Quibi 2020: Why a $1.75 Billion App Launched Into an Empty Commute

The room

The psychology: planning fallacy

In a 1977 technical report, Daniel Kahneman and Amos Tversky separated two kinds of evidence a forecaster can use. Singular information is about the case at hand: this team, this product, this schedule. Distributional information is about how cases like it have turned out. Intuition favours the first. Planners build a scenario of how their project will unfold, step by step, and the scenario rarely includes the delays, rivals and accidents that, taken together, almost always arrive. Any single setback is unlikely; the chance that at least one occurs is high. Kahneman and Tversky called the resulting optimism the planning fallacy and traced it to the neglect of base rates rather than only to wishful thinking.

Quibi’s plan was an unusually rich singular scenario: famous creators, a screen that rotates between two versions of every scene, sold-out advertising and a clear picture of a commuter with ten spare minutes. Distributional evidence was also on the table. go90 had closed after three years. An industry analyst said there was no proven case of premium short-form video as a paid platform. The company’s own launch had already slipped from a planned 2019 date to April 2020. The pandemic was a specific shock that no plan predicted, but the planning-fallacy question is different: how much margin did the plan leave for something, anything, going badly in year one?

The fallacy also shapes how results are read afterwards. Roger Buehler, Dale Griffin and Michael Ross found that people explain their own past overruns with causes that are external, temporary and specific, which makes those overruns feel irrelevant to the next forecast. That pattern does not mean an external cause is wrong. The pandemic was real and it hit Quibi’s use case directly, and the founders’ own letter weighed the idea alongside the timing. The research point is narrower: a plan that only works if nothing major goes wrong is fragile, whatever the shock turns out to be.

Why smart people fall for it

Three forces made a carefully built plan more fragile than it looked.

1. 1

   The vivid scenario beats the base rate

   The plan was a detailed story of stars, Turnstyle and commuters. The record of earlier mobile video ventures carried far less weight.

2. 2

   Plans assume nothing big goes wrong

   Content budgets, ad deals and a 7 million subscriber target were set for a year without major shocks, leaving little margin when one came.

3. 3

   Setbacks get filed as one-offs

   A disruption explained as unique and external feels irrelevant to the next forecast, even when it exposes how thin the margin was.

Was Quibi’s failure really about timing?

The case for timing is strong. Quibi was designed for moments outside the home, and it launched just as stay-at-home orders took those moments away. Katzenberg pointed out that Netflix and Disney+ gained viewers in the same months, because people at home had televisions and time, the two things Quibi’s mobile-only design did not use. A product built for the gaps in a busy day met a period in which, for millions of people, the day had no gaps.

The case against timing is also strong, and some of it comes from the founders. Their letter noted that other businesses had faced the same conditions and found a way through. The launch itself was not a failure of attention: 300,000 people downloaded the app on the first day. The problem came later, when free users had to decide whether to pay, and only about half a million did against a plan of more than 7 million. Users complained about choices made long before the virus, such as no TV viewing and no clip sharing. Katzenberg himself said the company asked people to pay before they understood the product, and before launch an analyst had already questioned whether premium short video could ever sustain a paid service.

The honest answer is that the two causes cannot be cleanly separated. Katzenberg said nobody may ever know how much each contributed. A reasonable reading is that the pandemic removed the one scenario the plan depended on, while the weakness of the idea meant there was no second audience to fall back on. A product with a broader appeal might have bent; Quibi’s design left it very little room to bend.

That is where the planning fallacy matters. The useful question is not whether anyone in 2018 should have predicted a pandemic. It is whether a plan that committed more than $1 billion to a single year of programming, priced for a single habit, had any buffer for a bad year. The record suggests it did not.

What does the research say about big launch plans?

Kahneman and Tversky’s 1977 report was written as practical advice for analysts. It proposed a five-step correction: pick a reference class of similar cases, estimate how outcomes in that class are distributed, make your own intuitive forecast, judge how predictable the case really is, and then pull your forecast toward the class average in proportion to that uncertainty. The authors were candid that the method had not yet been tested systematically and offered it as suggested practice.

Buehler, Griffin and Ross tested the underlying bias in five studies. Students writing honours theses predicted about 34 days and took about 55. In a study of everyday tasks, students predicted academic projects would take 5.8 days and they took 10.7; most of those with external deadlines met them, but only 38.7 percent finished by their own predicted date. When participants thought aloud while forecasting, about three quarters of their thoughts were future scenarios and only 7 percent drew on past experience.

Two findings point to remedies. Simply recalling past projects did not help. But participants who were asked to connect their past experiences directly to the current task made predictions that matched their actual completion time, and 60 percent finished on schedule, compared with 29.3 percent in the control group. Observers predicting someone else’s task were more conservative than the people doing it, forecasting 8.5 days where the actors forecast 5.5, and observers referred far more often to the actor’s past.

Translated to a launch like Quibi’s, the research suggests that the most useful voices are often the least invested ones. The analysts who cited go90 and the absence of a proven paid short-form model were acting like Buehler’s observers. The people closest to the plan, rich in singular information about stars, technology and sold advertising, were in the position the studies identify as most prone to optimism.

How would a modern team decide?

A modern team would start before the March decision, at the plan itself. It would list launches that resemble this one, such as new subscription video services and earlier mobile video ventures, and ask how many reached their first-year subscriber targets. That record becomes the starting point, and the team’s own forecast adjusts from it rather than the other way round.

It would then write down the assumptions the plan cannot survive without. For Quibi, one of the biggest was that people would spend time away from home with a phone in hand. Naming that assumption makes it possible to ask, in advance, what the company does if it fails: add TV viewing, broaden the format, or slow spending. When the assumption broke in March 2020, a team with that answer prepared would have been choosing from a list rather than improvising.

Third, it would stage the big commitments. Content, marketing and hiring would be released in tranches tied to evidence, such as a paid-conversion rate after the free trial, rather than committed for a full year upfront. A free trial is useful, but it measures curiosity. The number that decides whether the next tranche is spent should be how many people pay.

Finally, it would invite outsiders to forecast. Buehler’s observers were more accurate because they looked at the record rather than the plan. An investor, a board member or an analyst with no stake in the story can play the same role, as long as their estimate is recorded and compared with the team’s before money is committed.

The manager's playbook

1. Build a reference class firstList similar launches and how they did in year one, then adjust your own forecast toward that record.
2. Write down the plan’s fatal assumptionsName the conditions the plan needs, such as people commuting, and decide in advance what you will do if one fails.
3. Ask an outsider to forecastPeople without a stake in the plan lean more on past cases; record their estimate and compare it with yours.
4. Separate curiosity from willingness to payTreat free sign-ups as interest, and set a paid-conversion target that decides the next round of spending.
5. Stage your costly commitmentsRelease content, marketing and hiring budgets in tranches tied to evidence, so a shock costs one stage rather than the whole plan.

Questions people ask