December 4, 2017

Driving away dollars: A case study on Uber and Travis Kalanick

The rise of Uber has been nothing short of meteoric. In financials first revealed to Bloomberg in April 2017, Uber is shown to have crossed $20B in annual gross revenue in 2016, up from less than $1B in 2013. Uber has leveraged this revenue growth and the ambitions of its bullish investors to expand into more than 600 cities worldwide. In 2017, Uber faced a number of high-profile PR crises, but the question of which crises – if any – have actually dragged on Uber’s business is harder to resolve. For Uber it’s a question of not just crisis management but also resource allocation and long-term strategy. Uber’s ability to respond intelligently and swiftly to its crises depends on its ability to understand which issues and influencers are actually hurting its revenues the most, and why.

While we confirmed that these crises did in fact hurt Uber’s sales, it is our ability to drill into which crises mattered most and why that empowers management to respond correctly. We found that the root cause of revenue loss to Uber can generally be tied to the behavior of former CEO Travis Kalanick, but it was union loyalists driving calls for boycott on social media who took advantage of this opening to cause harm to Uber in the first half of 2017. It is surprising how quickly and efficiently these labor unions were able to cause tangible harm to Uber: their initiatives drove an immediate impact measuring as much $8M a day in lost business.

We see a cumulative effect on Uber’s brand where Uber’s 2017 Q3 US sales show significant slowing down in growth; our models indicate that revenue may have even declined quarter-over-quarter for the first time ever. Those same models show key competitor Lyft growing significantly year-over-year, picking up share amongst key demographics and growing both the airport and carpool businesses significantly.    

Note that Uber is not a client of ours, but its business makes for an excellent case study in what we can do. Quantifind focuses on finding impactful signals even when customer feedback is noisy and it is difficult to understand what part of it is important. Quantifind’s signal processing engine helps find the 10-20% of conversation that actually does inform what actions the consumer will take to impact the bottom line. Our platform connects social, forum and other online consumer conversations with sales data. In this case we leveraged predictive models for Uber sales (proxy panel data) built from online conversation to measure if and how these 2017 crises have impacted Uber’s financials. 


The first instance of a scandal driving customers away from Uber is well-documented. In late January 2017, left-leaning political activists declared a boycott of Uber. The acute offense was that Uber didn’t halt service at JFK airport – and in fact may have eliminated surge pricing to capture greater sales volume – in solidarity with those protesting President Trump’s immigration ban (as the New York Taxi Workers Alliance had). The #DeleteUber movement took off on social media, peaking at nearly 25% of all organic Uber consumer conversation on Twitter on January 28. Within a few days, The New York Times reported that 200,000 people had deleted their Uber accounts.

As the weeks went by, the boycott conversation fluctuated as the various controversies flared, though never returning to the level seen on January 28. The figure below details the movement of this thematic conversation, as a fraction of all Uber-related conversation. The salient question was, did any of this conversation matter to Uber sales?

Data chart on Uber's scandals
‍Figure 1: "Boycott Uber" topic time series for the first half of 2017, reported as a percentage of all unique Uber mentions on Twitter (sample size = 1.13M).

We tested whether this “Boycott Uber” topic time series impacted sales, with a focus on the five months after the late January boycott. Our test confirmed a negative correlation between changes in the boycott conversation and changes in Uber’s daily sales with 98% confidence. No other conversation theme from the data showed comparable negative association.

As shown in Figure 1, this boycott conversation flared up in response to a number of specific revelations:

Notably absent from this short list are the announcement of the Waymo intellectual property lawsuit (February 23), the firing of executive Amit Singhal over undisclosed sexual harassment allegations (February 28), the news that Kalanick took his fellow executives to an escort bar in South Korea (March 24), and news that an executive in Asia hired a special investigator to go after his rape accuser in India (June 7). These four events may certainly have offended Uber’s customers, but had no demonstrable immediate impact on Uber’s sales. Not all of Uber’s crises in 2017 mattered to the bottom line, and some mattered more than others.

The Fowler blog post and the surfacing of the Kalanick video each drove a boycott conversation that immediately reduced Uber’s daily US sales by $8M. For both the Tim Cook/privacy revelation and the reports of surge pricing in the aftermath of the London Bridge terror attack, the corresponding boycott conversation reduced daily US sales by approximately $3M.

Over a longer time period, we measured a much slower-moving decline in Uber sales whose consistency is therefore more concerning. Uber growth clearly slowed down drastically in Q3. In terms of year-over-year quarterly growth, we estimate that Uber achieved just 20% growth, whereas a year earlier the same metric showed 70% growth.

A Familiar Foe

In order to help inform management on how to respond, we can leverage the overlap of social graph data with that language of consumers driving those impactful boycott conversations. Intuition might expect that those boycotters are a politically-active subset of Uber’s left-leaning urban base. Over the entire year, where the majority of the conversation happened around the very first #DeleteUber movement (January 28 and vicinity), this is largely true. However, when we examine people behind subsequent surges in boycott conversation the profile narrows: 3 of the top 5 indexing affinities for this group are American unions (Table 1 below). Absent from this list are any feminist or privacy groups which you might expect to be enraged by the specific revelations behind the calls to boycott. Similarly, we see that demographics of this impactful group surprisingly skew more white, more male, and older relative to Uber’s base.  

Table 1: Top indexed affinities for social media users driving the Boycott conversation, relative to users mentioning Uber overall (February-October 2017).

‍It is apparent from the data that these boycott movements on social media are being led not by outraged Uber customers, but rather by “influential” union loyalists.

Lyft Seizes the Opportunity

While Uber has been harmed by these recurring scandals and boycott movements, Lyft has flourished. Lyft seized an opportunity to contrast its values with Uber’s by making a $1M donation to the ACLU the same weekend that Uber faced its first boycott; this reportedly drove a 60% increase in Lyft app activations at the same time that 200,000 people quit Uber. But that was just the start. In June, a snapshot from credit card transactions fueled reports that Lyft is eating away at Uber’s market share dominance in the US in 2017. Our models shows the trend continuing: a 60% year-over-year growth in Q3 sales for Lyft, whereas Uber sales growth slowed to 20% year-over-year.

Our platform determined that Lyft grew by eating into Uber’s historical advantages in the South and Midwest regions and with the 35+ crowd while maintaining its strong position with both women and in the Pacific region. We can also use our models to understand why Lyft has grown, and where Uber is fundamentally threatened, beyond the impact of PR crises. Three topics of conversation across the aggregate of data sources correlated strongly with driving revenue: Airport, Carpool, and Tourism.

Lyft has made great strides in cutting into Uber’s dominance at airports, with consumers often citing parking challenges and parking costs as a reason to choose Lyft instead of driving to the airport. For the last quarter Lyft saw stronger pull than Uber at all three of the major California airports (SFO, LAX, and SAN) as well as McCarran International Airport in Las Vegas. Lyft Line, Lyft’s analog to Uber Pool, is up significantly and consumers are pleased with and happy to recommend Lyft’s services in and around the Disney resorts in Orlando. Their recently-launched Disney-branded Minnie Van service in particular are resonating.

Fickle Support?

While millions of conversations about Uber and Lyft portray Uber as the insensitive, politically incorrect Goliath, and Lyft as the progressive, and emerging underdog, our models extract a more subtle message from the buried signal in consumer conversations that actually predict sales. Lyft would be well advised to recognize that the outcry of public condemnation of Uber that hurt sales the most was driven largely by union loyalists who capitalized efficiently on the opportunity. As Lyft continues to grow in airports, carpool and tourism - and unions begin to feel threatened by them – Lyft may find that the same “influencers” that helped them gain share vs. Uber in 2017, may turn against them in 2018.