Lecture 8: How to Start a Startup

You make some good general points in your comment, but since you're holding up DoorDash as an example of success, I thought I'd point out that DoorDash is just one of a number of "on demand" companies that are currently en vogue.

In many cases, their ability to scale is heavily dependent on their ability to shift major costs (taxes, vehicle operation, commercial auto insurance, etc.) onto their workers, who they classify as independent contractors. The big question is whether the workers are actually independent contractors. This is a topic of discussion in legal circles[1] and the trend does not look favorable for many of these companies[2]. Uber is already being sued for misclassification[3] and the same attorney says she's looking at Homejoy[4].

One adverse outcome in a legal action could change the economics of these businesses significantly, and potentially even put some of them out of business.

[1] http://www.pepperlaw.com/publications_article.aspx?ArticleKe...

[2] http://www.msnbc.com/msnbc/fedex-loses-court-case-against-dr...

[3] http://uberlawsuit.com/

[4] http://www.washingtonpost.com/news/storyline/wp/2014/09/10/a...

well, probably time comes for a "collective bargaining on demand" startup and an app for it

Okay, I heard that point in the lecture and read it again in your post.

Okay, now for me -- I'm slow this evening, and I still just don'e get it on the crucial role of a mobile pbone. A phone, sure. But what's the crucial role of the mobile part?

A guess: For a given delivery job, the candidate drivers typically are not close to a land line or desktop phone. Instead they are out and about typically in the car they will use for the delivery. So, the fact that their phone is mobile, that is, with them always, is what is crucial. A guess.

A second guess: The phone should be not just mobile but also smart so that it has GPS to say where the heck the driver is currently. Then, with an app on the smart phone, could let the central office or dispatching system pick the driver closest to the location of the pickup. But this would meant that they have an app, and I thought that maybe so far they didn't.

Gee, I had heard that it was standard for taxi services to be willing to pickup from a restaurant and deliver the food.

And I'm confused: Where I've lived in the US, it has long been standard for pizza shops and Chinese restaurants to be willing to deliver. So, sounds like the problem of delivering food for such restaurants had long since been solved.

Apparently there's going to be more of this stuff of a customer pushing some buttons and have some product delivered within, say, an hour. So, apparently one goal of Amazon is to have warehouses in major cities so that there can be same day delivery of orders. Okay, then, Doordash could be looking to do those deliveries?

There should be some economies of scale here: That is, for five pizzas, have two ways to proceed. First way, just pick up the pizzas one at a time and deliver them. The second way is to pick up all five pizzas and deliver them using a solution to the traveling salesman problem with six stops, the pizza shop and the five customers. So, of the pizza shop is busy, get to pick up several orders there within, say, 5 minutes. Or, maybe pick up from several restaurants in the town center and deliver to the surrounding houses and offices -- one Doordash with nearly all the restaurants in the town center would have an economy of scale.

Doordash got such fast response from their simple landing page that maybe a lot of taxi drivers, some of which are just sole proprietorships anyway, and, sure, now likely making heavy use of cell phones, that they could just offer a service like Doordash for any pairs of customers and restaurants that want a delivery. I'm failing to see why the taxi drivers in Peoria need Doordash to follow the same idea.

Then use a good editor to look at the HTML and see how to parse the thing and extract the links.

Write some code to do this and try on some more of the URLs. When see are not parsing the files just right, tweak the code.

Finally, get some okay code and get and parse all the HTML files from all the URLs with the code -- scalable!

So, start by hand and see how to automate the work and then have a scalable solution.

I just did that for a collection of about 5000 URLs!