The second thing I noticed was how, for a Web-based business, there were few computers -- maybe seven in the building,
Don't let that mislead you. Make no mistake about it, this is a highly automated business with a lot of thought, design, analysis, and most likely, very tightly-targeted software.
With all the talk here at hn about cool stuff like algorithms, scaling, UI design, CPM, etc., etc., etc., I still spend over 50% of my time helping customers move shit from Point A to Point B. That's it.
It may sound old school and boring, but it's anything but. When you're doing so much volume, every little thing matters. And lots of those little things can be dramtically affected by good systems. Here's where you have to get good or die trying. When the smallest mistake turns a 50 cent profit into a $100 loss, your systems must be on the razor's edge.
I bet the routing and error detection software in these warehouses is very sophisticated and handles outlying cases most people would never think of. Perhaps some journalist will take us on that trip with the next article :-)
Netflix has deployed the largest "legacy adapter layer" business process that I can recall. This is a great example of how to start a business where the end-state model is dependent on things outside of its scope of control (or even influence).
When Netflix started in 1997, Most of its potential customers were using dial-up, and there was no way Hollywood would have allowed digital distribution anyway. So, they rode on top of some existing infrastructure: (1) the US mail system, and (2) existing laws that allowed physical media movie rentals.
My wife and I have been Netflix subscribers for years. However, most of our viewing is now done via the little Roku box, and physical disk delivery is of a much secondary importance (their streaming media is better than DVD but not as good as Blu-Ray).
Netflix is in this fantastic position of having a large customer base that can shift from its relatively high marginal cost* business model to its new, much lower marginal cost* * one.
* postage and manual labor.
* * I presume the movie studios are getting paid per-view?
About marginal cost, it's not obvious to me that postage both ways plus sorting is more expensive than bandwidth plus royalties. Given that postage gets more expensive and bandwidth gets cheaper, streaming should be cheaper at some point if it isn't already.
I think I've read that Netflix has an agreement with the studios where they pay per shipment rather than per DVD. So, having 1000 of a new DVD on hand costs little more than having only a few. The studios like it because it aligns Netflix's cost structure with their own (basically fixed), so agency costs are minimized.
I'm sure that Netflix pays less for bandwidth than Amazon sells theirs for. So, we can presume streaming a two gig movie (probably much less in reality) would cost much less than 30 cents -- postage one way for a DVD. Add in a few more cents for data center, etc. and they're still better off relative to postage.
But:
number of customers that want to sit on couch in front of TV and watch rented movie with family
>
number of people who want to watch download on computer * (1 - fraction of computer capable watchers do this for free)
Tiny sales compared to DVD players, almost all download uses require you to do something 'techy'
Netflix market is families and film fans - not people that know what HDMI stands for.
Netflix will definitely be online only one day, probably before Amazon goes the same way, but not for a few years.
The genius of their system is that most titles never ever get re-shelved because the software is smart enough to do the match-making.
""It's a lot faster than a traditional shelving system where you have to run around and marry up orders," says O'Handley. Dennys Torres, one of four original operators when the hub opened, in January 2002, can clock more than 800 discs per hour. The Scan Tomorrows move faster, set aside by the handful. Ship Discs get an envelope and a pair of stickers. "When we first started last year," says Torres, "moving 4 boxes was a lot." He indicates sets of long, narrow boxes filled with DVDs, some 320 discs in each. "We were taking long breaks, two-hour lunches. Now, it's all work; everybody moves at least 10 boxes a day."
"Its biggest secret remains the warehouses themselves -- for two reasons. No. 1, each holds several million DVDs, not to mention expensive mail-sorting gizmos and dry-board posted statistics on how many discs were recently returned damaged, placed in the wrong sleeve or scratched. And No. 2, Netflix has grown leery of what happens when customers learn the location of a warehouse -- they drop off DVDs at the door."
This seems in-genuine for multiple reasons:
1. Other articles I've read on the Netflix Inventory Control system indicate that there are almost _no_ DVDs in the warehouses, the vast majority are either at customers, or in shipment. So the "Several Million DVDs" is an unlikely claim.
2. Expensive mail sorting gizmos? So? Are they afraid of theft? "Mail Sorting Gizmos" neither lends themselves to easy snatch and grabs, transport, nor resale.
3. Dry-Boarded Stats on scratches, damaged, wrong sleeve. Doesn't sound like competitively useful information to me, particularly after data like # of Sleeves per hour (650-900) per employee, and interval of exercise (every 65 minutes) has been made public.
4. Drop Off DVDs at the door - This is a bad thing, why?
Nope, My guess is that NetFlix is trying to delay the inevitable, which is unionization of their workforce by trying to keep the location of their shops confidential and avoid organized labor leaders paying their sorters/packers a visit...
> 4. Drop Off DVDs at the door - This is a bad thing, why?
Because their process is designed around 1000X as many DVDs coming in through a different entrance, probably with some "packaging" that they've optimized around.
DVDs at the door are an exception and exceptions are even more expensive in the physical world. And, that's ignoring the cost of dealing with folks who think that they can pick up new DVDs when they drop off their old ones.
Weird. In this video they have people hand stuffing envelopes but in the original article they have a machine stuffer. I wonder, is the video just outdated?
Am I reading this right? It seems to say that every day, every disc that’s currently is in stock is processed so that any that need to sent to customers can be pulled out for delivery. And my intuition tells me that would be crazy.
"Then discs are scanned a second time -- if a title is requested, and around 95 percent of titles get rented at least once every 90 days, the machine separates it and sorts it out by ZIP code. (The entire inventory of the building is run through this daily, a process that alerts other warehouses of the location of every one of the 89 million discs owned by Netflix.)"
If you DON'T scan the discs every day, then you need to organize them in some fashion so that they could be picked from the shelves when an order comes in.
If you DO scan them every day, then they can just go on the shelves in random order. The scanner will automatically pick out the discs that it needs to ship that day as they go through.
So, you pay extra for machines and labor to scan every day, but you save on the added cost for sorting equipment, as well as random-accessible storage units and people/equipment to locate and pull random movies from the stacks.
This makes even more sense if, as others have suggested, most of the Netflix inventory is deployed in the field at any one time, and not sitting on the warehouse shelves.
Surprisingly manual process - at least on the incoming side. I was interested to know how many discs are stored at one warehouse, but that is probably too specific for them to disseminate freely.
I remember reading an article (don't ask me to find it, it was a long time ago) about how they don't actually "store" many DVD's, they unpack them, see who is next in the queue to receive it, and put it in the pile to ship back out. This sped up the turnaround and reduced storage costs significantly.
I'm surprised too by the manual process described in the article.
I wonder why it itsn't more automated, they could photograph the disc to see if it matches the sleeve for example. It seems like a terrible job to manually check the discs.
How quick and accurate is image matching? They can recognise the sleeve quickly but not, it seems the discs. OCR-ing movie titles from discs would be near impossible with any degree of accuracy.
Is there not a unique code/barcode around the central hole on the back though? Or could they overprint a barcode?
afaik, they try not to store popular items. otoh they do the opposite for the long tail titles. probably from ny times article, they also have changed their package format (both material and shape) 10 times or more.
It's also the reason that the time of day you mail your DVD back has no effect on when you receive your next one.
Huh? Sure it doesn't matter if you send it before the last pick up of the day: Netflix will receive and ship a DVD the next day and I get the movie the day after.
But if I drop it off after the last pickup time (after 5:00pm for me), I get a DVD back a day later rather than the usual 2-day turnaround.
Interesting. I didn't expect Netflix to have a "fun" atmosphere like Zappos or anything, but it's funny to hear about a trendy web company with little, if any, character.
I wonder why the employees are required to wear the red Netflix t-shirts. It's not like they need to look presentable for customers or anything...
I would imagine that working as some sort of engineer in the corporate headquarters is worlds apart from working in the warehouse sorting DVDs, and that you couldn't draw inferences about one from the other...
This warehouse sounds like a job that is far too tedious for me. I'm glad they get full health insurance and a 40 hour work week though.
I can't wait until every title they offer can be streamed to the Xbox360. They have a great selection already, but it will make Netflix even better than it already is!
What they probably did was hand out shirts to the employees to look snappy for the visiting dignatary. Most companies I've worked for try to do the same. Some require higher "dress levels" when visitors are coming (like the current one says "no shorts & flip flops; use business casual" only on days when important-enough outsiders come by.
Don't let that mislead you. Make no mistake about it, this is a highly automated business with a lot of thought, design, analysis, and most likely, very tightly-targeted software.
With all the talk here at hn about cool stuff like algorithms, scaling, UI design, CPM, etc., etc., etc., I still spend over 50% of my time helping customers move shit from Point A to Point B. That's it.
It may sound old school and boring, but it's anything but. When you're doing so much volume, every little thing matters. And lots of those little things can be dramtically affected by good systems. Here's where you have to get good or die trying. When the smallest mistake turns a 50 cent profit into a $100 loss, your systems must be on the razor's edge.
I bet the routing and error detection software in these warehouses is very sophisticated and handles outlying cases most people would never think of. Perhaps some journalist will take us on that trip with the next article :-)