INDUSTRY: TRANSPORTATION
Client Background
XYZ Inc. has operations in over 785 metropolitan areas worldwide with over 110 million users. Ridesharing is a very volatile market and demand fluctuates wildly with time, place, weather, local events, etc. The key to being successful in this business is to be able to detect patterns in these fluctuations and cater to the demand at any given time.
Problem
Analysis
EDA on 29101 observations, 13 variables across 6 months of raw data e.g. pickup_date, pickup_month, pickup_time, pickup_hour, location, no. of pickups, wind speed, visibility, etc.
Solution
most mature market, key factors affecting pickups (ranking), demand recommendations (customer segments, high demand zones, fleet & routing optimization, optimal pricing strategy, etc.)
Outcome
INDUSTRY : MEDIA & ENTERTAINMENT
Client Background
ABC Ltd. is a leading OTT service provider launched in 2017 by the XXX Group. It offers a wide variety of content (movies, web shows, etc.) for its users. They were experiencing a sudden decline in viewership of content.
Problem
Decline in the number of people coming to the platform, decreased marketing spending, content timing clashes, weekends and holidays, etc.
Analysis
Analyzing and visualizing the data to draw inferences around the influence of different factors on the first-day viewership.
Evaluating a Linear Regression model and observing the coefficients of the model to determine the factors that influence the first-day viewership.
(The constructed dataset contained content release day view counts, along with other factors related to the content such as average visitors for the past week, the genre of the content, day of release, etc. It had 1000 row entries representing individual content pieces and 7 attributes that may affect the first-day viewership)
Solution
to make trailer reach a priority, drive visitors around launch, leverage the weekend — test Saturday launches, schedule tentpoles in Summer, Genre strategy — promote Sci-Fi, optimize ad spend mix — focus on quality reach vs raw impressions and so on
Outcome
INDUSTRY : HEALTHCARE
Client Background
Care provider CCC Clinics was facing issues regarding primary & remote care. The Clinic chain had 80-100 OPD patients per unit per day. Few of these patients were only checking up on their health and had no serious ailments, thus consuming physician time. The patients outnumbered physicians by a vast ratio, adding to waiting lines and other care delays. Another set of patients, failed to follow remote care instructions on time resulting in late interventions.
Problem
Primary care and remote monitoring gaps create late interventions and unnecessary clinic load (Limited access to clinician time; long wait times for appointments; Patients lack continuous monitoring; 56% of patients missed early warning signs; High volume of non-urgent visits (72%) and inappropriate ER utilization; Fragmented follow-up; poor medication adherence)
Analysis
NLP for conversation + diagnostic/symptom triage models + time-series anomaly detection for telemetry + personalization/recommendation models
Solution
Deployed a Virtual Health Assistant (VHA) that:
Outcome
INDUSTRY : AUTOMOTIVE (ELECTRIC VEHICLES)
Client Background
Japanese EV manufacturer & retailer XXX, had a large number of prospective buyers enlisted on their CRM. At the same time, there happened to be a shortfall of resources within the organization, especially sales. One of the issues faced by XXX was to identify which of the leads are more likely to convert so that they could optimize the resource allocation.
Problem
Poor lead conversion, lacking revenue growth, determination of buyer “metric of interest”
Analysis
EDA - converted vs non-converted leads; activity impact; moment of truth; cross-tab by referrals; conditional box plots; confusion matrix; Accuracy - Recall - Precision - F1 score; Resource Allocation, Budget optimization, product improvement, etc.
Solution
Outcome
INDUSTRY : CPG + RETAIL
Client Background
KLN Pvt. Ltd supports Retail + CPG brands and their consumer group comprising busy families, health-conscious consumers, budget shoppers via pantry & grocery automation services. KLN proposed a reduction in food waste and grocery spend, to auto-manage pantry, personalized meal suggestions, smart reorders while delivering upon privacy-first partner offers
Problem
Modeling & Methodology
Solution
Outcome
INDUSTRY : BANKING & FINANCIAL SERVICES
Client Background
Founded in 1993, YYY was one of the biggest banks in ABC. (>20 million customers servicing close to 2000 cities). Customer base was skewed towards young, urban and tech-enabled individuals, while banking resources were focused on traditional customer engagement touchpoints (inefficient, lacking scalability)
Problem
Customer base was rapidly growing. There was pressure to implement scalable and cost-efficient ways to improve customer service
NLP/ML
NLP was used for cleaning the raw data. Various ML techniques like Deep learning, classification, anomaly detection and clustering were used to learn from the data. Sentiment analysis was used to analyze customer’s emotional response to the resolution provided. Deep learning based techniques were used to help the chatbot work in languages other than English. Monitored the performance of the chatbot and iterate/tweak to make the chatbots’ performance comparable to, or even better than, that of human responses.
Solution
Data showed that over 80% of the queries were for basic banking tasks: checking balances, transferring funds, ordering checkbooks, etc.
Outcome