Sunday, June 20, 2010

796 acres of data

I just bought a new computer (haven’t had a chance to open the box yet, actually). As an accessory, I got a backup disk drive, a device about the size of your average hardcover novel. This drive cost less than $200 and hold 2TB of data. That two terabytes. Two trillion bytes.

So I thought about the first disk drive, the IBM 350, which held five million 7-bit characters and was able to access any record in an average of 800 ms (that’s eight tenths of a second, which is longer than it takes some people to tie their shoes).

The 350, as you can see from the Wikipedia picture, was the size of a side-by-side refrigerator and needed five feet of clear floor space around it. If you were to arrange units in a rectangular grid, I figure each unit would occupy about 65 square feet. It would take 533 IBM 350s to hold 2TB of data, and that means you would need close to thirty-five million square feet of space, about 796 acres of air-conditioned floor space. I suppose you could house them in, say, forty-story buildings, reducing the footprint from 796 acres to about forty acres (buildings need space around them). Then there’s the power requirement. And then there’s the question of how much longer than 800 ms it would take to retrieve the data.

Okay, that’s unfair
Let’s skip ahead to the first IBM disk unit I actually saw, in 1962. That was an IBM 1301. The model 2 stored fifty-six million characters and had slightly faster access times (100 to 800ms). But it had the same footprint as the 350, so now we’re down to seventy-two acres.

And what did that seventy-two acres of data cost?
You could have leased it all (exclusive of the real estate and without accounting for the cost of electricity to run and cool the units) a mere $168,000 a month. Or you could have bought the ball of wax for $8.9 million.

Now let’s talk about how the cost of doing business has risen.

Saturday, June 05, 2010

Print pricing 101

It never ceases to amaze me how many people make it through life with apparent financial success and yet cannot do basic arithmetic. I wonder, sometimes, how they have managed to stay out of jail and out of bankruptcy. And yet there they are. And a lot of the people who so amaze me are salespeople. In particular, they are printing salespeople, because I buy printing services on behalf of my publishing clients.

The basics
For the sake of simplicity, I am discussing only book manufacturing in this post. Book manufacturing consist of printing either digitally or by offset and binding the book with a soft or hard cover. There are further divisions within those categories, but we don’t need to concern ourselves with them for this exercise.

In digital printing, a company may have some administrative costs for setting up a new customer and for uploading and checking files, but these costs are modest and may or may not be itemized on a customer bill. Once the files for a book are loaded onto a server, every copy costs the same amount to produce, from the first to the millionth. The printer may offer some volume discounts as incentives, but the price per copy does not drop rapidly. If you’re going to save money, it will be in ordering quantities that are efficient to ship. The printing and binding costs don’t change much, if at all.

In offset printing, on the other hand, the cost to set up the job is significant, whereas the cost to produce each additional copy of the book is quite low. As more and more of the prepress tasks are automated, the setup costs keep coming down, but they are still important in pricing.

That’s where your trusty calculator comes in
When you get printing quotes, typically you will ask for pricing on two or three different quantities. A given printer may suggest that the lowest quantity might work out better printed digitally and the higher quantities should be printed offset, but you can get all quantities quoted both ways if you want to and if the job could go either way. Some books can only be printed offset, for a variety of reasons.

But now you’re looking at the quotes. You can see that the digital prices don’t change much, if at all, with quantity. That’s as it should be. But then you look at the offset prices and you see this:
  500    $9.20/copy
1,000 $5.05/copy
1,500 $4.27/copy
What can you learn from that?

The first step is to multiply out the totals:
  500 @ $9.20/copy = $4,600
1,000 @ $5.05/copy = $5,050
1,500 @ $4.27/copy = $6,405
The next step is to subtract to find the differences between totals:
  500 @ $9.20/copy = $4,600
1,000 @ $5.05/copy = $5,050 = $4,600 + $450
1,500 @ $4.27/copy = $6,405 = $5,050 + $1,355
What do those numbers tell us?

The difference in cost between 500 copies and 1,000 copies is $450. If we divide $450 by 500, we see that the second 500 copies cost $0.90 each.

The difference in cost between 1,000 copies and 1,500 copies is $1,355. If we divide $1,355 by 500, we see that the third 500 copies cost us $2.71 each.

Now this makes no sense, because once the job is running, every new copy should cost the same. Certainly we don’t expect the cost per copy to jump up! What’s going on here?

What’s going on is that the person who did the pricing is either working from bad source documents (charts, tables, or incorrect software) or else just made a mistake.

But suppose it really does cost $1,355 to produce 500 books. Then the cost for the first 500 should be no more than $3,695 ($5,050 - $1,355), not $4,600.

Or suppose that it really costs $450 to produce 500 books. Then the cost for 1,500 should be no higher than $5,500 ($5,050 + $450), not $6,405.

The numbers I started with are from a real quote from a real printer. I didn’t make them up as a fake example that would never really happen. In this case, the printer will fix the numbers or I’ll use someone else.

The point is that you always have to check the numbers a salesperson gives you. Not checking them can be expensive.