobile companies, and I think over $1 billion in MTX revenue was just their products.

I think those are great points in where the underlying business model could have underlying cracks. My take is that SaaS has desirable features in the current age of agile product development and data-driven business decisions. Selling an Oracle database is a bit like going door-to-door to sell vacuums. Ultimately, it is hard to gauge product adoption, CAC, etc. with a licensing model at first since you aren't selling a $5/month per user solution. The Oracle database could actually scale better as a business model as once the customer is set up they can move to the next sale versus always battling how much it costs the company to maintain the service. However, in this current day and age, it takes a little faith to run this model versus one that has daily performance metrics from the onset.