On the one hand: I have subscription fatigue. I've reached a point at which I'm hestitating to use services because the number of current $10-$20 subscriptions I have no longer fits in L1 cache and it often isn't worth it to me to evict cache lines to figure out what I'm spending --- and so I bounce.

On the other hand, next to the cost of having employees, my rational brain knows I can't possibly be spending real dollars on these things. Put in perspective: every Matasano employee has an Amazon account for which they can buy any book they want ever --- we even have a rule that says "don't worry if someone else bought the book, if you want the book it's yours and you keep it and you can share if you want to but don't worry about it" so that people won't waste time coordinating books instead of just getting what they want and getting back to work. This policy, which we concluded won't cost us "real" money ever, is much more expensive than all my $10-$20 subs ever will be.

But the subscriptions "feel" more expensive.

The answer to the original poster's question is: these things cost $10-$20 because the people running the service feel like they will generate approx. $10-$20 worth of value, and those are the terms under which being in business to generate that kind of value makes sense. In many cases, if the service actually turns out to be worth $3/user, they'll shutter it and pick something better to work on.

There are a very few conditions:

## Terms and Conditions

### Limits

This benefit includes no stated specific limit on the number of books you can acquire or the subject matter they involve. We'd like to keep the policy as simple as possible. Our only request is: please intend to read or use whichever books you acquire. If you think you'll want a book on Clojure programming 6 months from now, consider waiting a couple months before getting it.

### Multiple Copies

Please do not obtain more than one copy of the same title, unless you've lost or somehow destroyed your first copy.

### Ownership and Sharing

When you get a book under this benefit, it's your book. It belongs to you.

At the same time, we hope books acquired under this benefit will help build an office library. If you're not using a book, please give it a home in the office library.

We'll provide a wiki page to register what books are available in the library, so you can easily tell whether you should look for a title in the office or go ahead and order it.

If you know the library has a copy of something you want, please use the library copy until you're sure you need to own your own copy. We are not asking everyone to share a single copy of the IDA Pro book, but we also don't want to buy 10 copies of "Designing BSD Rootkits".

### Crazy Ridiculous Books

Please do not use this benefit to try to acquire collector copies of first-edition signed Knuth prints that cost $1000, or, for that matter, to acquire professional books (like ANSI standards) that cost hundreds of dollars. If you need a book that costs hundreds of dollars, use "How Do I Get Books" option 3 and talk to Cory or Dave; we'll figure something out.

---

Nothing in there says you can't use this policy to buy Denis Johnson's _Tree of Smoke_ to read on the train. Nobody has, but if someone did I'd probably just smile.

Glad to see I was mistaken :)

"Use your best judgment at all times."

For something you want, maybe. But a lot of these services fall into the category of "I might find it useful sometimes but I'm not sure" and most people aren't going to spend $15/month on something they're not sure they want.

The average Tarsnap revenue per user is about $3/month, and a large majority are less than that (yay lognormal distributions!) but I still work to convince people that it's worth spending that money.

You think this is a good thing, but it's not. Instead of setting a simple price based on what Tarsnap is worth to people --- a task that requires figuring out what Tarsnap is worth to different people and selecting the segment of the market that(a) you want to serve and (b) is sufficiently lucrative --- you have a fundamentally cost-based micro-metering structure that almost certainly radically undercharges most of your users.

You may think you're doing your users a favor by letting them off easy. But you're not. If Tarsnap is worth $11/mo to most of your users (hint: it is) and you charge $10/mo, you're generating value for them. But instead of doing that, you charge them almost nothing, and as a result are much more limited in what you can do for them. Last I checked, you were consulting instead of full-timing Tarsnap.

If you charged more, you could afford to:

* Radically overhaul your website and educate users about what secure backup actually means

* Commission a better user interface for people who you'd like to offer secure backup to but can't because they're not going to use a Unix command-line tool

* Implement more and better features on the backend of your system for tracking data, scheduling backups, backing up different kinds of data, and whatnot

This is all apart from the obvious point that your micro-metered pricing scheme is a Mensa test that almost makes fun of people's attempts to figure out what your service costs. Yes, the fact that it costs much too little means the joke is on you, but that doesn't really take the sting out of it for your prospective customers.

One example, which I use for hosting, is nearlyfreespeech.net. Amazon's and Google's cloud services have a similar utility-pricing model, though I'm not sure how much I personally trust them to price in line with costs, especially whether they're going to pass on future cost decreases. Prgmr.com is a flat fee rather than a combination of micro-prices, but it has a little of the same vibe of, "I calculated how much it should cost to operate these servers, divided by the number of VPSs, added on a reasonable profit margin, and that's what I'm charging", which is one reason I went with them.

It's not even necessarily the overall price that worries me so much as liking the general warm-and-fuzzy feeling that I'm not getting a hugely marked up service. Same as when shopping for, say, a mechanic: I feel most comfortable with a mechanic I can trust to charge me what the repair costs, plus a reasonable profit, not one who's trying to size me up to figure out how much the repair is "worth to me".

Just to set the record straight on one point, though:

Last I checked, you were consulting instead of full-timing Tarsnap.

Yes, I do consulting, but it's a very small fraction of my time. And honestly, I think what I learn from the consulting is worth far more than the money I get.

It is, as Patrick will testify after many an evening conversation in Chicago, "all love" with you & me & these critiques. Your project should be epic by now, and it's driving me a little crazy.

* you have to find out what the value your service generates is to different groups of customers (an MBA would call them "segments").

* you have to figure out the cost of acquiring and maintaining customers in each segment.

* you price based on the value (if it creates $10/mo apparent value, that's the number with a gravitational pull).

* bust out cleverness to figure out ways to capture money from high-value low-volume segments without scaring away low-value high-volume segments.

I'm close enough to Colin's target market to know that "just stick the server's filesystem in tarsnap and don't even bother taking an inventory of what's on the machine" is worth multiples of S3+80% just in recovered admin productivity.

Here is a hint, Colin: any time a prospect says he's "run the numbers" on your pricing, you just received a strong signal that you need to simplify pricing. Purchasing decisions should be made in the limbic system.

Think about it: if you charged $5/mo and put an obscenely high cap on storage for the $5/mo account, then by your own math you'd be hugely improving your revenue, and at the same time you'd be charging half of this original poster's pain threshold number --- which is referring not to crazy encrypted backup solutions but to things like to-do lists!

At EnterAct, the ISP I ran tech-ops for, my old boss Mike Cloran came up with what he called the "5/30" pricing scheme, which we used to great effect to grab customers from other ISPs. Other ISPs were either metered (which scared the shit out subscribers) or flat-rate (which chased away light-users). The 5/30 plan was: $5/mo + $1/hour, capped at $30.

This is actually not a great deal and it is not particularly sophisticated (it is... wait for it... a capped metered plan). But it simplified the buying decision for people, addressed their perceived risks, and made EnterAct seem even more flexible than it already was.

Pricing is the hardest part of product marketing. There's a reason why established companies don't let just anyone do it. Your pricing scheme punts on it.

... and also hugely increasing my costs. With all due respect, I've talked to a lot more Tarsnap users than you have, Thomas, and I also have the advantage of seeing how much data people upload and how much they delete.

Your suggested pricing model would result in people uploading far more data and never deleting any of it, and would very rapidly move Tarsnap from "profitable" to "bankrupt".

If Mozy can charge 0.50/mo/gig, 0.30/mo/gig is way too cheap.

The security is of lesser consideration for me, the convenience was the main selling point.

A critical part of trust is transparency, and in the long term I think I'll do better by having transparent and obviously reasonable pricing than by obscuring things and trying to maximize my (short-term) profits.

I didn't know about either of your companies but by the discussion, its clearly possible for Tarsnap users to derive value from the trust associated with transparency and feeling like their dollar is maximized. tptacek, pricing schemes aren't the only component of the purchase decision or profitability. His users are making a more emotional decision. cperciva should stick to the core of what his users are finding valuable...then get bought out and turn into an evil profit maximizing company

Using S3 would cost me more per month unless I implemented a similar system to tarsnap.

It is like the Iron Mountain of backup solutions.

Not specific to cperciva: If you have customers who value your service less than a cup of coffee, fire them. If this leaves you with no customers, consider a new app or marketing strategy.

Vancouver, Canada. (Well, Burnaby technically, but close enough.)

I want to get a coffee some time [...]

Sure! Let me know if/when you're ever in Vancouver.

If you have customers who value your service less than a cup of coffee, fire them.

I disagree. The money I get from Tarsnap's $0.01/month users is insignificant, obviously -- but they're worth many $/month for their willingness to proselytize. The utility pricing model is like the freemium model, except it weeds out the people who aren't willing to pay anything at all.

And those with just too much data and too little budget :)

Check out what tarsnap does so you can compare tarsnap with a tapedrive.

This (S3's storing everything so redundantly) is what I see as the main benefit that tarsnap would get you, since the encryption itself is the easy part---just GPG your dump before writing to tape.

Now, I might be strawmanning you. What does tarsnap get you in particular that tapes do not?

When I saw the pico-dollars bit on the site I instantly knew this was not someone I wanted to do business with. The sort of person who tries to prove his intelligence constantly rarely has the ability to cut through the fat to get into the meat of the matter. They'll play in the weeds all day long, but never climb the mountain for a good view.

I agree with Patio11, buck up grow a pair and charge good money. It looks like you've got solid infrastructure, it just needs to be charged for and sold.

Seriously, switch to $/Gig, and sell the hell out of it. I don't care how smart you are. Show me how you can make my life better/simpler then get the hell out of it.

1. True story. I've gained a good nose for these people and I just walk away after the first encounter.

The word "picodollar" is basically a dollar sign with a <blink> tag around it.

I thought - this guy has DJB sized balls, I really want to use it. portsnap was great, this will be too.

And I don't even have a beard.

Colin has a platinum backed BSD pedigree. Tarsnap appears to be aiming at BSD minded users in the first instance ie people who mostly know what they are doing. Colin's grown tarsnap out of what he knows, the BSD aesthetic. What you know is a really good place to start.

But now, to grow tarnap beyond the BSD commandline pico-niche, something I really hope Colin does, because he's earned the success, it could / should adopt a Jungledisk like model. With an OS X desktop client, clearer pricing, and Apple level UX/marketing. And clients for home storage devices like Qnaps etc. And yes he should charge much more for that. Tarsnap does seem bizarrely, wonderfully cheap. And long may it remain that way... on the commandline.

Not listening to your customers will make it rough road for your company. Your customers say don't price in picodollars. Your customers say price flat rate and figure out how much storage you can afford to give them for that rate while still making a healthy margin. You'd do well to listen to them instead of justifying why you should continue to scare potential customers away.

Tarsnap should white label the service and keep the existing pricing. It will increase their volume drastically and the really low pricing will allow a lot of people to build a very profitable business on top of the service.

Someone then can charge 20$/month and a buck a gig for storage with some flashy mac/iphone/ipad app and kill it.

If you want to build a service on top of Tarsnap, please let me know and I'll do my best to provide you with whatever interfaces you need.

- I spent $18 a month on Netflix without a second thought.

- $15 is one less meal per month at Chipotle (for 2 people). I'd gladly give that up for a decent service

- Average cable bill in the US is probably $80+. Give that up and you not only waste less time, but can afford at least 6 of these services.

- Once you're out of college, you will GLADLY give up $15/month to save time. Time with family is precious and anytime I can spend a small amount of money to gain more time I consider it a win-win.

- Most middle class families probably spend more than $15/week at Starbucks alone.

- From a business standpoint, $2/user/month is simply not sustainable.

(b) Yes.

Continued: the green trains are older, they have no airflow, they're slower, they're consistently much more crowded, and they're rowdier.

The #91 bus down Austin is so much better than the green that it's a no brainer just to bus down to the blue line station. It helps that the train stops right outside our office, though.

increases since then seems to be 5-7% per year.

good guess!

I would be happy if our 20k users all payed for a year up front discounted by 30%, rather than a full-price for every month. Why? Because we don't have much income now, and getting the money now would allow us to expand much faster, so the next 20k users would signup quicker, and would not have the big discount for an entire year up front.