# Portfolio Summary — NYC Taxi Routes --- ## Project ``` name: NYC Taxi Routes slug: nyc-taxi-routes type: DA stage: Phase 5 — Analysis + Insights complete, all 9 business questions answered target: JFK departure share (share of trips departing from JFK airport) stack: Python · Pandas · Scikit-learn · Matplotlib/Seaborn · Plotly · Folium · wgnd-toolkit · Jupyter period: 300.000 Taxifahrten, NYC 2016 rows: 293,369 (cleaned, from 300,000 raw) notebooks: 5 findings: 6 dashboard: — (not deployed, DA project, no dashboard planned) ``` --- ## Storyline ``` thesis: JFK trips are a small (1.93%) but disproportionately valuable (~4x revenue) niche with strong weekday and hourly seasonality — a flat fleet allocation would systematically under-serve Mondays and early mornings while over-serving Saturdays. hook: A JFK trip earns roughly 4x the revenue of an average trip, yet only 1.93% of trips actually depart from there. proof: Route classification (JFK/NYC/Other) → revenue-share comparison → weekday breakdown → hourly breakdown — each step narrows the "how many taxis" question into concrete, time-sliced allocation factors. so_what: Fleet allocation should follow a weekday/hour-weighted schedule, not a flat share — captures the JFK revenue premium without stranding taxis during off-peak windows. ``` --- ## Problem ``` kpi_name: JFK departure share kpi_ist: 1.93% (5,649 of 293,369 trips) kpi_soll: n/a — descriptive fleet-sizing question, not a target-vs-gap KPI kpi_gap: n/a problem_statement: | A NYC taxi fleet operator wants to know how many taxis to station at JFK airport. JFK sits far from Manhattan's core business district, so taxis dropping passengers there take long to reposition — but JFK-bound passengers travel longer, more lucrative routes. The 2016 Yellow Taxi dataset (300k trips) is used to quantify JFK's actual share of business and its time patterns, to inform a fleet allocation that isn't just a flat percentage. ``` --- ## Key Findings ### F1 — JFK Departure Share ``` finding: Only a small fraction of all trips actually depart from JFK. number: 1.93% (5,649 of 293,369 cleaned trips) source: 03_analysis.ipynb ``` ### F2 — JFK Revenue Premium ``` finding: Trips touching JFK (pickup or dropoff) generate far more revenue per trip than average. number: 7,613 trips (2.6%) generate 9.99% of fare revenue — ~4x an average trip source: 03_analysis.ipynb ``` ### F3 — Weekday Skew ``` finding: JFK demand is not flat across the week — Monday leads, Saturday lags. number: Monday 2.60% JFK share (highest) vs. Saturday 1.44% (lowest); Monday is 1.35x overrepresented among JFK trips, Saturday 0.75x underrepresented source: 03_analysis.ipynb ``` ### F4 — Hourly Pattern ``` finding: JFK's peak hour comes earlier than the network-wide peak, and early-morning JFK share runs far above general traffic at that hour. number: JFK peak 17:00 vs. network peak 18:00; 5–6am JFK share ~2x general traffic share source: 03_analysis.ipynb ``` ### F5 — Route Dominance ``` finding: The overwhelming majority of trips and distance stay entirely within the NYC metro area — JFK is a small, distinct niche, not a core volume driver. number: 97.21% of trips, 83.77% of distance are NYC→NYC source: 03_analysis.ipynb ``` ### F6 — Data Cleaning Impact ``` finding: Cleaning removed a small but meaningful share of rows via technical-error and business-scope masks (duplicates, invalid geo/fare/distance/passenger values). number: 6,631 rows removed (2.2% of 300,000 raw rows) source: 02_preparation.ipynb ``` --- ## Model Results n/a — DA project, no modeling in scope. --- ## Figures ```yaml geo: - img/pickup_locations_map.png # Pickup/dropoff scatter, classified JFK/NYC/Other temporal: - img/jfk_share_by_weekday.png # JFK departure share per weekday, Monday highest - img/weekday_distribution_overall_vs_jfk.png # Weekday distribution, all trips vs. JFK - img/hourly_distribution_overall_vs_jfk.png # Hourly distribution, all trips vs. JFK routes: - img/route_distribution_log.png # Trip count per route type, log scale ``` --- ## Recommendations ``` r1: title: Monday uplift detail: Station ~35% more JFK taxis on Mondays than the weekly average — Monday accounts for 20.61% of all JFK trips vs. only 15.24% of all trips (1.35x skew). r2: title: Saturday reduction detail: Reduce JFK presence ~25% on Saturdays vs. the weekly average — Saturday accounts for only 11.38% of JFK trips vs. 15.23% of all trips (0.75x skew). r3: title: Early-shift coverage detail: Double early-morning (5–7am) JFK coverage relative to general demand at that hour — JFK share is ~2x the overall traffic share during those hours (likely early flight departures). r4: title: Shift the evening peak detail: JFK's own peak hour is 17:00, an hour before the network-wide peak of 18:00 — schedule the JFK evening shift accordingly instead of following general traffic patterns. r5: title: Prioritize JFK return fares detail: Favor NYC→JFK return trips over idle repositioning — JFK-touching trips earn roughly 4x an average trip's revenue. ``` --- ## Status ``` generated_by: /project-case story generated_at: 2026-07-11 summary_version: 1 portfolio_check: ✅ passed report_html: ❌ pending slides_html: ❌ pending dashboard: ❌ not deployed — DA project, no dashboard planned ```