Uber’s Decline: A hiccup or sign of trouble | Product Case Study | Root Cause Analysis

Problem Statement - You are the PM of Uber. The Appstore/ Playstore rating dropped from 4.5 to 3.8 in the last 2 weeks. Find the root cause of the problem and come up with solutions to solve the problem.

Disclaimer: This case study represents hypothetical scenarios and solutions and is written for educational purposes only.

As tempting as it sounds to come up with solutions for the problem at hand, comprehending the problem by deconstructing it into mini-problems is where we’ll start.

Assumption— Since we’ve been asked to find the root cause of the drop in ratings, we assume the origin can be found by investigating Mobile App only.

A drop in ratings on Appstore/Playstore could be a result of soaring user discontent over the past two weeks due to factors which we are not sure about yet. To probe deeper, let’s start with an overview of the user journey on the Mobile App.

Uber: User Journey

The journey can be divided into the following activities :

1. Decision to Choose
2. Account Setup
3. Request Ride
4. Ride User
5. Arrival

User journey helps us understand various touchpoints, user goals and actions. Any issue originating from these touchpoints could prove to be fatal for the overall experience, for instance, any technical issue with account setup for new users could prevent the users from even using the app.

(In practice, Product Managers will have a deep understanding of users and user journey beforehand)

Assumption — As a Product Manager of Uber’s operation in India, we are going to focus on problems and factors originating in India only. In addition, we are focusing on Uber India APK for this case study.

In an attempt to narrow down on potential problem/s, we move on to brainstorm the factors listed below :

Internal Factors — These factors come into play due to the actions of internal parts of the organization. They could be related to technology, marketing, human resources, operations, etc. Therefore, it is paramount to ask the right set of questions to the stakeholders involved.

External Factors — These factors come into play due to reasons beyond the organization’s control. These factors may or may not have resulted due to internal actions.

Our questions for stakeholders are -

We also have to remember that Uber depends on its partnership with drivers and unions to supply the rides. Therefore, our questionnaire considers their experience as well.

(In practice, the Product Manager would already know answers to some of these questions)

We are going to assume the responses to the questions mentioned above. Out of all the responses, the following are worth mentioning :

• We haven’t made any significant changes to application UI/UX in the past 3 weeks.
• We have rolled out UPI payments (beta) last week along with minor updates to the ride search algorithm (3 weeks ago).
• Uber has changed the driver’s compensation process across 3 major states in India. Uber now mandates drivers to clock certain active hours (by staying online/available for ride) during the day to be eligible for bonuses.
• A few driver unions who had partnered with uber are now demanding the company to rescind the changes made to the compensation system. Some of these private partners and unions are on indefinite strike. Additionally, a lawsuit has been filed against Uber in Chennai
• Ola has ramped up their operations across India with fresh investment. Ola has introduced a company-owned fleet of 1200 (along with special discounts) cars to counter Uber’s dominance in 15 major cities across India.

Based on the abovementioned responses, we’ve come up with 4 major hypotheses -

1. If there is a recurring failure of new UPI services, then it could have led to customer dissatisfaction because customers will end up wasting time and retrying payments.
2. If there are some issues in the updated ride search algorithm, then it could have led to customer dissatisfaction because customers would find it difficult to book rides or connect to the drivers.
3. If drivers are leaving Uber because of the compensation system, then it could have led to an imbalance in rides demand and supply because of lesser cars on roads.
4. If customers are switching to Ola because of discounts, then it could have led to dissatisfaction in the customer segment who switches between the apps because they would be paying more using Uber.

Now that we have established the hypotheses, it’s time we use the data and analytics available at hand. Our objective is to look for anomalies in the metrics and the reviews.

Qualitative Analysis of Playstore/Appstore reviews: Since we’ve observed a significant drop in user ratings, it makes sense to start our investigation by analyzing the reviews. Our goal is to look for any patterns of discontent originating from a specific or a collection of different problems. We also have to make sure that the analysis is carried out by independent research teams to avoid any bias. The team can decide to start their investigation by looking for lower ratings (1–3) first which have comments attached to them.

Assumption — The dip in Application ratings has started only 2 weeks ago and has gradually come down

Filter through recent poor ratings — Uber

Noticeable Results:

1. We have received ~30,000 1 star (1/5) ratings in the span of the last 48 hours.
2. A considerable section of users is complaining about issues with Uber Pool. The issue revolves around heavy fluctuations in UberPool prices from the time of search to the time of ride selection. Some users claim to have witnessed a 100% hike

Social Media/Press Analysis: Due to ongoing strikes in some cities and states, we will also analyze current social media/press trends.

Noticeable Results:

#boycottuber has been trending on Twitter for the last 5 days

Posts: 19000

Users: 10200

Likes: 125000

Analyzing the Metrics: This step could very well be the first stop for many managers and teams because of readily available data. We will use this data and our understanding of the user journey to get relevant metrics that could hint at the problems.

Noticeable Results:

• A drop in Daily Active Users: Witnessed a sharp 20 % decline in DAU the last 5 days. 50% of this drop is attributed to 3 major states across India. Additionally, there has been a drop in Driver’s activity (DAU) by around 32% in the abovementioned states.
• Bounce Rate: 10% Increase in bounce rate on UberPool’s ‘Search Ride’ page. 70% of this drop is attributed to peak hours (8 AM-11 AM & 4 PM — 8 PM). There is no other noticeable change in bounce rate on other service pages viz. rental, premium, or ride.
• UPI Payment Success Rate: We have witnessed a staggering 85% success rate on UPI payments (beta). 20% of users have switched from alternate payment methods to UPI payments since release. The bounce rate on the payments page is normal.
• Average waiting time: This metric is used to measure the average time it takes for the user to enter their ‘destination’ and find a ‘ride’. Uber has witnessed a 10% overall spike in this metric. 70% of this spike is attributed to 3 major states across India. To find more about increment in average waiting time metric, we also decide to club these results with ‘Search Timeout’ per user. There is an increase of 30% in the results.

Considering these findings, we can now expand on or neglect certain hypotheses accordingly.

1. UPI Payment success rate along with its acceptance by the users indicates that there have been no considerable failures or issues with this service. Additionally, there have been negligible mentions of UPI failure/issues on Playstore/Appstore and social media. Therefore, we are neglecting the related hypothesis.
2. Specific bounce rate on the UberPool booking page along with the negative reviews about price fluctuation complaints on Appstore/Playstore indicate a potential problem. We will recognize this hypothesis as a root cause for the problem.
3. There is a steep drop in DAU (riders and drivers) in 3 major states across India. Negative press (#boycottuber on Twitter and 1-star ratings) could be a result of drivers and unions leaving (or going offline) Uber. Additionally, the negative anomalies in Average Waiting Time and Search Timeouts per user in these 3 states is an indicator of discontent among drivers and riders. We will recognize this hypothesis as a root cause for the problem.
4. There is no considerable change in the bounce rate metrics for any other rides services except UberPool. Since we do not have concrete evidence (social media analysis or Playstore/Appstore ratings) of discontent due to discounts or better offerings by competitors, we are neglecting the related hypothesis.

Root Cause :

After backing our hypotheses with data and logic, we have concluded that the root cause for our problem lies in the updated ride search algorithm which is impacting UberPool the most, and the ongoing discontent among drivers and unions which is a result of changes made to their compensation system. We have also found that the strikes and unrest (result of discontent among drivers) in a few states have caused massive damage to the company’s reputation in the media and press.

Solutions :

• Damage Control: Our first step should be to discuss with the legal team and executives to develop a strategy to contain the damage done by the new compensation system. High-level talks with the unions and state government should be initiated followed by a press release on the steps taken. We should assert our goal of providing world-class services to our customers and partners.
• Handle the irrational reviews: After analyzing recent poor ratings (within the last 3 days) on Playstore/Appstore, we have concluded that most of these ratings are not related to the App. Since we do not control the rating system, we should request support from Google/Apple to filter the irrelevant ratings and remove them.
• Issues/Bug Handling: As we have identified a problem with the Ride search algorithm, we have to work rapidly on catching these issues. We should start with working on the reviews and identify the problem from the comments if possible(sometimes we prompt users to fill forms explaining the issue in detail). Our approach should be to start by talking to Quality Analysts and Tech leads and letting them come up with a plan.

In the following impact matrix, the higher the ‘Impact’ rating, the higher the stakes. Lower the ‘Time to handle’ rating, more time it takes to resolve.

Finally, when we come up with our solutions, we go back to metrics to verify the improvements we were hoping for.

Disclaimer: This case study represents hypothetical scenarios and solutions and is written for educational purposes only.