Broken is beautiful.

An ode to wonderfully imperfect products.

Alex Taussig
Lightspeed Venture Partners
7 min readFeb 11, 2019


Recency bias is powerful, especially in technology. Our favorite consumer products are elegant and sleek — the results of years of iteration, testing, and user feedback. It is easy to forget the path a product took to reach its present state of perfection. The starting point is, too often, mere dust in the wind.

In the beginning, most consumer products were broken in some meaningful way. An early product doesn’t need to be perfect to be valuable. Succeeding in spite of clear product flaws demonstrates a visceral user need.

As an investor, this lesson took me years to internalize. I have since grown a taste for broken products that are clearly “working” nonetheless. Users will put up with a ton if aggravation if a product truly delivers value. Ignore the rough edges and focus on users, and you might find a gem.

History is full of examples of this phenomenon. I’ll list a few below for delight and nostalgia.

Uber screenshot (2011).

I moved to San Francisco from Boston in 2012. At the time, Uber was a black car service. You could use an app, or more often SMS, to hail a relatively expensive driver to show up in a 5–10 minute window somewhere in the city.

The context here is important. San Francisco has notoriously bad public transit. Taxi services are unreliable at best. Because street parking is a nightmare, only those with garage access usually own a car. Adjacent neighborhoods are often partitioned by giant hills that make walking even short distances unpalatable, despite the city’s high population density.

For all these reasons, Uber’s value proposition was incredibly strong for its hometown users. Yet, the UX was frequently miserable — smelly cars, bad drivers, frequent delays or no shows, “surge” pricing, and annoying phone calls to clarify pickup location. Still, Uber was worth putting up with. Unless you owned a car, Uber was literally the only way to get home from some neighborhoods, especially at night. It subsequently fixed many of these issues with scale and driver feedback, especially as markets matured.

Stitch Fix* screenshot (2011).

In the beginning, Stitch Fix barely had a website. Instead, the company had a set of forms that requested your email, measurements, and fashion preferences. Here’s a description of the makeshift operation CEO/founder Katrina Lake started with:

“A year after Lake started Stitch Fix in her apartment, the company had a small office in San Francisco where every Monday all employees helped package and ship out “fixes” — the boxes of clothing sent to clients. As someone who never liked taking risks, Lake concedes she made some risky bets during this time, such as trusting that clients would pay for the clothing they kept. (This was before Stitch Fix had a system that took people’s credit card numbers. At the time, customers were sending Stitch Fix money through PayPal or writing their credit card numbers on a piece of paper and sending it back with the clothing they didn’t want.)”

I recall my wife’s experience with Stitch Fix in those early days. She paid $25 for a fix twice, each time keeping zero items. She loved the idea of an affordable personal shopper, but each time the clothing she received was the wrong style or size.

Weeks later, I saw a third Stitch Fix box appear in my doorway. I was puzzled why she would risk yet another $25 on a service that failed her twice before. Clearly, she was so invested in the idea of Stitch Fix working for her that she was willing to try it over and over again, until the service performed.

Eventually, it did. In retrospect, the company was initially limited by both inventory and data. That much was unavoidable. If a customer’s preferences either (1) didn’t match current inventory or (2) generated an incorrect fix, she had a bad experience. The fact that people like my wife were willing to put up with the service over and over again, until the company had sufficient quantities of inventory and data, demonstrated that the company was solving a real problem for consumers. Both issues were eventually solvable with enough scale and operational excellence.


Twitter fail whale, circa 2009.

The original “microblogging” service, Twitter took off at SXSW in 2007. The service tripled its volume to 60,000 tweets per day during the event. Growth was rapid from there. Tweets per day grew from only ~4,000 in 2007 to ~1 million in 2008 to ~50 million by the beginning of 2010.

All this growth melted Twitter’s fragile infrastructure. Sometime around 2008, either Nick Quaranto or Jen Simmons coined the term “fail whale” to describe the image of a nonchalant whale held up by some hardworking birds. It appeared when Twitter’s core service was down. The original image was drawn by Yiying Lu.

The fail whale was an endearing way to channel the frustration of users who desperately wanted to connect with others on the service. In a way, it downplayed the real technical issues the team was facing and set expectations with users that “Hey, we’re trying our best over here.” The company grew past the fail whale phase a few years later, but those early days remind us how much users loved the service, despite its technical flaws.

HQ Trivia*

Scott Rogowsky, a.k.a. lag daddy.

If Twitter had a fail whale, HQ Trivia had “lag daddy” — aka pixelated host Scott Rogowsky.

HQ launched to the public in October 2017 and ramped to over 1 million peak concurrent users within 3 months. It doubled that within another 3 months. Put into perspective, only two online games sold through Steam have ever surpassed 1 million peak concurrent users: PlayerUnknown’s Battlegrounds (PUBG) and Dota 2. Fortnite by Epic Games* is not on Steam, but revealed more than 8 million peak concurrent players in late 2018. So, only 6 months after launch, HQ achieved a peak concurrent scale comparable to a top 5 online game today!

Unlike a text-based service like Twitter, HQ had to deliver live mobile video to all those players. Lag daddy was the result of the company’s nascent infrastructure breaking down under unimagined strain. Like the fail whale, lag daddy became a meme, often promoted by Scott himself:

HQ succeeded in spite of these disruptions. I like to think that it was user love, plain and simple, that kept people coming back each night to give HQ a second chance while it worked out the kinks.


Android 1.0 running on a T-Mobile G1 phone. (Source)

When Apple launched the iPhone in 2007, Google knew it needed a response. At the time, industry incumbents derided Apple’s “keyboard-less” approach to mobile. Remember this gem from then Microsoft CEO Steve Ballmer?

“[The iPhone is the] most expensive phone in the world, and it doesn’t appeal to business users because it doesn’t have a keyboard. We have the greatest phones in the market today.” 🤦‍♂️

According to this oral history of Android, Google’s engineers were secretly enamored of the iPhone, but couldn’t resist keeping that physical keyboard. When Google partnered with T-Mobile in 2008 to create the G1 phone, which ran Android 1.0, it innovated around multi-tasking, pull down notifications, and copy/paste functionality (all things iPhone lacked at the time). Despite some cool new features, the first Android phone was clunky. The OS still looked and felt like it came from a past generation.

Android (2008) vs. iOS (2007). Which one feels more like the future?

Because its underlying software was open source, however, Android improved at a rapid pace. It only took Android another year to release Cupcake (v1.5), which was the first version to support an on-screen keyboard. Donut (v1.6) allowed Android to be ported to a wider variety of hardware devices. At this point, there was no Play Store. All the individual apps like Gmail were updated alongside the firmware refresh, so downloading new versions of the OS was extra important.

I recall early Android users putting up with its clear disadvantages because they believed in Google’s relatively open approach compared to Apple’s. They had faith that Google would continue to improve the software and that the open ecosystem would offer more of the internet’s benefits in the long term. It also probably helped that Android-enabled smartphones were significantly cheaper at a time when only 6% of adult Americans owned one.

My takeaway

Evaluating consumer products at an early stage requires a high tolerance for flaws. The price of success is greater usage than most companies can adequately support in the short term. If the product offers enough value to a user, she will put up with almost anything for the privilege of using it. The broken customer journey of early users is a badge of pride for companies that have shown true breakaway growth and product/market fit.

That’s why, as an investor, I try to focus on the heart of what’s working. Nearly everything else can be fixed with enough time and resources.

If you have a beautifully broken product, send it my way! For more of my thoughts, consider subscribing to my weekly newsletter:

* Designates a Lightspeed portfolio company.

Enjoyed this post?

Give it a clap or two so that others can find it!

Image result for medium clap gif



Alex Taussig
Lightspeed Venture Partners

Partner @ Lightspeed. Current: All Day Kitchens, Archive, Daily Harvest, Faire, Found, Frubana, Muni, Outschool, Zola. Past: $TDUP, $TWOU. Writes