Do It Manually
Solve problems first, automate later.
Welcome back to Jab’s Lab, my weekly musings on startups, tech, business, and life. Today I am talking about how a lot of the early work at a startup can and should be manual. This post is a spiritual successor to Start Small. If the goal of starting small is to test your hypothesis and gain certainty, let’s also find manual ways to test as efficiently as possible. While I see a lot of people rush to build a tech solution, I argue that they should start by doing things manually instead.
In the same ethos as Paul Graham’s famous Do Things That Don’t Scale, this post gives my thoughts on how you should build only what’s absolutely necessary at the start, and do the rest manually. Let’s dive in.
As the resident software engineer in business school, people often look to me for my renowned expertise in all things technology. And by that, I mean they ask “can you build my app?”
I started diving deeper with founders to learn more about their app ideas. What I realized is that while many people have an incredible elevator pitch1, they often lack the knowledge of how to get started.
I’ve become incredibly curious about how to get business ideas off the ground. It’s easy to see existing companies and know exactly what they do. What’s harder is to see how they got there. And I would argue that more often than not, they started pretty manually.
DoorDash, Airbnb, and Doing it Manually
I had the pleasure of hearing Evan Moore (one of the founders of DoorDash) speak the other day at SeedCon, a startup convention hosted by UChicago’s Polsky Center. His story was fascinating. DoorDash was started by four guys who lived in the Bay Area and decided to put up a website to let others order food. They were solving their own problem, because they could not get takeout delivered from their favorite restaurant.
How did it work? They had a small website where they uploaded restaurant menus as PDFs. They had a phone number on the website (it was their own cell number). People called “DoorDash” and placed their order. Evan and his friends wrote the orders down, called the restaurants, and placed the orders. 30 minutes later, they drove to the restaurants, paid for the order, grabbed the food, texted the people who ordered the food that the order was on the way, delivered the food, and collected the cash.
You can see how every step of this process is the same as in the DoorDash app today. But instead of automated technology for each step of the way, they did it manually.
Imagine instead that they decided to build a food delivery app. They could have spent months or years building integrations with different Point of Sale systems, perfect every single user experience flow, and make sure it was the easiest, most reliable app on the market. But if they couldn’t get customers to use the app, then what would have been the point?
Only when they reached the limits of their own manual labor did they decide to build tech. And reaching the limits of doing it manually is a good problem to have — it means people want what you’re building.
With their initial website, DoorDash validated the demand for their service, and did the rest manually. Today, DoorDash is a public company with an $88B market cap. Not too bad for some guys that started by calling restaurants to place orders.
Airbnb had a surprisingly similar story. They started by renting out space in their houses online to people visiting SF for a conference. They literally blew up air mattresses2 and sold some floor space. Later, they decided to build a simple app, but nobody was booking. Realizing that many of the hosts’ photos were low quality, they flew out to NYC and famously helped new hosts take professional pictures of their places to increase bookings. It wasn’t the app that made Airbnb successful in the early stage, but rather the manual steps they took to get people to start using it.
Solve Problems First, Automate Later
Everyone wants to build a tech product. There’s no doubt about that. It is sexy and flashy to have a well-designed product. What’s not as sexy is a buggy app that almost solves a problem for someone, but not quite. I don’t think it’s a black and white “you either have a tech product or you don’t.”
Oftentimes, you can build the bare bones app and do the rest manually. It’s probably best to leave the really hard technical stuff out of v1. Take a page out of the DoorDash playbook, and make it as manual as possible. These hard technical challenges are often the easiest things to do manually. For example:
Ordering from a restaurant via a complex Point of Sale integration: hard.
Ordering from a restaurant by calling them: easy.
Start by doing the manual things first, knowing you can automate these things later.
How Do You Do It Manually?
With all of the tools available today for founders, I think it’s more important than ever to understand that the essence of building a startup is focusing on identifying and solving a problem for a specific user. Ideally one that people will pay for. And then identify the ways you can solve it manually.
If you have an idea, instead of jumping right to building tech, ask yourself: “what can I do manually that will be easier than building custom software?”
Want to build a platform to bring people together? How can you start building that community before you have the tech? Why not start by spinning up a Circle community?
Want to build a new database of AI tools? How can you aggregate them first and show them to people? Why not have a public Notion page that people can pay to access?
Want to package your knowledge into a product so that others can use it? Why not offer to do it yourself for them as a service, and leave the self-service portal for later?
Figure out what you can do manually, and what you actually need to build custom tech for. Software is good at increasing usability and scalability. But if you’re building something that people don’t want, it’s better to put that solution in front of people quickly, get feedback, and iterate. Trying to “perfect it first” is a surefire way to fail.
Hopefully you get to a point where you cannot do it manually anymore. At which point, you’ve earned the right to build some custom software.
And hey, perhaps you’re reading this article 6 months from now and AI tools have gotten to the point where you can spin up custom software effortlessly. Maybe it’s so easy to build tech that you just automate things to start. But in that world, an entirely different question comes to mind: What problem will you decide to solve? Perhaps a topic for another post.
Until next Tuesday,
-Cory
When can you call it “a startup”? When it’s just an idea with an elevator pitch? When you take the first step towards building something? When you have a tech product to go along with it?
I think that anyone who has taken any risk to create is an entrepreneur. As my wonderful professor Greg Bunch had us declare together at the beginning of every class: “THE WORLD NEEDS MORE ENTREPRENEURS.” Who am I to gatekeep?
Today I learned that the “Air” in Airbnb described the air mattresses that they used while doing it manually. The founders blew up some air mattresses in their SF apartment and called it “Airbed and Breakfast,” or airbedandbreakfast.com.
Excellent read!