← tanbedawy.workbook
Case Study · MENA Ride-Hailing Leader · Analytics Product 02 / 07

Marketplace Intelligence

The suite that answers one question from every angle: why do rides fail? Booking-funnel analytics, acceptance-rate decomposition, peak-pricing analysis, and anomaly matrices — with the metric glossary written into the dashboards.

Funnel AnalysisTableauSQL SegmentationPeak PricingAnomaly Detection
product scorecard — structural factsproduction
14
peak-multiplier brackets in the pricing analysis
9×11
duration × distance anomaly matrix
24h
hour-of-day heatmaps per peak bracket
2
parameterized views — daily & hourly

Note on numbers: no figure from inside a client dashboard is presented here as an achievement. The build is the story; published screenshots have absolute figures blurred.

01The Business Problem

In a two-sided marketplace, a failed ride has many possible causes: the customer never got an offer, got one and waited too long, cancelled before or after assignment, or the price didn't clear. Aggregate completion rates hide all of that. The commercial and ops teams needed to know where the booking funnel leaks, and how peak pricing and trip distance change captain behavior.

The goal: decompose every failed ride into a named, measurable failure mode — then track each mode daily, hourly, and by city.

02My Role

Sole analyst on the product: self-written SQL for every view, hourly and daily parameterized architectures, and the in-dashboard metric glossary that made the suite readable across departments. One person owned the questions, the queries, and the answers.

03What I Built

Bid-Ask Dashboard + Daily Report

A dual-tab navigation system covering the full booking funnel — app open → dropoff added → CCT selection → ride → completion — per city, with funnel drop-offs per stage, LO metrics with in-view definitions, and demand/supply performance blocks with WoW deltas.

Bid-Ask Rides Deep Analysis

Failed rides split by zero-ask vs ask-received, then by timed-out vs customer-cancelled — each with requested-vs-suggested price comparison, captain-offer status breakdowns, and median timing metrics (creation → first ask → assignment → cancellation).

Peak Analysis Workbook

Sessions, bookings, trips, and paid cancellations across 14 peak-multiplier brackets — as overview bars and as 8 hour-of-day heatmap matrices (bracket × 24 hours) with in-row relative highlighting.

User Segments & Anomaly Matrices

User segmentation (Power / High / Medium / Low × Flexi / Marketplace / Shared dependency) with WoW pivots; a duration × distance cross-tab (9 × 11 brackets) that surfaces anomalous rides; and zone-level booking contribution and C/R tables.

04Signature Build Details

A glossary inside the dashboard: every non-obvious term — ask received, timed-out, PAC, suggested price, captain offers — is defined in colored text inside the view itself. Cross-department readers never needed a translator.
Failure modes, not failure rates: "not completed" splits into zero-ask vs ask-received, then timed-out vs cancelled, each against price positioning — so pricing and supply teams saw their share of the leak separately.
Per-city funnel accounting: the daily report repeats the full funnel table per city with WoW deltas — one scroll compares Cairo, Alexandria, Damietta, and Tanta stage by stage.
Toggle architecture: Dates-aggregate vs Hours views share one layout; the same funnel logic reads by day for trends and by hour-of-day for operational patterns.
Timing medians for cancellations: creation-to-first-ask, first-ask-to-assignment, assignment-to-cancellation — the difference between "no supply" and "slow supply" quantified.

05KPIs Defined & Governed

Acceptance RateCNFPAC / POAC Impression RateConversion RateC/R R/SPeak IntensityAssignment Rate

06Decisions Enabled

Pricing teams saw acceptance behavior per peak bracket and distance band before touching multipliers.
Ops distinguished supply gaps (zero-ask) from matching friction (ask-received but timed out) — two different fixes.
The segment pivots showed which user tiers depended on which marketplace model, week over week.
The anomaly matrix flagged implausible duration-distance combinations for quality review.
SCREENSHOT SLOT
Bid-Ask Dashboard — funnel overview
(blur all absolute figures)assets/mkt-bidask-funnel.png
SCREENSHOT SLOT
Peak Analysis — hourly heatmaps (bracket × hour)
(blur all absolute figures)assets/mkt-peak-heatmaps.png

07Skills Demonstrated

Funnel decompositionMulti-dimensional segmentation Peak / surge analysisAnomaly matrices Parameterized view designSemantic layer / in-view definitions Advanced SQLTableau navigation systems
Mohamed Eltanbedawy · Business & Data Analyst — BI & Analytics
mohamed.tanbedawy@gmail.com · linkedin.com/in/tanbedawy