Project submission · Team Aakhri Mile
The last 2 kilometres: why Delhi's metro journeys fall apart after the exit gate.
Delhi Metro is fast, clean and on time. Then you tap out, and the journey turns into a negotiation: haggling for an e-rickshaw, guessing where the bus stops, walking on a broken footpath in the heat, or calling a cab you didn't plan to pay for. We mapped why this happens and designed two connected interventions, one digital and one physical.
The metro covers the long part of the trip well. The short part at both ends is slow, unpredictable, unsafe at night and unfair on price, especially for women, older people and anyone carrying a load.
It isn't a lack of options. There are e-rickshaws, buses, autos, cabs and live bus data. The problem is trust and information at the exit gate: nobody knows the fare, the wait or the route until they start bargaining.
Exit Board: a fair-fare, shared-ride board for the last mile, on your phone and on a physical bay at the station exit. Fixed fares by zone, a visible queue, and shared rides that fill faster.
Team
Riya (UI/UX) · Arjun (UI/UX) · Meher (Product) · Kabir (Product) illustrative
Scope
One busy metro interchange and the 2 km around it, evening peak
Duration
6 weeks · following the course process
Tools
Paper, FigJam, Kumu, Figma, Claude, Delhi Open Transit Data
Week 1 · Frame
What exactly is the problem?
We started with a drawing, then pinned the problem down in words, narrowed it to something we could study in six weeks, and dug for causes. No solutions this week; ideas went into our hunch log.
1.1 · Rich picture
What it looks like at the exit gate, 7 pm
Our first team drawing of the scene. Tap the blue dots to see what each part means. Crossed swords show conflict.
Selected
Tap a dot
Each dot is a person, a thing or a tension we saw or heard about.
illustrative field notes1.2 · Problem statement canvas
The problem, before any solution
We answered every box before allowing ourselves a single idea.
Metro riders lose time, money and sometimes their sense of safety in the short trip between the station and their real destination.
Around busy metro exits, within about 2 km. We focus on one interchange station in the evening peak.
Office commuters, students, women travelling alone, older people, people with disabilities, anyone carrying luggage.
WRI India's survey of 7,200 commuters [3] verified; Safetipin audits rating station surroundings from 1.5 to 3.3 out of 5 [8] verified; our own exit observation illustrative.
Evening peak, after dark, in rain and in the May–June heat. Also on days when e-rickshaws are cleared from the exit.
Metro riders keep growing year on year [1]. E-rickshaws grew fast and mostly unplanned [5]. Live bus data arrived in 2018, but people still struggle to find their bus [6].
Lost time, higher costs, people dropping the metro for bikes and cars, more traffic and pollution, unsafe walks at night.
E-rickshaw drivers, traffic police, shopkeepers at the exit, residents of the lanes around the station.
How fares are actually set. Why shared rides don't fill. Whether women avoid certain exits after dark.
Building new metro lines. City-wide bus reform. The long middle part of the trip.
"An app for e-rickshaws." "More feeder buses." Parked until Week 4.
We started thinking this was a traffic problem. By the end of the week it looked more like a trust problem.
Delhi Metro riders face a slow, unpredictable and unfair last two kilometres after they exit the station. Because fares, waits and routes are invisible until people start bargaining, many rely on costlier or less safe options, or give up the metro altogether.
1.3 · Narrowing the topic
From "Delhi traffic" to something we can actually study
Big topics feel important but are impossible to map well in six weeks. We narrowed four times.
Level
Topic
Why we moved on
Status
Delhi traffic and pollution
Far too big. The real levers are regional and political.
droppedPublic transport in Delhi
Still huge: buses, metro, policy, funding.
droppedGetting to and from metro stations
Better. Clear users, clear places. Still many stations.
narrowedThe last 2 km from one busy interchange, evening peak
Small enough to observe, talk to people and test ideas. Big enough to include the causes.
our scope1.4 · Problem tree with one 5 Whys branch
Causes below, consequences above
Tap any box to see our reasoning and how sure we are. Then follow one root down with the 5 Whys.
Selected
Tap a box
Red boxes are consequences, the black box is the core problem, blue boxes are direct causes, yellow boxes are deeper causes.
5 Whys on one branch: "Fares are unclear and negotiated"
Careful: the fifth answer is not "the" root cause. It's one strong branch. Other branches (safety, footpaths, heat) have their own whys. Week 3 turns this tree into loops, because causes in a real system feed back on each other.
1.5 · Framing question
The one question our whole project answers
"How can we build an app for e-rickshaw booking at metro stations?"
A solution in disguise. It assumes the problem is booking, that an app is the answer, and that e-rickshaws are the only mode.
Problem, not solution ✔ · Causes and consequences separated ✔ · More than one cause explored ✔ · Evidence vs hypothesis labelled ✔
Week 2 · Research & people
Who is in this system, and what do they see?
We read the research, pulled real data, stood at the exit for two evenings and talked to people on every side, not only commuters.
2.1 · Research plan
What we needed to find out, and how
| Question | Method | Who or what | Owner |
|---|---|---|---|
| How do people actually get from the exit to their destination? | Observation, 2 evenings × 1 hour; tally by mode | Main exit, 6–8 pm | Kabir |
| How are last-mile fares set? | Short interviews | 5 e-rickshaw drivers, 2 auto drivers | Meher |
| What do riders find hardest? | Intercept interviews, 5 minutes each | 10 commuters of mixed age and gender | Riya |
| What do existing apps show, and what don't they? | App teardown | Google Maps, One Delhi, Uber/Rapido metro tickets, DMRC app | Arjun |
| What does the research already say? | Desk research with AI support; every claim checked at source | WRI India, Safetipin, DMRC, Open Transit Data, news | All |
| What rules shape the exit area? | Expert interview | 1 traffic constable, 1 transport researcher | Meher |
2.2 · Evidence register
What the research already tells us
Real sources, each linked. We wrote down what each one supports and how far we can trust it.
| Finding | Source | What it supports | Trust |
|---|---|---|---|
| Walking and shared services make up more than three-quarters of last-mile trips; shared autos are the most popular shared option. | WRI India & Toyota Mobility Foundation, 7,200-commuter survey in Delhi, Bengaluru and Nagpur [3] | The last mile runs on walking and informal shared rides | high |
| People will spend up to 20 minutes, including waiting, to reach a station. That defines the catchment. | Same study [3] | Waiting time matters as much as distance | high · 3-city figure |
| Women travel shorter last-mile distances but pay more, and are more averse to waiting, so they choose faster, costlier options. | Same study [3] [4] | The last mile is unequal by gender | high |
| 2,04,131 e-rickshaws registered in Delhi (Jan 2026); the real number is believed to be higher. The government is considering a registration window. | Delhi govt data via PTI [5] | A huge, partly informal fleet | medium · news report of official data |
| Delhi publishes live bus GPS data every 10 seconds, since 2018, and apps use it. Yet commuters still struggle to find their bus. | Open Transit Data portal [6]; Swarajya ground report [7] | Data exists; the information still doesn't reach people at the moment they need it | medium |
| Safety audits of 17 Yellow Line station areas (500 m, 5–10 pm) scored from 1.5 to 3.3 out of 5; dark spots in Delhi fell from 7,500 (2015) to 2,700 (2018). | Safetipin reports [8] [9] | Station surroundings feel unsafe, and it can improve | medium · self-published, perception-based |
| The quality of the route matters more for walking than whether a footpath simply exists. | Study of Delhi metro stations, Urban Planning journal [10] | Walking needs good routes, not only infrastructure | high · peer-reviewed |
| Vehicles are among Delhi's largest PM2.5 sources: 20% in IIT Kanpur's 2016 study; a range of 9–38% across studies. | WRI India review [11]; Down To Earth [12] | Last-mile failure that pushes people into cars and bikes also affects air quality | medium · studies differ |
| Uber and Rapido started selling Delhi Metro QR tickets through ONDC in May 2025. | Inc42 [13] | Metro and last-mile apps are starting to connect | high |
| DMRC has run electric feeder buses since 2019; the Delhi government took over 100 of them in 2022. | Deccan Herald [14]; Business Standard [15] | Formal feeder services exist, but at small scale | medium |
2.3 · Interviews & observation
What people told us
Ten commuters, five drivers, one constable, one shopkeeper. Here are six voices that changed our thinking. illustrative · written to show the format
Exit tally · 2 evenings, 6–8 pm illustrative
Walked: about half. E-rickshaw (shared or solo): about a third. Bus: under a tenth. Auto or cab: the rest. Average wait we timed for a shared e-rickshaw to fill: 6–9 minutes, and longer after 8 pm.
2.4 · Stakeholder map
Everyone in the last 2 km, by how close they are
Tap a dot to see what each stakeholder wants and what they carry.
Selected
Tap a dot
Inner ring: directly involved. Outer rings: affected from further away.
2.5 · Power–interest grid
Who can change things, and who cares the most?
What this tells us
The people who care most (commuters, drivers) have the least power. The people with the most power (transport department, DMRC, platforms) don't feel the problem every day.
Design implication: our intervention must give commuters and drivers a shared voice, and make the problem visible to the powerful.
2.6 · Journey map with a system layer
Priya's journey home, and what happens behind it
The top rows are a normal UX journey map. The yellow rows show the system underneath: other people, rules, money and materials.
| Tap out | At the exit | Choosing a ride | Waiting | The ride | Last 200 m | |
|---|---|---|---|---|---|---|
| Priya does | Taps her card, checks the time | Pushes through the crowd at the gate | Asks three drivers the fare | Waits for the shared ride to fill | Squeezed in with 4 others | Walks a dark lane home |
| Feels | Calm | Overwhelmed | Annoyed, distrustful | Impatient, watched | Okay, relieved | Anxious |
| Screens | Metro card / QR | None | None (no app shows the fare) | Maps shows a bus "in 6 min" that doesn't come | UPI payment | Shares live location with mum |
| Other people | DMRC staff | Shopkeepers, police, 30+ drivers calling out | Drivers compete for the same passengers | Driver loses money every empty minute | Driver, other riders | Residents, guards |
| Rules & money | Fixed metro fare | Police may clear rickshaws from the exit | No agreed fare; women often quoted more | Driver only earns once full | Cash or UPI, no receipt | Streetlights maintained by civic body |
| Materials & energy | — | Crowded, narrow footpath | Battery rickshaws charged overnight | Idling in traffic | Overloaded vehicle | Broken footpath, puddles |
illustrative persona based on the pattern in the WRI India findings on women's last-mile costs [3] and Safetipin audits [8].
Week 3 · Find the loops
What keeps the problem in place?
We moved from a tree to loops. Causes don't sit in a line; they feed back on each other. This week changed our project more than any other.
3.1 · Iceberg model
A fight over ₹20 is only the tip
Tap each level.
3.2 · Causal loop diagram
Four loops that keep the last mile broken
Built in Kumu, rebuilt here. Tap a loop button to light it up, or tap any box to see its connections. + means "moves in the same direction"; − means "moves in the opposite direction".
Selected
Pick a loop
Each loop is a story. Read it out loud from any box and you should end up back where you started.
3.3 · System archetypes
Two patterns we recognised
Clearing rickshaws from the exit
When the exit gets jammed, rickshaws are cleared or impounded. The exit is clear for a few days. But riders still need rides, so drivers come back and crowd the same spot, often with less order than before.
The fix treats the symptom (crowding) and ignores the need (a planned place to pick up).
"Just drive to the station" or "just take a cab"
When the last mile fails, each person solves it alone: a bike, a cab, a lift from family. It works for them, so nobody pushes for a shared fix. Meanwhile, the shared system gets weaker.
Individual workarounds take the pressure off fixing the system.
Week 4 · Leverage & ideas
Where could a small change shift the whole system?
We looked at what already exists, found the gaps, placed our ideas on the leverage ladder, and only then wrote How Might We questions.
4.1 · Solutions landscape
A lot has been tried already
| Who | What | What works | What's missing |
|---|---|---|---|
| Delhi govt + IIIT-Delhi | Open Transit Data: live bus GPS for all apps, since 2018 [6] | The data is public and good | It doesn't reach people at the exit gate, and it doesn't cover e-rickshaws |
| DMRC / Delhi govt | Electric feeder buses from stations [14] [15] | Formal, safe, cheap | Small scale; fixed routes don't match scattered destinations |
| Uber, Rapido via ONDC | Metro QR tickets inside ride apps [13] | One app for metro and ride | Built for cabs and bikes, not shared low-cost rides; most last-mile trips stay outside |
| Delhi govt | Free bus travel for women since 2019, now with Pink Saheli smart cards [16] | Cost barrier removed for women on buses | Doesn't help where buses don't go |
| Delhi transport dept | E-rickshaw registration drives and planned policy [5] | Brings the fleet into the formal system | Focused on control, not on service quality or fair fares |
| Safetipin + civic bodies | Night safety audits and lighting fixes [9] | Measured drop in dark spots | One-off audits; riders can't see or act on the data |
4.2 · Impact gaps canvas
Between the problem and the solutions, three gaps
No one makes last-mile fares visible or fair. Every trip starts with a negotiation.
Good data exists, but not at the one place and moment it's needed: just after you tap out.
No designed, lit, shaded, orderly space where rides and riders meet, so the exit becomes a fight every evening.
4.3 · Leverage ladder
We placed every idea on a rung
Ideas on the bottom rung are easy but weak. We chose ideas on rungs 2 and 3, marked in yellow.
If fares, waits and seats are visible and fixed by zone, both at the exit and on the phone, then haggling stops and shared rides fill faster, because riders trust the price and drivers can see the demand. This breaks loop R2 and weakens R1.
hypothesis · tested in Week 54.4 · How Might We
Only now did we let ourselves ask "how might we"
How might we make the last-mile fare clear and fair before anyone has to bargain?
How might we help riders and drivers find each other faster, so shared rides fill without long waits?
How might we make the exit a calm, lit, orderly place to wait, especially for women after dark?
4.5 · Our concept
Exit Board: one system, two parts
The UI/UX and product parts only work together. The screen without the bay is just another app; the bay without the screen is just a shelter.
In the metro app, before you tap out
- Pick your destination zone: see fixed fares, wait time and seats left for shared e-rickshaws, plus the next real bus.
- Join a shared ride and get a bay number.
- Women can choose a women-only shared ride where available.
- Drivers get a simple screen: riders waiting per zone, fair fare, queue position.
Builds on Open Transit Data [6] and the ONDC ticketing link [13].
The Exit Bay
- A modular bay just outside the exit: shade, light, seating, numbered zone posts.
- A large fare board by zone, readable without a phone.
- A queue lane so rickshaws line up instead of crowding the gate.
- Made from standard steel sections; fits footpaths from 3 m wide.
Responds to the Safetipin findings on lighting and visibility [8].
| Concept sheet | |
|---|---|
| Leverage point | Rung 3 (fixed fares, queue rule) + rung 2 (fare and wait visible to both sides) |
| Loops it changes | Breaks R2 (haggling and distrust); slows R1 (car spiral); stabilises B1 (exit crowding); helps R3 (night safety) |
| Whose behaviour changes | Riders join shared rides more; drivers queue instead of crowding; police manage a bay instead of clearing |
| What stays the same | Drivers keep their vehicles and their income model; no new fleet is needed |
| What we prototype | A clickable phone flow (below) and a 1:20 cardboard model of the bay |
4.6 · Theory of change
From a board at one exit to a city-wide shift
Assumptions to test are marked under each arrow. If an assumption fails in testing, the chain breaks there.
Week 5 · Prototype & test
Does it work, and what else might it change?
We built the cheapest versions that could test our hypothesis, tried them with real people, and looked hard for side effects.
5.1 · Prototype
Try the phone flow
This is the clickable version we tested with commuters. Tap through it.
1:20 cardboard model; zone posts A–D; fare board readable from 10 m. illustrative
5.2 · Unintended-consequences wheel
What else might Exit Board change?
Inner ring: direct effects. Outer ring: effects of those effects. Tap any circle.
Selected
Tap a circle
Green = good effect, red = harmful effect.
5.3 · Premortem
It's 2030 and Exit Board failed. Why?
| Why it failed | Likely? | Early warning | What we'd do |
|---|---|---|---|
| Drivers refused fixed fares because they felt fares were set without them | High | Few drivers sign up in month 1 | Set fares together with driver groups; review every 3 months |
| Only people with smartphones benefited | Medium | Older riders bypass the bay | The physical fare board and a staffed help point work without a phone |
| Police went back to clearing the exit | Medium | Bay blocked by enforcement drives | Make the bay an official, permitted pickup point with the civic body |
| Fares rose for the shortest trips | Medium | Complaints from Zone A riders | Minimum fare kept low; zones based on real trip data |
| It worked at one station but never scaled | High | No second station after a year | Design the bay as a standard kit; publish the fare method openly |
5.4 · Stakeholder tests
What happened when we tested it
illustrative results · to show the format
| Who | What we tested | What happened | What changed on our map |
|---|---|---|---|
| 6 commuters (3 women) | Phone flow, think-aloud | All understood zones in under a minute. Women valued "seats left" and "women-only ride" most. | Added a safety variable to R2: trust depends on safety as well as price |
| 4 e-rickshaw drivers | Fare board + driver screen on paper | Liked the queue; two worried zone fares were too low for night trips. | Added "night fare" and a new link: driver income → willingness to join |
| 1 traffic constable | Bay model | "If it's official, we can manage it. If not, we'll still have to clear it." | Moved "permission from the civic body" to the top of the theory of change |
| 1 shopkeeper | Bay model | Worried the bay would block his shopfront. | Bay redesigned to sit on the road edge; new stakeholder added to the map |
Week 6 · Present
What we learned, and how our thinking changed
Our final package: the updated map, how to measure success, an honest AI log and every source we used.
6.1 · Map versions
Three versions of our understanding
"It's a traffic problem"
Too many vehicles at the exit. Solution in our heads: more feeder buses.
"It's a trust problem"
Hidden fares and waits drive haggling and push people to cars. Loops R1 and R2 at the centre.
"It's a shared-income problem"
Nothing changes unless drivers earn fairly too. Driver income now sits at the centre, linked to trust, safety and shared rides.
6.2 · Key performance questions
How we'd know if the system is shifting
| Question | Signal we'd watch |
|---|---|
| Has haggling stopped? | Share of trips started at the board fare |
| Do shared rides fill faster? | Average wait from tap-out to departure |
| Are drivers earning at least as much? | Driver earnings per hour, before and after |
| Do women feel safer at the exit after dark? | Repeat Safetipin-style audit; share of women using the bay after 8 pm |
| Are fewer people giving up on the metro? | Station entries over 6 months, compared with similar stations |
6.3 · AI use log
How we used AI, honestly
| Week | Tool | Used for | Kept / changed | How we checked |
|---|---|---|---|---|
| 1 | Claude | Challenged our first problem statement | It spotted the "solution in disguise"; we rewrote it | Faculty review |
| 2 | ChatGPT | Found reports on last-mile access in Indian cities | Kept 4 of 9 suggestions; 2 links didn't exist | Opened every source |
| 2 | Claude | Clustered anonymised interview notes | Kept 3 themes, merged 2, rejected 1 | Checked each theme against our notes |
| 3 | Claude | Asked what loops we'd missed | Added the night-safety loop R3 | Confirmed with the Safetipin reports |
| 5 | Claude Code | Turned the Figma flow into a clickable prototype | Rewrote the copy in our own words | Tested with 6 commuters |
illustrative log following the SoD Authorship and Acknowledgement Policy.
6.4 · References
Sources
Every number in this submission links back to one of these. Figures were checked on 22 October 2026; check again before you quote them.