Uber: The feedback loop disrupting transportation

| September 1, 2014 | 3 Comments

When Uber was valued at tens of billions of dollars, there was significant criticism from valuation experts (Aswawth Damodaran, no less) and an equally compelling response from one of Uber’s earliest investors, Bill Gurley. One of the best things that emerged from these discussions (in conjunction with a back-of-the-napkin sketch tweeted by David Sacks, founder of Yammer) was a much more nuanced discussion of the feedback loop that drives Uber’s network effect.

Uber is a classic two-sided marketplace where more cars on the network attract more travelers and vice versa. But as with any network effect, it isn’t simply the number of cars and travelers that attract each other but the levels of participation of both sides. A higher participation from drivers is useful only if it results in higher availability, and consequently, lower waiting time for passengers. Similarly, a higher participation from passengers is useful to drivers only when it means lower down time and, potentially, the ability to charge higher prices, thanks to Uber’s much-maligned surge pricing. (Surge pricing increases rates as demand overtakes supply)

Uber’s feedback loop works in the following way:


More drivers equals shorter pickup times

More drivers available at any given point in time result in shorter pickup times as the probability of matching a request to a drive in the vicinity is that much higher. Shorter pick-up times, in turn, lead to greater and wider usage.


More usage leads to higher coverage

As more drivers fuel more usage, more usage, in turn, brings in more drivers on the road. Uber’s network is a city-level network and usually starts from the centre of the city. Over time, as this feedback loop picks up, demand starts sparking up in the fringes and drivers start getting onto the platform and serving the fringes as well. Hence, greater usage increases saturation within the city, since the city has a finite limit. With greater saturation, the pickup times further fall, thereby attracting more demand, leading to a positive feedback.


Greater liquidity leads to better prices

As more demand and supply flood in, the waiting time for drivers falls. With lower waiting times as well as higher availability of drivers, the platform can offer better prices to travelers. This, in turn, brings in more travelers into the system and the virtuous feedback loop gets strengthened further.


Uber has a fairly nuanced feedback loop. As with the case of the data-driven feedback loop in peer lending, and the curation-driven feedback loop for many market-making platforms, it pays to understand the factors that contribute to a positive feedback loop. Paying close attention to these factors and architecting the conditions that encourage these factors strengthens the network effect over time.

Network effects don’t simply happen. They are most often the result of carefully building and fueling a feedback loop.

Image Credits: David Sacks

About the Author:

  • Charis Rafailidis

    Thanks Sangeet.

    I partially remember some of your past posts. Will read again a bit.

    What you are saying is that in a platform the initial tests have more to do with the core interaction the platform enables its users to do instead of the products or services being produced or consumed, right?

    And for the platform to try and diminish all the potential scenarios of failure of those interactions, because a failed interaction means no value created in the platform, right?

  • http://platformed.info/ Sangeet Paul Choudary

    Yes, but there are different ways of testing a platform. If you refer some of my posts from end 2012, I lay out how platforms can be tested in their very early days. The point is not to test features and functionalities as that’s what I see in most of the suggestions above. It is to test scenarios that lead to failure of the core interaction on the platform. I’ll write more on this in the days ahead.

  • Charis Rafailidis

    Another great one Sangeet, as usual.

    One question on platforms, would love your view on this:

    A company that works as a pipe, like let’s say a new company that sells a kind of special shoes through the internet

    has huge differences from a platform that for example is a marketplace of different companies selling lots of different kinds of shoes and clothing, right?

    And please focus on this:

    On the early days, when those two companies are just building and want to start selling their products,

    Again there are huge differences right?

    I mean, a pipe company can even start pre-selling its new show, present some photos of its prototype, maybe even advertise on it and instantly take customer feedback, measure how the market responds before actually build or ship its product.

    But for a platform, this can’t be done, right?
    There’s no point for a platform to start selling its “product” – meaning the platform itsself, as without it being compete and without a critical mass actually using it, there is no value for anyone.

    The platform’s “initial tests” come at a stage later than the pipe, and that’s after the actual platform is complete and users start using it. Then the platform company can get feedback, measure and respond to thr market’s needs, by trying to get the producers to produce and the consumers to consume.

    Am I getting this right?
    Or is there a point you disagree or want to add/change something?