As a leader in the ride-hailing industry, Bolt prides itself in on providing convenient, affordable, and accurate services to their customers. However, recent reports of overcharging have raised questions about the efficacy of the in-app pricing model. This study was conducted to investigate the overcharge issue and offer practical solutions.
There has been a recent uptick in the reports of overcharged tickets from our riders. This problem appears to be tied to the GPS Confidence, a key component in our price and duration prediction algorithm. Approximately 60% of the reported overcharges occurred when the GPS Confidence for the ride was inaccurate.
Based on the dataset acquired from the Machine learning team. Our team began an in-depth investigation to understand the scope of the issue. We focused on the remaining 40% of cases where the GPS Confidence was positive to understand why these rides still resulted in overcharge tickets. We found that 50% of these rides, despite having positive GPS Confidence, generated overpriced ride tickets. Out of these, 52.9% were upfront overcharges, while 47.1% were based off the Predicted Duration/Predicted Distance of the trips.
Based on our findings, we recommend the following actions to tackle the overcharging issue:
- Improve GPS Confidence Accuracy: GPS Confidence is a significant factor leading to overcharges. We should explore partnering with a more reliable GPS provider or investing in our own technology to ensure the accuracy of ride entry data.
- Refine Prediction Algorithm: Our current prediction algorithm is overestimating the cost in about half of the cases with accurate GPS Confidence. The algorithm could be improved to more accurately predict the duration and distance of trips, thus preventing overcharges.
- Implement User Review Mechanism: A feature that allows riders to review the estimated fare before confirming their booking might help in reducing the number of overcharge tickets. This will also provide the opportunity for the user to flag potential overcharges before they occur.