Rahul Aswani

- Was brought in as a management consultant by the VC firm to help their investee team build a startup from scratch and mentor the founders on the 0 to 1 journey.
- Led the team into customer discovery, asking open-ended questions to the customers, trainers, and all stakeholders involved to uncover the real insights to understand the customer pain point.
- Established the problem, solution, and implementation hypothesis for the idea and market and validated them via open-ended customer calls to invalidate some hypothesis and proceed on building the startup based on the validated ones.

Rahul was brought in to consult a scooter sharing startup with the objectives to
1) Improve operations process on the supply side
2) Improve asset utilization and
3) Enhance on-ground ops staff productivity.

Due to the high entropy and realtime nature of the business, which leads to scooters being driven and dropped in unpredictable locations across the city. Scooters need to be rebalanced periodically to ensure that assets are not lying around in underutilized areas.

At this stage of the startup, there was no data team or analysts available with minimal data logging. Worked Independently with available location data and identified proxy data in the absence of accurate data to build upon. Created queries to identify key hotspots in the city which should be prioritized for rebalancing over other scooter stations. Iteratively improved the query and later productized it to be used by non-technical operations teams which increased their productivity.

This data-driven approach prioritizing the top hotspots and modeling the radius around them led to an overall improvement of 7.4% in the utilization of the city's scooter supply

Ride-hailing is highly realtime and hence prone to higher risk.
With that goal to ensure the safety of all stakeholders on our platform. We explored a breadth of possible solutions to verify every single driver taking orders in realtime and in a scalable way.
We finally narrowed down to build a facial recognition product with mainly 3 components
1) Machine Learning-based facial recognition engine which checks the driver's selfie against his KYC.
2) Suspicious Activity Reporting (SAR) engine to automatically detect possible situations based on the risk factor in the situational context and trigger the app to ask the driver to verify himself.
3) Camera module to identify facial features via a selfie in the input format expected by the backend FR-engine
I led this initiative under the Platform Trust & Safety workstream at Careem. Starting by developing a nuanced understanding of the on-ground conditions, the quality of phones used by our drivers, and thinking through every edge case. We then invited and evaluated RFPs by 11 vendors to finalize one. Then various modules were scoped and prioritized for development and iteratively launched through A/B tests, continuously incorporating the learnings in next release

Over several iterations and new feature additions, the order acceptance screen had become crowded with useless information, impacting the understanding of the driver. Leading to them not being able to accept order beneficial for them and missing out on their earnings as well as the GoJek's core metric of acceptance rate.

To solve this, I went back to the drawing board to conceptualize a new clutter-free experience with just the right amount of information to enable drivers to make a decision within 10 seconds.

Organized group discussions and usability tests to empathize and understand from the drivers on what was leading to their poor acceptance rates. Not leading with a solution and asking them on what can be improved we finally found out that they were not able to identify how much will this order make them and what's the distance they need to travel for it.

Decluttered the order screen and prioritized only the most important four elements to be shown to the driver for him to make a decision within ten seconds whether to accept or pass on the order.

This led to significant gains on the order acceptance rates across all categories as the right drivers were accepting the orders they found right for them

While leading the app performance improvement project. We integrated facebook's open-source network monitoring library https://github.com/facebook/network-connection-class to measure the network stability and speed of our customers. We had set out to optimize our API request and response payloads to work with the flaky and slow internet in Indian tier 2-3 cities back in 2015-2016.

On analyzing the data for this project it was discovered that Ola customers lose internet at a lot of critical events such as app launches and just while confirming their ride. This affected our bottom line as well.

Analyzed this data along with the existing conversion funnel drops to make a coherent picture and build a business case to increase 0.5% revenue via this feature. Presented this project plan to the CEO, CTO, director of product, and customer apps. Got the approval to build the feature.

I envisioned the product, collaborated with designers on the UX, worked out the details with engineers on the SMS gateway, integration and SMS input ingestion engine to building an MVP of Ola offline booking via SMS. This was then beta-tested in some select cities before collecting the feedback for the next round of iteration and India-wide launch.

Right segmentation is the key to understanding your user better. Getting it right helps retaining high LTV users and saving money on the ones which are already engaged on your platform and do not need further discount or incentivization to use your product.

By building the segmentation engine we were able to understand our customers better with respect to all the products offered by us. It helped teams such as Ola share, Ola pass, Ola select, Ola loyalty and more to identify and target the right user segments with precision improving the conversion of all these programs more than 15%. We later used data science and supervised learning to detect potential users who are going to churn and prevent that by offering promotions specifically to them in a personalized fashion.

Drivers try to find land exploit loopholes in the system to get more incentives than they deserve.

The driver's incentive calculation depends on multiple factors including but not limited to the order acceptance rate which is further dependent on an order acknowledgment rate. Only orders that are acknowledged on the device would get accepted or rejected.

(details are confidential)

This was leading to millions of dollars of overpayment in incentives to drivers who were actually not performing the work.

We solved the loopholes by rethinking the order flow from first principles and were able to achieve 26.5% improvements on the system.

Rahul defined the charger persona, created their user journey map and identified pain/gain points. Prioritized the top issues to work on based on the RoI and ease of development.

He gathered feedback, conceptualized features, wrote requirement docs and user stories and created a backlog for the following problems

Prob 1) After locating scooter on the map, its frustrating to reach there and not be able to find it

Sol 1.1) Feature to find the scooter in the real world with the help of 360 videos shot by the previous user who dropped it

Sol 1.2) Hardware feature to make the scooter loudly beep during the day and blink its flashlight during the night

Prob 2) Accessing scooters parked in locked/private property

Sol 2.1) Reverse geocoding the location of scooters and fetching the property type to identify If it's private property. Then set the right expectation for the charger or assigning it to ops.

Prob 3) Showing right scooters for chargers to pickup

Sol 3.1) Showing the user too many scooters around them might make them try far away scooters, which might not stay there until the user reaches there

3.1.1) Dynamic radius based on the availability of chargers in the area

3.1.2) Locking 1 scooter for 5 mins

Consulted a client launching a video commerce app for SEA

1. Understood the landscape for 3 industries - Social media (Instagram, TikTok, and YouTube), video streaming, and eCommerce to find an overlap opportunity

2. Conducted market research to identify target customer segments and build personas for potential users

3. Ideated new set of features for content creation, content sharing, transacting over the app, and mobile checkout flow

4. Prepared mockups and put together an InVision clickable prototype to gather user feedback on the user experience of the prototype and iterated as per the target user's needs and feedback

5. Ideated hypothesis to validate at a large scale

6. Launched a minimal MVP to collect data and iterate on the features and then finalized the first version to launch publicly

7. Planned the technical implementation of the Android/iOS mobile apps, video streaming back end system, CMS platform, eCommerce checkout modules

8. Scoped the work in quarterly roadmaps, broke the roadmap in product backlog with epics and user stories defined to the final details

9. Planned the go-to-market strategy and marketing plan

Delivered the artifacts to the client's team and assisted in the execution.

Navigating to the customer's location is a big part of the end-to-end trip experience. Also the most crucial as the delivery of entire cab service starts from this point.

Seeing the huge volume of trips in the ride-sharing industry and usage of google maps on most android phones for free. Google started working on its own navigation SDK which would combine its multiple web-based APIs in the form of a mobile SDK

1. Directions API - Provide directions for transit, biking, driving, or walking between multiple locations.

2. Distance Matrix API - Calculate travel times and distances for multiple destinations.

3. Roads API - Determine the precise route a vehicle travels.

4. Geolocation API - Return the location of a device without relying on GPS, using location data from cell towers and WiFi nodes.

Google chose to partner with GoJek for the beta trials of this SDK.

I led this collaboration from GoJek, helping the google team understand the nature of our business, evaluating their SDK in our context, scoping out the integration work, creating and adding stories to the product backlog, launching the nav SDK to our beta users and then A/B testing with larger set and give feedback to google over weekly sync up calls.

With the hyper-growth of GoJek in Indonesia during 2015-2017, it attracted attention from the Indonesian Government to enforce stricter adherence to KYC compliance of the Drivers registered on the platform.

This posed a big challenge in front of us as with a large fleet of more than a million drivers at that time. We had to not only build a productized flow to ensure KYC compliance for the thousands of drivers joining us every day but also ensure the same compliance standards on the existing fleet with very little proper documentation in place.

Led the development of the platform with these modules

1) Document submission module in the driver apps

2) OCR module to extract the information from photos and do verification before submission

3) Verification dashboard for ops teams to manually verify and cross-check the accuracy of information from previous modules

4) Reporting on top of all the modules, including ops agent performance.

The first iteration of this platform was built within a month, including ideation, gathering requirements, finalizing tech stack, system architecture, and a minimal UI. This product was more impactful than any bottom-line impacting products as it allowed us to operate legally in Indonesia.

Being able to track the exact location of an asset has a huge impact on the bottom line. Though it also has an equally negative effect on the bottom line in the case the asset cannot be located correctly either by:
1. The user intending to book
2. The Ops team/Hunters recovering missing scooters

The battery backup of shared scooters only lasts 24-48 hours and an asset needs to be recovered in this time otherwise it can be lost and written off.

Challenge was to come up with a way to improve the accuracy of the GPS coordinates sent by the Hardware GPS tracker attached to the scooter.

- Started by understanding the location transmission process of the Oyster 2 device.

- Quantitatively analyzed the location sent by 10 scooters in the last 24 hours.

- Did 6 hours of a field test by driving the scooter along with sending a location from a GPS based mobile app to compare the deviation

- Conceptualized and implemented a solution to triangulate location values on a rolling basis and identify the most accurate location to then send it downstream to the other system.

- A/B tests of text and control showed a 23% improvement in locations send via this algorithm

Driver Platform at GoJek is a suite of driver apps used by 3 Million GoJek drivers. As a part of the apps built payment wallet related functionality

Payments product included the following systems

1) Payments and dues re-conciliation to check if the driver has settled all his dues and there is no outstanding payment.

2) Check if the driver has sufficient balance before being given an order.

3) Withdrawing earnings from wallet to the bank account

4) Driver to Driver top-up via GoPay wallet

5) Customer GoPay wallet top-up via cash through drivers

6) Phone and other recharge for customers via driver GoPay wallet

It's always hard to co-relate customer experience KPIs changes to impact on the bottom line.

Finding the most pressing issues which are leading to low trip ratings and customer churn in millions of bookings is like finding a needle in the haystack.

You can't have feedback options like
- Our driver was sweating and smelling
- Our driver kept talking the whole ride whereas I wanted to sleep
- My driver appeared suspicious

A 1 to 5-star rating does not help much except pain a quantitative picture. Comments are where the real pot of gold lies.

We set out to build a scalable engine which can go through the comments of all the trips rated below 3 and perform the following data manipulation
1) Identify the sentiment of the comment as Positive, Neutral, Negative, and Threatening.

2) identify keywords in the comments which can give a list of the whole comment

3) classify the comment based on these keywords under various buckets such as unprofessional behavior, car issues, and more.

Built the MVP withing 2 weeks to present it to the Director of product and later plan, scope and create the backlog for the full-fledged tool.

In the first month of my joining GoJek, I was exposed to the huge problem growing day by day - Identity theft of drivers.

Drivers who not habitual email users were using the then flawed fixed email and password pattern created en masse for logging into the GoJek app which had their profile, financials details, and earnings.

Drivers were susceptible to Identity theft via pattern recognition, social engineering, and fake documents.

We decided to create a new identity system for our drivers.

To ensure that the new system is full proof we laid out the first principles the new system would be built on.

1. The identity needs to be verified in realtime
2. Identity needs to be verified every day using OTP verification
3. At any given point in time, only 1 instance of the account can be active.

I led my team to build the new migration feature which

1. Requested each driver to register their login number
2. Govt identity verification flow for no mismatch cases
3. Self serve identity verification without visiting the ops center or losing day's work

Technologies used
1. Facebook SMS kit, Twilio SMS kit, and local SMS gateway provider APIs
2. OCR reading vendor
3. Redis

The client is building a new social network based on common human interaction principles in a format never tried before.

He hired me with just a vision doc where he had written down all the ideas he had around this new social networking app. At first, it was hard to understand even for me as the app was trying to solve too many problems at the same time and did not have a clear target user persona in mind.

I took on this project as a challenge and took the client through a process of design thinking with intense brainstorming on
- The core value proposition of the product
- What problem does it solve or what need does it fulfill for the end-user
- How would certain users use this app in the real world scenarios
- Who are our target users who would be quick on adopting such a new social network

Together with an approach of asking hard questions to break the idea and a lean approach to designing mockups which helps us visualize the concept more clearly. I delivered a clickable prototype of the app which the client now uses as a low-fidelity MVP to explain his idea better to target users and Investors.

The client came in with just an idea and wanted to make sure they are starting on steady ground.

Lead the entire project as an interim founder/CEO to build a start from 0 to 1.

- Worked on understanding the Saudi B2B logistics market
- Analyzed the trends via market research and competitor analysis
- Interviewed people at the grassroots level to understand the stakeholders involved in a B2B shipping process
- Crated customer personas and interviewed them with open-ended questions to extract insights, not just data
- Create a process diagram and identified problems for the stakeholders in the process they follow
- Planned operations and features around the pain points identified
- Trained the operations team on interviewing customers
- Setup operations processes to run an MVP-pilot for the actual business on the ground
- Planned the product roadmap for V1 which support the on-ground operations and business teams

Took the project from an idea and delivered a working pilot program to the founder to fundraise with strong traction, business, and product KPIs results that the investor needed in order to do their successful due diligence