2013-03-08

Don't Automate: Lessons From The Hacks Powering Outbox, The USPS-Killer

How do they get into all those mailboxes? “We came up with this idea of the user taking a picture of the key using a cell phone and emailing it in,” says the CTO. “We’ve written some software, and with some light user-interaction, can generate a code for a key. We have a key machine--it only takes a couple of minutes.” That hack and more are what's driving Austin's most insane startup.



Getting snail mail onto your MacBook is both an enormous technical challenge and the entire point of Outbox. The Austin company has had to solve discrete hardware and software problems in every part of their business; nothing is off the shelf. In the quest to digitize people's mail en masse they've built their own tools, some of which are novel enough that they could be spun off as separate businesses. But the founders stay focused on a singular goal: combine digital and paper mail in a single portal. (Yes, they're trying to kill email, too.) Here's how it works.

What’s the technical problem you’re solving?


Jason Seriff, CTO of Outbox: The basic problem is getting pictures of mail. You can't run this business until you have a way to do that at high volume, high accuracy, low touch. Your basic options are hand scanners, $200 flatbed scanners, and $10,000 page scanners, all the way up to your quarter-million dollar high speed scanning equipment.


Whoa, whoa--why invest so much time and money taking pictures of peoples’ mail?


Will Davis, Cofounder of Outbox: There's a strange government monopoly on what can be put inside a mailbox, and that leads to some very strange incentives on the part of the government. They're incentivized to grow their employee base so they can increase the efficiency of how junk mail is delivered, because it's their bread and butter--worse for citizens, environment, and the government because the costs are just running away from them Evan Baehr, Cofounder of Outbox: The idea is not to be a scanning business to scan people's mail. The idea is that we want to be a content portal where people want and consume content. Let's create a new form of email--that’s one way to think about the goal. The only way to get there is to take over this older "content portal," which is mail, by making that portal much better--by making it digital. After people sign up, they’ll need to go to this content portal: long term, we get a bunch of customers logging in daily that they have to visit because something good might be in there. We want to be out of the scanning business as soon as possible. We're doing this to get users and bring in consumer behavior into this consumer behavior. We're not FedEx. We’re more like Netflix. When they started digital delivery, they began the business with a pure digital view of distribution for videos. That's our view. It's a physical product that only makes sense in light of the digital product.

You’ve had to solve a lot of problems with hardware. How did you know what to build and what to buy off the shelf?


Baehr: Last summer we met with Andy Rendich, the COO of Netflix. When Andy showed up at Netflix eight years ago, they were shipping 5,000 DVDs a day--these were DVDs in cardboard boxes and on folding tables they got at Sam's Club. Andy taught us how to do lean hardware development. And, like every other element of lean startup, we only automate or develop hardware for process bottlenecks. He showed up [at Netflix] and they had a 15-step process from intake to reshipment, all manual labor. He applied a rigorous evaluation process: How accurate is it when a human does this? What's the cost when a human does this? What's the expense of a hardware solution and what's the payback schedule? That meant a very high bar for automation. He has a nice anachronistic counsel: be slow to automate.


So what did you build to scan mail?


Seriff: We built our own scanning rigs from an off-the-shelf DSLR, a bunch of lighting, custom stands, and some computer vision software to strip out backgrounds and determine the size of the mail. The image that comes out has a very life-like feel. Baehr: We chose to build those things for a reason. It comes down to one question: at what point does the letter opener that you bought for 69 cents at Staples become the bottleneck in the process? We looked at these industrial scanners, $250,000, bank line of credit to get the machine, with a three- or four-month lead time. In demos, we’d bring our own bag of mail and say, can we try? The machine would usually break. So, be rigorous applying an ROI-payback metric to your decision to do any kind of automation.


How do you get into users’ mailboxes?


Seriff: That was the next problem. We built this thing and we started getting beta customers who were willing to give it a go. In Austin about 85% of mailboxes are in apartment buildings or in a locked cluster. We played with a bunch of ideas and the original ideas had no technical solution. They'd mail us a copy of a key—very high touch, very painful. We quickly figured out that wasn't going to work. We came up with this idea of the user taking a picture of the key using a cell phone and emailing it in. We’ve written some software, and with some light user-interaction, can generate a code for a key. We have a key machine. It takes a couple of minutes. At first we used a locksmith which was about 40% accurate; Software gives us 98% accuracy. If we get one we're not sure about, we cut several keys and send them to the customer and they keep the one that fits. Software not yet connected to the machine--not anyone does that. We've joked about letting people send us pitures of their keys and send copies as a business--but not something we're gonna pursue.

How do you programmatically sort the mail after it’s been scanned?


Seriff: The stumper here is identifying the mail. We capture the images and show them to users. But what if we want to show a digest? That's something we're exploring now for the next few months. We’re doing OCR and checking [mail against other mail] so that if we get 100 pieces of mail that are the same--we know this is clearly junk mail. We think that's interesting, but we’re also working on full-text search of mail, auto-categorization and data extraction. For a bill, you mostly care about the due date or cost; most people don't flip through their entire bill. We'd love to be able to extract out that critical data and hide the rest. Then the images become a less-needed piece of the software, something you can dive into if you need context. It's really about indexing and searching the data on paper, freeing it, and looking at personal analytics about mail.


Did you ever approach USPS about a partnership?


Davis: Back in 2011, we were doing a test with USPS in Austin that was working really well. We were asking the postal service: please don't take this person’s mail to their mailbox; don't even put it in the truck, we'll take it from there. But eventually we were called into the Postmaster General’s office to meet with the chief of digital strategy and the USPS legal counsel. They said, We'll never work with you. Baehr: Then the head of digital innovation for the USPS looked at us and said: "That will never work anyway. Digital is a fad. It will only work in Europe."


And that didn’t convince you to quit?


Seriff: At that point we had to change direction. That’s when we came up with the un-postmen: the fleet of drivers that pickup peoples’ mail [to be digitized]. We've managed some basic road routing software, so our drivers go around with iPhones that have access to our data and assigned routes. Davis: Also, it's not actually illegal to remove things from the mailbox, like a lot of people think it is. Although it is illegal to put things into a mailbox.


Is this one of those "we’re safe because no one else will be crazy enough to try" businesses?


Baehr: Yes, with an exclamation mark! This is a long-term play of enabling a new digital communication channel. At the level of operation, we're laser-reading, and driving trucks, and opening peoples’ mailboxes. There are other ways to do this easier as a business--have people forward their mail--but we reached this model because we think it's the only way to deliver the level of quality that we want to give to our customers. And that makes it totally insane.