Finding your idea — Hackcache

A two-hour method for turning a problem statement into one clear idea you can build in 48 hours, with a worked example and starting points for every theme.

What a good hackathon idea looks like

Before any method, know what you are aiming for. A strong idea for a 48-hour hackathon has all six of these:

One specific user. "A 52-year-old applying for an income certificate", not "citizens".

One real, painful moment that you heard about from real people, not one you imagined.

A result you can show working in 90 seconds.

A first version your team can build in 48 hours , with the tools you actually have.

Something new about it , even if the problem is old. A fresh angle counts.

Room to score on all five criteria: Innovation, Technical Execution, Design & UX, Impact and Presentation.

This guide is a two-hour method for getting there. Use it on Thursday 22 October, when the problem statements arrive. Better still, practise it once this week on one of the four themes, so the real run is your second time.

The two-hour method

Read the problem and list who is affected (15 minutes)

Talk to three real people (30 minutes)

Find the root cause with the "five whys" (15 minutes)

Turn it into a "How might we" question (10 minutes)

Sketch eight ideas in eight minutes (15 minutes)

Check what already exists (15 minutes)

Score your ideas and choose one (15 minutes)

Write your one-sentence pitch (5 minutes)

Each step below has a worked example: the team behind Kagaz, the example project used in our other guides, working on the Reimagine Government theme.

Step 1: Read the problem and list who is affected (15 minutes)

Read the problem statement aloud together, all of it. Then list every kind of person it touches: the people who use the service, the people who run it, and the people who help others through it.

Kagaz example. Theme: rethink how citizens interact with government services. People affected: first-time applicants, parents applying for their children, elderly people, office staff at the counter, and cyber café owners who fill in forms for others.

Circle the one or two groups you can actually reach today. You need to talk to them next.

Step 2: Talk to three real people (30 minutes)

This is the step most teams skip, and it is the one that makes your problem slide convincing. Three short conversations, by phone or WhatsApp, with people from the group you circled: a parent, a neighbour, a classmate, a colleague.

Follow the three rules of The Mom Test , Rob Fitzpatrick's well-known book on talking to users:

Talk about their life, not your idea. Once you describe your idea, people get polite and stop telling you about their problems.

Ask about specific things that happened, not what they would do. "Would you use an app for this?" gets a polite yes that means nothing. "Tell me about the last time..." gets the truth.

Talk less, listen more.

Questions that work:

"Tell me about the last time you had to [apply for / pay / sort out] something like this."

"What was the hardest part? What did you do about it?"

"What did it cost you: time, money, a day off work?"

"Have you tried anything to make it easier? What happened?"

A message you can send right now:

Hi Mama, quick question for a project I'm doing. When did you last have to get a certificate or document from a government office? What happened? Did anything go wrong?

Kagaz example. Suresh, 52: "I took half a day off work, queued for two hours, and was sent home for one missing photocopy." A classmate: went twice for her driving licence because the website's list didn't match what the counter asked for.

Write down their exact words. A real quote is worth more on your problem slide than any statistic.

Step 3: Find the root cause with the "five whys" (15 minutes)

The five whys comes from Toyota. Taiichi Ohno, the engineer behind Toyota's production system, described it this way: "By asking why five times and answering it each time, we can get to the real cause of the problem, which is often hidden behind more obvious symptoms." Take the most painful story from Step 2 and keep asking why.

Kagaz example.

Why was Suresh sent home? He didn't have a photocopy of his address proof.

Why not? He didn't know he needed one.

Why not? The list he found online was for a different case, and the office's own list was only on a notice board inside the office.

Why was it the wrong list? What you need depends on your situation: first application or renewal, adult or minor, whether your address has changed.

Why is that hard to find? Official information is written as general rules, not as a checklist for one person's situation.

Root cause: people cannot easily turn general rules into a checklist for their own case.

If your fifth answer is "the government should fix it", go back a step. You want a cause that something you build could help with.

Step 4: Turn it into a "How might we" question (10 minutes)

A "How might we" question turns the root cause into an invitation for ideas. IDEO, the design firm that popularised the method, uses it to frame problems so there is room for more than one answer. The Nielsen Norman Group's advice for writing good ones:

Start from what you found , not a generic goal.

Do not put the solution in the question.

Make it broad enough for several ideas, narrow enough to act on.

Say what you want to happen , not what you want to stop.

Kagaz example.

Too broad: "How might we improve government services?"

Solution already inside it: "How might we build a document checklist app?"

Just right: "How might we help first-time applicants feel sure they have every document before they leave home?"

Step 5: Sketch eight ideas in eight minutes (15 minutes)

This is the "Crazy 8s" exercise from Google's Design Sprint method. Each person folds a sheet of paper into eight boxes and sketches eight different ideas that answer your question, one minute per box. Rough boxes and arrows are fine. The point is to get past your first idea, because the first one is rarely the best.

Online, everyone sketches on paper, then photographs it and posts it in your team chat. Spend a few minutes each explaining your sketches, then everyone votes for their two favourite ideas.

Kagaz example. Ideas included: a checklist built from three questions about your case; a WhatsApp bot you ask "what do I need for X?"; a phone camera that checks you have each document; crowd-sourced reports of what each office actually asked for; a reminder the night before your appointment; a Hindi voice version for elders.

Step 6: Check what already exists (15 minutes)

Search before you commit, so you are not surprised when a judge mentions an existing app. Look in:

Google and the Play Store , using the words your user would use.

The official portal for that service. Governments often have more tools than people know about.

Devpost's project gallery at devpost.com/software, which holds thousands of past hackathon projects.

If something similar exists, that does not end the idea. It tells you where the gap is, and your pitch gets stronger for naming it.

Kagaz example. Passport Seva already has an official Document Advisor that tells passport applicants what to bring. So Kagaz does not compete with it on passports. Its gap is everything else: income, domicile and other certificates, whose rules differ from state to state, in plain language, in one place, shareable on WhatsApp with the family member doing the queueing.

Step 7: Score your ideas and choose one (15 minutes)

Take the two or three ideas with the most votes. Score each one from 1 to 5 on these questions, and add up the scores:

Can we build a working first version in 48 hours with our tools?

Can we show it working in 90 seconds?

Impact: would the people we talked to actually use it?

Innovation: is there something new in it, even if the problem is old?

Do we care about it? You will be working on it at 2 a.m.

Pick the highest score. On a tie, pick the one your team is more excited about.

Step 8: Write your one-sentence pitch (5 minutes)

For [one user] who [painful moment] , [your project] lets them [do one thing] so they can [get this result] .

Kagaz example. For first-time applicants who make wasted trips because they brought the wrong papers, Kagaz asks three questions and gives them the exact document checklist for their case, which they can share on WhatsApp.

If your sentence needs "and" twice, you have two projects. Pick one. From here, "Your working demo in 48 hours" takes over.

Starting points for each theme

These are facts and open resources to get you thinking. The real problem statements on 22 October will be more specific.

Reimagine Government

The scale of frustration is real. Central government ministries and departments received more than 51 lakh public grievances on CPGRAMS, the official grievance portal, between 2023 and 2025.

Accessibility is a requirement, not a nice-to-have. India's official guidelines for government websites and apps (GIGW 3.0) require them to meet the international accessibility standard WCAG 2.1 at level AA. Redesigning a real government page to meet it is a legitimate project.

Useful resources: data.gov.in , India's open government data portal, publishes government datasets you can download or call through APIs. API Setu lists government APIs for records such as driving licences, vehicle registration and CBSE marks. Most need approval before live use, so in a hackathon use them for realistic sample data and to show what a real version would connect to.

Questions to start from: How might we help an applicant know where their request is without calling the office? How might we make a government form readable for someone who does not read English well?

Social Impact

A large group is underserved. India's 2011 Census counted about 2.68 crore people with disabilities, 2.21% of the population, and 69% of them live in rural areas.

Look close to home. A local NGO, a school, a clinic, a residents' association: the people running them have problems they would describe to you in five minutes.

Questions to start from: How might we help a volunteer group fill its weekend shifts without a hundred WhatsApp messages? How might we help a visually impaired person use a service that is only on a website today?

Banking & Finance

Digital payments are everywhere. UPI processed a record 24.51 billion transactions in August 2026, according to NPCI data.

Understanding has not kept up. Only 27% of Indian adults were found to be financially literate in the national survey by NCFE, the National Centre for Financial Education (2019).

Fraud is common by number. In RBI data, card and internet frauds made up about two-thirds of reported bank fraud cases, though a small share of the money involved.

Keep it safe: never handle real money or real account details in a hackathon demo. Simulate payments with sample data, and say so.

Questions to start from: How might we help a first-time UPI user recognise a scam request before paying? How might we help someone on their first salary feel in control of the month?

Open Innovation

No theme, so start from your own frustration. Judges often say the strongest projects come from "a workflow you've personally found annoying". Then run the same eight steps: an idea you care about still has to pass the one-sentence test.

Using AI to help you ideate

AI assistants such as ChatGPT or Claude are good at widening your options and at arguing against you. They are not a substitute for Step 2: an AI's "users" are guesses, and judges can tell. Three prompts that help:

Map who is affected (Step 1):

Here is a hackathon problem statement: "[paste it]".

List every type of person affected by this problem,

including people who run or support the service.

For each, give one specific situation where it hurts them.

Argue against your idea (Step 7):

Here is our hackathon idea: "[your one-sentence pitch]".

Judges score on Innovation, Technical Execution, Design & UX,

Impact and Presentation. Give the five strongest reasons a

judge might score it low, and one change that fixes each.

Find what already exists (Step 6):

List existing apps, websites and government services that

solve "[your problem]" in India. For each, say what it does

and what it does not do.

AI tools sometimes invent apps that do not exist, so check every name it gives you with a real search before you mention it to a judge.

Ideas that look good but lose

"A platform for everyone." If you cannot name the one person it is for, judges cannot picture it either.

A solution looking for a problem. A Devpost judge put it this way: some teams are "just looking for a problem for their solution". Start from the problem.

A chatbot with no reason to be a chatbot. If a simple form or checklist does the job better, build that.

An idea that needs what you cannot get in 48 hours: live government or bank systems, real hospital data, a partnership. Simulate it with sample data instead.

Five features instead of one. Pick the one that makes the 90-second demo work.

Practise before 22 October

The four themes are already public. This week, take 60 minutes with your team, pick one theme, and run Steps 2 to 8 quickly. When the real problem statements arrive on Thursday 22 October, you will already know the method and can spend your two hours on the problem.

More Hackcache guides

The event

Hackcache

Open on CachedInfo