Scootistic DC

Author

Lola Nurullaeva, Sachi Nandurkar, Shubham Patil, Dhanyasri Bolla, Harshal Dinesh Soni

Published

December 8, 2023

Abstract

This project explores the feasibility of an integrated umbrella subscription service for e-bike sharing platforms in Washington, D.C., encompassing providers like Capital Bikeshare, Lime, and Veo. The service aims to enhance user convenience, streamline pricing, and provide e-bike manufacturers with valuable insights into consumer preferences and usage trends.

The survey design will examine key product attributes, such as vehicle type options (scooters only, bikes only, or both), ride duration limits (30, 60, or 180 minutes), and extended ride fees ($1, $5, or $10 per additional minute). Critical research questions include understanding consumer preferences across bike-sharing services, identifying optimal pricing schemes for user retention, and assessing potential partnerships with existing D.C. e-bike platforms.

Results include coefficient estimates for each attribute, visualized with 95% confidence intervals to highlight the impact of ride characteristics on user choice. Power analysis indicates that a sample size of 450-500 is optimal for reliable coefficient estimation, particularly for the “vehicle_typeBoth” variable, which shows higher variability at smaller sample sizes.

Introduction

With the growing popularity of e-bike and scooter-sharing services in urban areas, cities like Washington, D.C., are becoming hubs for micromobility options that promise convenience and environmental benefits. However, navigating the diverse offerings of various e-bike providers, such as Capital Bikeshare, Lime, and Veo, can be a challenge for frequent commuters. In response to this, our project explores the feasibility of an integrated umbrella subscription service designed to simplify the user experience while providing flexibility and cost-effective options.

This project seeks to enhance the convenience of e-bike usage in Washington, D.C., by offering a unified subscription model that covers multiple providers. The proposed service will streamline pricing structures and enable e-bike manufacturers to gain valuable insights into consumer preferences and usage patterns. Our analysis leverages a comprehensive survey design to investigate the key attributes influencing ride-sharing decisions, such as vehicle type preferences (bikes, scooters, or both), ride duration limits, and fees for extended rides.

We aim to address critical research questions, including consumer preferences for vehicle types, optimal pricing strategies for user retention, and the potential for strategic partnerships with existing micromobility platforms. Using multinomial logit (MNL) models and the logitr package in R, we analyze factors driving ride-sharing choices and identify patterns that can inform the development of a robust subscription service. Through a detailed evaluation of user behaviors and preferences, this project will contribute valuable insights for designing an efficient and appealing solution for D.C. commuters.

Survey Design

Eligibility Requirements

The ideal respondents for this survey are DC commuters who frequently travel within or to the city, particularly those familiar with rental bikes and scooters. Our general target population includes all DC commuters, regardless of their mode of transportation—whether they use bikes, scooters, trains, cars, or buses. Eligibility will be determined by a simple yes/no question: “Do you commute within or to DC?”

Respondent Information Collected

We collected critical demographic information from respondents, including:

  • Age

  • Gender

  • Employment status

  • Household income

  • Level of education

Conjoint Choice Questions

Attributes and Levels:

Vehicle Type: Scooter only, Bike only, Both

Ride Duration Cap: 30 minutes, 60 minutes, 180 minutes

External Ride Fee: $1 per minute, $5 per minute, $10 per minute

Purpose of These Questions:

Price Sensitivity: To gauge how much users are willing to pay and to identify optimal pricing strategies that balance profitability and user satisfaction. Vehicle Preference: To determine which vehicle type is favored, aiding resource allocation and negotiations with service providers. Ride Duration: To understand preferred ride limits and assess the viability of tiered service offerings. Combination of Features: To discover the most appealing combinations of pricing, vehicle types, and ride durations for potential subscription plans. Strategic Rationale By varying these attributes across different scenarios, we aim to observe:

Price Sensitivity: Understanding the willingness to pay for specific services and how price impacts perceived value. Vehicle Preference: Identifying whether users favor specific vehicle types or prefer having access to both. Ride Duration Importance: Assessing preferences for shorter, cost-effective rides versus longer, pricier options. Combination of Attributes: Analyzing how various attribute combinations affect consumer choices, guiding the development of a financially viable and attractive service for DC commuters.

Data Analysis

Sample Description

The cleaned survey sample captured a total of 87 respondents and recorded 1566 choice responses. Summary statistics for key demographics are provided below.

Demographic Distributions

Gender

Race

Education

Income

Data Cleaning

The survey collected a total of 264 respondents, but after applying the data cleaning process, only 87 respondents remained in the dataset. To ensure data quality for the analysis, several filtering steps were implemented to remove incomplete or low-effort responses from the raw data.

First, any responses where the survey was exited early, indicated by a non-missing value in the time_p_end_screenout variable, were excluded. This ensured that only fully completed surveys were retained for analysis. We also removed respondents who skipped any of the six choice-based conjoint (CBC) questions, ensuring that all key questions had consistent responses. Additionally, any responses showing a uniform pattern of selecting the same option across all CBC questions were filtered out, as this behavior suggested a lack of engagement.

To further improve the dataset’s reliability, we calculated each respondent’s total survey completion time (time_total) and set a minimum threshold based on the 10th percentile of all completion times. Responses submitted faster than this threshold were discarded, as they indicated potential speed-through behavior. These steps ensured that the final dataset contained only thorough and engaged responses, improving the consistency and quality of the data for analysis.

Modeling

By using our baseline linear logit model, the following utility model was estimated:

\[ \begin{align} v_{j} = 0.00671975*price_{j} + 0.00015938*ride\_dur\_cap_{j} + 0.01477000*ext\_ride\_fee_{j} \\ - 0.16315444*vehicle\_type\_bike\_only_{j} - 0.21807098*vehicle\_type\_scooter\_only_{j} \end{align} \]

The coefficients for vehicle types suggest a preference against “bike only” and “scooter only” options relative to the baseline alternative (“both” bikes and scooters), although their p-values indicate weak statistical significance.

Results

WTP for each attribute with 95% CI

knitr::include_graphics(here::here("figs", "mnl_wtp.png"))

These figures helps us to explore the willingness to pay (WTP) for various service features, focusing on ride duration caps, extended ride fees, and vehicle type options. The results highlight user preferences, the variability in their responses, and the implications for service pricing and optimization.

  1. Impact of Ride Duration Cap on WTP
Key Findings:

At 30 minutes, the mean WTP is $0, acting as the baseline.

At 60 minutes, the WTP decreases slightly to -0.64, with a wide confidence interval ($-8.07 to $6.78), suggesting varied user preferences.

At 180 minutes, the WTP drops significantly to -3.19, with an even wider confidence interval ($-40.34 to $33.92), indicating greater uncertainty and diverse reactions to longer ride durations.

This means there is a negative correlation between ride duration caps and WTP, suggesting that longer durations lead to diminishing value perceptions. The increasing variability in WTP at higher caps reflects differing user preferences, pointing to potential opportunities for personalized pricing or service packages.

  1. Impact of Extended Ride Fee on WTP
Key Findings:

At 1/min, the mean WTP is $0, serving as the reference level.

At 5/min, the mean WTP drops to approximately -8.65, with the confidence interval spanning -37.61 to 20.07, indicating significant variability.

At 10/min, the mean WTP decreases further to -19.47, with a broader confidence interval (-84.62 to $45.17), suggesting a high degree of uncertainty.

Higher extended ride fees result in lower WTP, signaling a diminishing value for users as prices increase. The growing uncertainty with higher fees highlights the need for careful fee structuring to balance revenue generation with customer satisfaction. Segmenting the customer base based on fee tolerance could help tailor offers and enhance customer retention

  1. Impact of Vehicle Type on WTP
Key Findings:

Bike Only services have a mean WTP of 23.86 with a confidence interval ranging from -50.22 to 98.45.

Scooter Only services show a higher mean WTP of 31.93, with a wider confidence interval (-59.98 to 124.02).

The Both option, offering both vehicle types, has a mean WTP of 0, serving as the baseline.

Users display a stronger preference for scooter-only services, as reflected in the higher mean WTP. The wide confidence intervals, especially for scooters, highlight considerable variability in user preferences, which suggests a need for customized service offerings to cater to diverse user needs.

Interpretation of the WTP Comparison Across Different Attributes

knitr::include_graphics(here::here("figs", "mnl_wtp_barplot.png"))

The barplot shows a comparison of willingness to pay (WTP) for changes in four attributes: Ride Duration Cap, Extended Ride Fee, and Vehicle Type (Bike vs. Both, Scooter vs. Both). The values have been adjusted for comparison, and the results reveal insights into how different factors affect user willingness to pay.

  1. Ride Duration Cap (+30 min)

Mean WTP: $-0.64 Confidence Interval: $-8.07 to $6.78

Interpretation: When the ride duration cap is increased by 30 minutes, the mean WTP is negative, suggesting that users are less willing to pay for longer ride durations. The broad confidence interval indicates significant variability in user preferences. Some users may still value the longer duration, while others may not find it worthwhile.

  1. Extended Ride Fee (-$5/min) Mean WTP: $10.82 Confidence Interval: $-25.09 to $47.01

Interpretation: A decrease in the extended ride fee by $5/min results in a positive mean WTP, indicating that users appreciate lower fees. The confidence interval is wide, reflecting uncertainty in the exact WTP, with some users showing strong preference for the reduced fee, while others have a much smaller or even negative willingness to pay. This highlights the importance of balancing fee adjustments.

  1. Vehicle Type: Bikes vs. Both Mean WTP: $23.86 Confidence Interval: $-50.22 to $98.45

Interpretation: Users are generally willing to pay for bike-only services compared to the “Both” option, with a mean WTP of approximately $23.86. However, the wide confidence interval suggests significant variability in preferences. Some users may find bike-only services highly valuable, while others might prefer access to both vehicles or have lower willingness to pay for bikes.

  1. Vehicle Type: Scooters vs. Both Mean WTP: $31.93 Confidence Interval: $-59.98 to $124.02

Interpretation: The scooter-only option shows the highest mean WTP, $31.93, indicating that users are more willing to pay for scooters than for both vehicle types combined. The confidence interval is quite broad, with a high degree of uncertainty in individual preferences, but overall, the scooter option appears to be more highly valued than the combined offering of both vehicles.

Simulations

Single Market Simulation
knitr::include_graphics(here::here("figs", "single_market_simulation.png"))

Multiple Market Simulation
knitr::include_graphics(here::here("figs", "multi_market_simulation.png"))

The market simulations were performed to evaluate market share across competing alternatives: Bike Only, Scooter Only, and a Combined Option. These scenarios were designed to reflect realistic competitive conditions. Below is a summary of the attributes for each alternative:

library(knitr)

market_table <- data.frame(
  Alternative = c("Bike Only", "Scooter Only", "Both"),
  `Price ($)` = c(15, 25, 20),
  `Ride_Duration_Cap` = c(60, 30, 180),
  `External_Ride_Fee` = c(5, 1, 10)
)

kable(
  market_table, 
  caption = "Attributes of Product Alternatives in the Simulated Market", 
  align = c("l", "c", "c", "c")
)
Attributes of Product Alternatives in the Simulated Market
Alternative Price…. Ride_Duration_Cap External_Ride_Fee
Bike Only 15 60 5
Scooter Only 25 30 1
Both 20 180 10

This configuration ensures a diverse set of options in terms of price, usability (duration caps), and external fees, allowing us to study trade-offs between affordability and flexibility.

Simulated Market Shares with Confidence Intervals:

The single market simulation plot shows the market share and 95% confidence intervals for the three alternatives. Bike Only captures approximately 25% of the market, with a confidence interval indicating limited uncertainty. Combined Option is the most popular, with market share nearing 50%, reflecting its balanced attributes and extended ride duration cap. Scooter Only has a lower market share (~25%) but exhibits the highest variability in the confidence interval. The multiple market simulation plot extends the analysis across multiple scenarios, showing consistent performance trends for all alternatives. Variability is observed across scenarios, particularly for the Combined Option and Scooter Only alternatives.

Key Insights from Simulations:

Market Preferences:

The Option is the most competitive, likely due to its extended ride duration cap and mid-level price. This suggests that flexibility is a significant driver of market adoption. The Bike Only alternative struggles to compete due to its shorter ride cap and higher relative external fees compared to Scooter Only.

Confidence Interval Interpretation:

The large confidence intervals for the Scooter Only option indicate high uncertainty, possibly due to niche appeal or sensitivity to external market conditions.

Opportunities for Increasing Demand:

Attribute Sensitivity:

Ride Duration Cap: Increasing this for Bike Only to match or exceed the Combined Option could make it more competitive. Price: A small reduction in price for Scooter Only could improve its appeal, potentially boosting its market share closer to the Combined Option.

Competitive Design: Focus on optimizing attributes that balance flexibility and affordability. For instance, lowering external ride fees for Bike Only could provide a competitive edge over Scooter Only.

Attribute Ranges for Competitive Outcomes:

Price: The Combined Option demonstrates that a mid-range price (~$20) is optimal for capturing market share while maintaining revenue potential. Ride Duration Cap: A significant driver of adoption. Extending caps for Bike Only and Scooter Only alternatives could challenge the dominance of the Combined Option. External Fees: Reducing external fees, particularly for Bike Only, offers an opportunity to increase adoption among price-sensitive users.

Sensitivity

Market Share Price Plot
knitr::include_graphics(here::here("figs", "share_price_plot.png"))

This plot illustrates the relationship between market share and price for the bike alternative. As price increases from $10 to $30, the market share shows minimal sensitivity, remaining relatively stable at approximately 25-30%. The shaded region represents the confidence interval, indicating the level of uncertainty around these predictions.

Key Observations:

Price Sensitivity: The low sensitivity of market share to price changes suggests that customers of this product segment are relatively price inelastic within the tested range. Uncertainty: While the confidence interval slightly widens at higher prices, the predictions remain reasonably reliable across the tested range.

Revenue Price Plot
knitr::include_graphics(here::here("figs", "rev_price_plot.png"))

This plot shows the sensitivity of total revenue to changes in price. Here, revenue increases consistently as the price rises, which is a direct result of the minimal decline in market share and the arithmetic relationship between price and revenue. Revenue uncertainty increases at higher prices due to compounded uncertainty in market share predictions.

Key Insights:

Optimal Price Point: From the observed trends, higher price points (approaching $30) are more favorable for maximizing revenue, assuming market share remains relatively constant. Uncertainty in Revenue: The broader confidence interval at higher prices suggests a need for caution when making revenue predictions for pricing beyond $25.

Tornado Plot
knitr::include_graphics(here::here("figs", "tornado_plot.png"))

This plot evaluates the sensitivity of market share to various product attributes, including price, ride duration cap, and external ride fee. By altering these attributes by ±20%:

Price has the largest impact on market share, albeit still relatively small. A 20% increase or decrease in price results in slight changes in market share, reinforcing the earlier observation of low price elasticity. External Ride Fee and Ride Duration Cap have smaller impacts on market share compared to price, indicating that customers are less sensitive to these attributes within the tested ranges.

Best Price Point and Design Decisions

Pricing: The analysis suggests that higher price points (near $30) are optimal for revenue maximization, as market share remains relatively stable. However, this comes with increased uncertainty. Product Design: The tornado plot indicates that design decisions, such as adjusting ride duration caps or external ride fees, have limited effects on market share compared to price. Hence, these features could be optimized for cost-efficiency rather than demand generation. Uncertainty Management: While confidence intervals are narrow at lower price points, the widening at higher prices necessitates careful consideration of market conditions and further data validation before implementing premium pricing.

  • Limitations of our analysis:

The analysis is based on a sample of 87 respondents, which may not fully represent the broader population, potentially limiting the generalizability of the results. The small sample size could lead to biases or reduced accuracy in estimating preferences, especially for niche vehicle options like scooters. Additionally, the study focuses on specific attributes (ride duration cap, extended ride fee, vehicle type) and may not account for other influencing factors such as customer loyalty, regional preferences, or broader market trends. Future research with a larger, more diverse sample and expanded variables would provide more robust insights.

Final Recommendations and Conclusions

Summary of Findings

  1. Impact of Ride Duration Cap on WTP:
    • As the ride duration cap increases (from 30 minutes to 180 minutes), the mean WTP decreases, indicating diminishing value perception.
    • The confidence intervals are wide, suggesting significant variability in user preferences for ride durations.
  2. Impact of Extended Ride Fee on WTP:
    • Higher extended ride fees result in a decrease in WTP, with confidence intervals indicating substantial uncertainty in user responses.
    • The WTP drops significantly as fees rise, indicating that users are sensitive to price changes.
  3. Impact of Vehicle Type on WTP:
    • Users show a stronger preference for scooter-only services over bike-only services, with scooters having the highest WTP.
    • Both bike-only and scooter-only options show significant variability in user preferences, reflected in wide confidence intervals.
    • The “both” vehicle type option has a neutral mean WTP, serving as the baseline.
  4. Market Simulations:
    • Single Market Simulation: The “Both” vehicle option captures the largest market share (~50%), while “Bike Only” and “Scooter Only” each capture around 25%. The Scooter Only option shows high variability in market share.
    • Multiple Market Simulation: The trends from the single market simulation hold across multiple scenarios, with the “Both” option consistently performing well.
  5. Sensitivity Analysis:
    • Market Share Price Plot: Market share for the bike alternative is relatively inelastic to price changes within the tested range.
    • Revenue Price Plot: Revenue increases consistently with price but shows higher uncertainty at higher price points.
    • Tornado Plot: Price has the largest impact on market share, while ride duration cap and external ride fee have a smaller influence.

Conclusion

The analysis of ride-sharing service preferences reveals several key insights that can inform strategic decisions for service optimization.

  1. User Sensitivity to Pricing:
    • The results demonstrate that users are highly sensitive to ride duration caps and extended ride fees. As these factors increase, the willingness to pay (WTP) decreases significantly. This highlights the importance of carefully balancing pricing models to maintain user satisfaction and revenue generation.
  2. Preference for Scooter-Only Services:
    • The preference for scooters over bikes, with scooters showing a higher mean WTP, suggests that the service should prioritize scooter availability and possibly offer incentives for scooter use. However, the variability in user preferences suggests that offering a variety of vehicle types may also attract a broader customer base.
  3. Market Share Insights:
    • The “Both” vehicle option consistently captures the largest market share, indicating that users appreciate the flexibility in choosing between different vehicle types. This could be a key strategy for increasing market penetration. However, pricing and vehicle type options should be adjusted periodically to cater to shifting user preferences.
  4. Revenue Optimization:
    • Pricing has a direct impact on both market share and revenue. As such, understanding how different pricing strategies affect demand is crucial. In particular, keeping extended ride fees competitive and offering flexible duration options will likely increase user retention and overall revenue.

In summary, the findings suggest that adjusting pricing models, emphasizing scooter options, and offering flexible ride configurations can improve customer satisfaction, enhance market share, and drive revenue growth. Further experimentation and sensitivity analysis are necessary to refine these strategies and ensure they align with evolving market dynamics.

Recommendations

  1. Optimizing Pricing and Fee Structure:
    • Lower extended ride fees to enhance customer satisfaction, as users show high sensitivity to price changes.
    • Consider offering tiered pricing models to cater to different user segments, especially with regards to scooter and bike options.
  2. Focusing on Vehicle Type:
    • Emphasize scooter-only options, as they show a higher WTP compared to bike-only or combined vehicle types.
    • Explore offering specialized packages for different vehicle types to cater to varying user preferences.
  3. Service Customization:
    • Given the variability in user preferences for ride duration caps, consider offering personalized ride durations to appeal to different segments of users.
    • Implement flexible ride duration packages and experiment with a variety of ride durations to optimize user satisfaction.
  4. Market and Revenue Optimization:
    • The “Both” vehicle type option should be prioritized as it captures the largest market share, but efforts should be made to boost the market share of scooter-only services through pricing adjustments.
    • Adjust pricing strategies for bike-only services to make them more competitive, potentially through reduced external fees or extended duration caps.
  5. Long-Term Strategy:
    • Implement simulations and sensitivity analyses regularly to fine-tune pricing models and service offerings.
    • Monitor customer preferences and adjust the service configuration (e.g., ride duration caps, vehicle types, fees) based on ongoing feedback to maintain a competitive edge.

Limitations

  1. Restrictions on Sample Size:

Despite offering insightful information, the final sample size of 87 respondents might not accurately reflect the broad community of commuters in Washington, D.C. The findings would be more reliable and broadly applicable with a larger sample.

  1. Demographic Diversity:

Despite efforts to gather data from a variety of demographics, it’s possible that some groups were underrepresented, which could have skewed the results and missed particular user needs or preferences.

  1. Restricted Range of Attributes:

A particular set of characteristics (vehicle type, ride duration cap, and extended ride cost) were the subject of the poll. Although they weren’t mentioned, other elements that can affect user decisions include weather, docking station accessibility, and integration with public transit.

  1. Assumption of User Behavior:

The conjoint choice model makes the assumption that users make logical decisions, which might not account for all behavioral subtleties, including impulsive or emotional choices that users make in real-world situations.

  1. Geographic Scope:

Because the study is restricted to Washington, D.C., commuting patterns, user demographics, and micromobility infrastructures may differ from those of other cities.

  1. Static Pricing Evaluation:

Dynamic pricing methods, which are frequently employed in actual micromobility services to control supply and demand variations, are not taken into consideration in this research.

  1. Survey weariness and Engagement:

It’s possible that respondents encountered survey weariness, especially when answering choice-based conjoint (CBC) questions, which may have impacted the caliber of their answers.

  1. Technology-Specific Preferences:

Preferences for specific brands or technology features (e.g., app usability, vehicle quality) were not explored, potentially overlooking factors that could influence user satisfaction and retention.

Attribution

All members contributed equally.

Appendix