What the Uber-Lyft war teaches us about success and failure in the on-demand economy

This article is based on the book Platform Scale, written by Sangeet Paul Choudary. Platform Scale is available for free download for a limited period between October 5th and October 9th. To access additional bonus content, check the book website here.

The on-demand economy is bringing together technology and freelance workers, to deliver us services in exciting new ways. We are increasingly using our cell phones as a remote control for the real world.

Every week, we see a new platform come up that connects consumers with freelance labor. New companies are forming in almost every service vertical. But not all these companies clearly understand what determines success and failure of on-demand business models.

Success in the On-demand economy

There are two critical factors that will determine the success of a company in the on-demand economy: multihoming costs and interaction failure.


In computer networking parlance, mutihoming refers to a computer or device connected to more than one computer network. In the world of platforms, this notion is an important one. If your producers and/or consumers can co-exist on multiple platforms, you face a constant competitive threat. Eventually, it may be difficult for clear winners to easily emerge.

Multihoming costs are relatively high for developers to co-develop for both Android and iOS. Multihoming costs are high for consumers also because of the cost of mobile phones. Most consumers will own only one. However, multihoming costs for drivers to co-exist on Uber and Lyft are relatively low. Most drivers participate on both platforms. Given the ease of booking rides, multi-homing costs are very low on the consumer side as well.

In a previous article on TechCrunch, I had elaborated in detail how multi-homing could be prevented by creating long-term stored value within the platform.

For on-demand platforms, this is important because multihoming allows producers (service providers) to co-exist on multiple platforms. With a limited supply of service providers available, this can lead to interaction failure.


Interaction failure happens when the producer or consumer (or both) participate(s) in an interaction, without the interaction reaching its logical, desired conclusion. Imagine a merchant setting up a listing on Ebay that never gets any traction, or a video enthusiast uploading a video on YouTube that fails to get a minimum number of views. Quite often, these outcomes could be the result of poor quality listings or videos, but they could also be owing to the platform’s inability to find the right matches. Producers and consumers who experience interaction failure get discouraged from participating further and abandon the platform.

If reverse network effects set in, this can eventually lead to an implosion of a platform. In the initial days, interaction failure regularly leads to the chicken and egg problem. To understand these phenomena better and how they could impact your platform, I’d recommend this post on reverse network effects and this one on the chicken and egg problem.

Bringing it together – The Uber-Lyft war

Interaction failure is especially important for on-demand platforms. Imagine a consumer requesting for a service and the service never arriving. Imagine, in turn, a producer receiving a request, preparing to fulfill that request, only to find that the request gets cancelled. In both cases, the consumer or the producer may decide to abandon the platform.

This is exactly what Uber had in mind when it waged its war on Lyft. Unethical as that was, I’d like to focus on that to glean lessons for building the next Uber for X.

In some of the largest cities we see drivers drive for both Uber and Lyft, and other competitors. It’s not uncommon for these drivers to switch between the two platforms multiple times a day. With a limited supply of drivers in a city and the cost for a driver to connect to an additional platform so small, we see drivers multihoming on both Uber and Lyft. This has naturally led to intense competition between the two companies and Uber infamously resorted to a playbook to create interaction failure on Lyft using questionable tactics.

Uber decided to target interaction failure on Lyft, by contracting third party employees to use disposable phones to hail Lyft taxies. Before the Lyft taxi arrived at its pickup location, the Uber contracted employee would cancel the ride. With so many cancelations on the Lyft platform, drivers would become frustrated driving for Lyft and, in some cases, switch over to Uber. Lower drivers would lead to further frustration for consumers as they would have to wait longer for their requests for cabs to be fulfilled, eventually spurring them to abandon the platform. This loop is illustrated below.

When multihoming costs are low, producers and consumers will connect to many platforms. With multiple platforms sharing the same producers and consumers, it is difficult for a business to build defensible networks. Thus, it is difficult for a clear winner to emerge in the market. With many platforms operating and defensibility low, interaction failure becomes a key factor in determining long-term winners.

What does this mean for you?

SangeetpaulblogIf you’re building the Uber for X, you need to ensure that you’re tracking a metric that helps you determine the degree of interaction failure on your platform. Freelancers that don’t get business within X days, requests that don’t get satisfied within Y minutes, may all be indicative of interaction failure. The exact measure of interaction failure will vary by platform and the importance of tracking interaction failure will, in turn, depend on the multihoming costs.

About the author: Sangeet Paul Choudary is the author of the book Platform Scale, available for free download for a limited period between October 5th and October 9th. He also writes the blog Platform Thinking.

The Three Design Elements for Designing Platforms

What does the traditional world of manufacturing have in common with the way networked platforms work? How can a basic understanding of factory design help us change the way we think about designing internet platforms, marketplaces and social networks?

I’ve written previously about the distinction between pipes and platforms. If you haven’t already read it, you must definitely check it out. It lays out many of the basic principles that underlie the strategies I discuss on this blog.

The fundamental shift, I believe, the world has seen, is the move from linearity to networks. Putting 3D printers and other recent maker movement trends aside, factory-based manufacturing has traditionally remained the same. And that’s what I refer to as Pipes, as captured by the graphic below:

Pipes are characterized by the firm as the producer and a linear flow of value.

In contrast, Platform Thinking allows for users as co-producers. Value doesn’t flow linearly from a firm to the consumer. Producers and consumers are connected with each other over a network. The platform’s role, unlike a Pipe’s, isn’t production. The platform connects producers and consumers over a network and provides them the tools to interact with each other. It then uses data to match them with each other.

Business Design for Fifth Grades

Let’s look at business design at a very high level. At a fairly abstract, stratospheric level, the goal of business is the creation of value and the capture of some part of it to generate a profit.

If we get back to Pipe Thinking, and take the example of manufacturing, value is created in a factory or on an assembly-line.

The factory’s goal is value creation. 

The factory does this by setting up a set of value creating actions that add value on to the product.

There is an end-to-end process that is responsible for value creation.

But here’s the most important, yet obvious, bit. There is a unit of production and consumption which moves through these processes AND gathers the value that is added to it by the factory. Finally, in the hands of the consumer, this unit delivers value to the consumer.

We call it the product, the object that is manufactured in a plant. It’s the car running through Toyota’s assembly line or toothpaste tube getting filled out at Colgate-Palmolive.

This object is the basic unit of production and consumption in the industrial world. 

So if we think of it, the design of a manufacturing plant goes about in the following manner:

  1. Start with the object you’re creating; the product.
  2. Lay out the steps required to create it (value-creating actions)
  3. Design the process that encapsulates this flow of action
  4. Design the factory (and organization) that can execute this process.

The factory design enables the process.

The process design enables value creation.

But all of this starts by first looking at the unit that is being created, determining what value needs to be created on it and designing the entire business in a way to facilitate that value creation.

It doesn’t work the other way round. You don’t start with building a factory and an organization and then deciding how you structure the process. Any change in the organization or factory infrastructure is started by a change in the process which is started by a change in the need for value creation (often because of customer feedback owing to which you tweak the product).


So all this sounds great, you say! You just told us a bunch of obvious things using some fancy words. Where are we going with this?

I believe the fundamentals of business never change whether you’re in an agrarian economy, an industrial economy or an information economy. Let’s use what we already know from the above and figure how this applies to platforms.

Business design, as we just noted, doesn’t start with the factory, it starts with the unit that is produced. Unfortunately, in the design of a lot of networked platforms, we see the design process start from the factory i.e. the website or the app. That, instead, should be the last step in the design phase.

Let’s take a step back and get back to the Pipe:

Now, let’s watch it change when we talk about Platforms:

There are two key changes that happen as we move from pipes to platforms.

  1. The value creation shifts from inside the Pipe to outside the Platform.
  2. The producer role shifts from inside the Pipe to outside the Platform


These are the two fundamental shifts captured in all the talk about Open Innovation. And these are the two fundamental shifts we need to be aware of while designing platforms.

(Note: The platform can be the producer in many cases but most platforms will allow for external production in some way or the other.)

So let’s go about the job of designing platforms.



As in the case of Pipes, let’s start with the unit that is produced or consumed. This is the most interesting part. When we think of platforms or any form of internet businesses, we rarely think of what is being produced or consumed, we think of it in terms of a website or an app, or some other physical/visual manifestation. In actuality, since we’re talking about information businesses, the unit being produced and/or consumed should be content/information.

Seed and Parties, no Platform

What is YouTube? A website? An app? A platform for the hosting (production) and viewing (consumption) of videos?

Kickstarter? A platform for the hosting (production) and backing (consumption) of projects!

Twitter? A platform for the creation (production) and consumption of tweets!

Etsy? A platform for selling (production) and buying (consumption) actual physical products leveraging information about them (listings)!

Uber? A platform for booking a car leveraging information (car availability) to match producers (taxi drivers) with consumers (taxi seekers)!

Irrespective of the actual exchange being physical or digital, the unit that powers the matchmaking of producer with consumers (on the platform) is an information unit.

Start with the unit! If you’re designing the Twitter for X, look at what the Tweet for X looks like.

If there is one thing that’s central to Platform Thinking, it’s this. Large platforms look clunky, you don’t know where to start. Start with the unit that is being produced or consumed – I like to call this the Seed – build out from there. More on this in one of my subsequent posts.

Start with the Seed!


Great! So we started with the unit. What’s next?

In the case of Pipes, one moves from the unit to the process that adds value to the unit all through.

There’s one interesting difference between Pipes and Platforms though. Consumers don’t typically add value in real-time to this unit. They just consume. On platforms, consumers may add value as well. A creator may take a picture on Flickr but consumers may tag it. A creator may upload a video on Youtube but consumer votes determine how often it gets consumed. There are various ways in which consumers add value.

So what’s the counterpart of the process in Pipes? A set of actions involved in the creation and consumption of value?

I like to think of this as the Interaction. Every Platform has at least one Interaction.

Seed, Parties, Interaction

Creator uploads video, Consumer watches it, votes upon it. This is the primary interaction of YouTube.

Creator starts a project, Consumer consumes it, backs it etc. The primary interaction on Kickstarter.

The interaction is essentially a set of actions required for the creation and consumption of Seeds on the Platform. (This is an incomplete view but I’ll get further into it in a subsequent post, to avoid detracting from the main point.)

Note that a platform may have multiple seeds and multiple interactions but there will specifically be one that is core to the value proposition of the platform. YouTube is not a place where you go to create and consume comments, it’s a place to create and consume videos. In most cases, the primary seed and interaction are fairly obvious. In subsequent posts, we’ll look at a few cases where these are not.


Finally, we come to the platform. Once you know the seed and design the interaction as a set of actions, platform design is a lot simpler.

You’re just creating a network and an infrastructure that enables this interaction. 

Seed, Parties, Interaction, Platform

Well, there’s a lot more depth there but for the purposes of this article (remember, we addressed this to fifth graders somewhere up there), the key point here is that platform design should start with the seed, flesh out the interaction and then design the platform as a consequence. Not the other way round.

Even with platform design, one must distinguish between the system and the interfaces. YouTube is a complex system but has many different interfaces/functionalities for creators, viewers, brands, advertisers, media houses etc. on different channels like web and mobile.

The design of the interfaces should be true to the design of the system. And that is achieved when one starts by focusing on the seed and the interactions it enables.

I’m going to be digging into this in detail over the next 2-3 months to lay out a detailed framework for designing and running networked platforms. I’d love to hear your thoughts first up, whether you agree or disagree.


Let’s quickly recap the three key principles of platform design:

1) Platform design should start with defining the value that is created or consumed, the Seed.

2) The Seed should lead to the actions that enable the creation and consumption of that value.

3) Only in the last step should one go about designing the system and interfaces that enable those actions.

There is a fourth key element missing here, the role of data. I will be covering that in future posts. For this post, in particular, I wanted to focus on the contrast between pipes and platforms. The role that data plays on a platform is very unique to a networked world.

It’s been a late start to the year but I’ve spent the last three weeks bringing a lot of my thinking on platforms together in an effort to start creating a structure around it. I’m getting that stared with this post.

Everything old is new again! Hopefully, the magic lies in using the old to interpret the new!

This article was first featured on Sangeet’s blog, Platform Thinking (http://platformed.info). Platform Thinking has been ranked among the top blogs for startups, globally, by the Harvard Business School Centre for Entrepreneurship

The One Feature That Changed Social Networking Forever

What is the single most important innovation that Facebook ever came up with?

Before I answer that, let’s think of the real value for users on a social network. Social networks are a classic example of the platform business model where users create all the value and there is very little value until users come on board. Real value for every individual user, then, is the value that his network is capable of creating.

Most communication and networking products have never truly succeeded in capturing this value on an ongoing basis. E.g. I might be connected to a lot of users on a communication product but I need to be actively engaged in a conversation to benefit from these connections.

This is why the News Feed is Facebook’s most important innovation. It allows users to constantly benefit from the surge of activity in their network neighborhood. It’s a stalker’s delight, a lurker’s guilty pleasure. But the News Feed is the single most important innovation that changed social networking from a user-centric to a network-centric activity.

It changes the use case for an entire product category. Post the News Feed, social networking was no longer about staying connected with a friend or even with a group. For the first time, social networking was about staying connected with one’s entire network.

The move from transaction to engagement

The News Feed represents a leap in the evolution of online communication and networking products. Online communication and networking have traditionally been transactional in nature. Email has always been a transactional product. Chat is transactional as well. We use these products only when we want to engage in an actual exchange (of information) with someone else.

Early social networks were built along the same lines. Imagine the days of Orkut, Bebo and, even, MySpace. Social networking, back then, was an extension of the existing communication models around email and chat. You typically logged on to connect with friends. If you didn’t want to connect with friends, you just never logged on. The dominant use case on these social networks was transactional.

The News Feed changed that! It moved social networking from a transactional use case to an engagement-driven use case.

Engagement products need to provide a minimum guarantee of activity to keep the user engaged. Transaction products, in contrast, need to ensure liquidity and the assurance that the user can complete a transaction conveniently.

If you think of Facebook pre-News-Feed, users used the platform largely to communicate with others. The News Feed shifted value in the platform from mere connections (and communication) to content (and engagement).

First among equals? 

By no means is the news feed the only determiner of engagement on Facebook. The decision to allow developers to build out an app economy on top of Facebook and the creation of social use cases on top of Facebook (most notably gaming and gifting), clearly helped the engagement cause. However, across all initiatives that Facebook ever took, the one that has been most persistent and that eventually took over as the default home screen – the first ‘feature’ that a user is exposed to on every log in – is the news feed.

The representation of the network effect

In traditional social networking, the feature that the user kept returning to was his own profile, with some notifications alerting specific network activity. This is why having the News Feed as the default Home Page is rather important. It changes social networking from a user-centric to a network-centric activity.

The News Feed embodies the very concept of the network effect. It shows that the network effect isn’t simply a function of the number of other nodes you are connected to but also of the nature of the links that connect you with them. A user’s past interaction with other nodes is a great determiner of the strength of ties between nodes. A real world network would have certain ties stronger than others. The news feed captures this and creates a user-centric view of the network.

This is also why I believe Facebook deserves credit for pioneering the news feed. An activity stream or news feed like feature was already present in Twitter, and before that, on Flickr. But these never gave an accurate representation of the network and were at best, mere activity streams aggregating the activity at neighboring nodes. There was no focus on the nature of the link with neighboring nodes. This is where Facebook’s focus on optimizing the news feed algorithm creates a more accurate representation of one’s network than ever before.

A stronger network effect? 

One might argue that the news feed also creates a stronger network effect. With traditional social networks, you could have a few hundred friends but it was arguably the same 10 friends bringing you back to the platform. This meant that losing those 10 friends to another platform could signal the need for you to make the move as well. With the news feed, a stream from a much wider circle of friends constantly hits you. When the central use case shifts from communication with individual friends to interaction with the overall network (via the news feed), it could potentialy make the network more resilient to a situation with reverse network effects.

Beyond social networking

Moreover, this doesn’t apply only to social networks. Any business model which relies on user-generated content can benefit from a well-architected news feed. Even marketplaces, which have traditionally been transactional, are creating engagement with a news feed.

Design principles

The key design consideration is relevance. A news feed should help with personalized discovery. This introduces another tension. Relevance and personalization often tend to reinforce things that we are already interested in. A personalized feed should factor in some form of serendipity to ensure that users do not get increasingly served only those objects that reinforce their preferences. (Designing for serendipity is far from trivial and merits a post in itself, sometime soon.)

If you’re building a networked product, think of what embodiment of the network can be delivered to the user. The News Feed is brilliant because it takes a user’s network and individual actions and builds out something that results when the two are superimposed on each other. This is exactly how our social experience works in the real world. Our world is shaped by a consequence of the actions that we take with our environment. None of that is, of course, simple enough to be replicated through a mathematical model. But the news feed is a great approximation.

Tweetable Takeaways 

How social networking moved from a transaction-first to an engagement-first model.Tweet This.

How to build products that deliver value from network effects. Tweet This.

The one feature that transformed social networking forever. Tweet This.

Design principles for building products that capture network effects. Tweet This.

This article was first featured on Sangeet’s blog, Platform Thinking (http://platformed.info). Platform Thinking has been ranked among the top blogs for startups, globally, by the Harvard Business School Centre for Entrepreneurship