17 project ideas, one per goal.

Map the actual cost of being poor in your neighbourhood. Poor families often pay more per gram of rice, oil, or shampoo than middle-class ones do. Prove it, or disprove it, with numbers.

A list of 10 essentials, a weighing scale for loose goods, and time at your nearest kirana shop and the nearest supermarket. A phone camera helps for MRPs.

Record unit price (per 100 g or per mL) for the same 10 items at both. Repeat across three kiranas and two supermarkets so one shop's oddity does not skew you.

Add sachet prices vs. bottle prices. Sachets almost always cost more per mL, and sachets are what poor households buy.

Grow the same variety of methi in four conditions in identical pots: compost, kitchen scraps, no fertiliser, commercial NPK. Which "free" option beats the paid one?

Four identical pots, the same soil base, the same seeds, and one sunny balcony or patch. A kitchen scale to weigh yield.

Same watering schedule, same light. After three weeks, harvest and weigh yield per pot. Repeat twice so one pot's bad day does not decide the answer.

Add cost per gram of yield. A slightly lower yield at zero cost may beat a higher yield you had to pay for.

Measure PM2.5 outside your school gate at 8am, 1pm, and 5pm across 10 school days. Compare to WHO limits. Is the school run the most polluted part of your day?

A low-cost PM2.5 monitor (Atmotube, PurpleAir portable, or a school-owned unit). A fixed spot at the gate. A notebook.

Same spot, same three times, same duration (5 minutes). Log weather. Compare morning and afternoon peaks against the WHO 24-hour guideline of 15 µg/m³.

Compare gate readings to 200 m indoors from the gate. Quantify how much a closed classroom actually protects a child.

Same text, same students, printed vs. on a phone screen. Do students actually retain less from screens? A small, local answer to a question everyone argues about.

Twenty willing classmates, a 500-word passage, a 10-question comprehension quiz, and parental consent forms.

Randomly split the group. Same reading time, same quiz, same room. Swap the groups the following week with a different but comparable text so each student does both formats.

Add time-to-completion and self-reported focus. A small effect on retention may hide a large effect on attention.

Audit airtime by gender in one week of a specific news channel or podcast. Who speaks, for how long, on what topic? Count. Report. That is the whole project.

A stopwatch, a spreadsheet, and access to one week of recordings (YouTube, podcast feed, DTH replay).

For each segment, log speaker gender, seconds spoken, and topic category. Do a second pass on 20% of segments to check your own counting error rate.

Split "hard news" from "soft news" airtime. The overall gender split may hide a much sharper one inside politics or business.

Chlorine test-strip readings from your tap at 6am and 6pm across seven days, plus from a nearby RWA overhead tank. How much chlorine actually reaches the last house on the line?

A pack of free-chlorine test strips (0–5 ppm), permission from your RWA to sample the tank, and a fixed collection method.

Same tap, same time, same first-flush protocol. Compare against the IS 10500 residual-chlorine target of 0.2 ppm at the consumer end.

Sample from the first and the last house on your line. Distance from the tank almost always shows up in the numbers.

Log your household's actual electricity use hour-by-hour for a week with a plug-in meter. Which three appliances account for 50% of consumption? Almost certainly not the ones your family blames.

A single plug-in energy meter (about ₹500). A rotation schedule so it visits each appliance for at least 48 hours.

Meter the AC, the fridge, the water heater, the TV, the router, and any always-on plug loads in turn. Log kWh. Rank.

Model the effect of raising AC setpoint by 2°C. Small habit change, measurable saving, honest number.

Interview 10 informal-sector workers in your neighbourhood — auto drivers, delivery riders, domestic workers — about hours worked and take-home pay per hour. Anonymised, ethics-cleared. Publish the median.

A short structured questionnaire, consent forms in the participant's language, and an ISEF Form 4 for human participants.

Ten-minute interviews. No names, no addresses. Same questions, same order. Compute median hourly pay and compare against Haryana minimum wage.

Add unpaid waiting time. An auto driver's "hourly" wage looks very different once you count the hours parked at the stand.

Time how long it takes to walk from your house to the nearest five essential services — school, hospital, market, park, bank. Compare with a friend's route in a different colony. Whose neighbourhood is actually built for people?

A phone with a walking-route app, a stopwatch, and one friend in a different pin code willing to run the same protocol.

Walk at a normal pace, log time and distance. Repeat each route twice on different days. Compare totals and count how many crossings had no footpath.

Overlay accessibility — how many of your routes could a wheelchair or a stroller actually use? The gap is often larger than the time gap.

Compare the number and condition of public benches, water fountains, and shaded bus stops in two neighbourhoods in your city — one high-income, one low-income. Photograph. Count. Map.

A phone with GPS-tagged photo, a fixed area (say, 1 km² around a central point), and Google My Maps or similar.

Walk a defined grid in each neighbourhood. Log every bench, tap, and shaded stop with a condition rating (working / broken / missing). Total per km².

Cross-reference against municipal ward budgets, if the data is public. Argue — carefully — with the numbers, not the neighbourhoods.

Count and classify every piece of litter in one 10 m × 10 m patch of your local park, weekly for four weeks. What comes back? Where does it come from?

Gloves, a marked-out square (four sticks and twine work), a categories sheet (plastic, paper, food, cloth, metal, other), and a way to dispose of what you count.

Same day of the week, same time, same square. Count, categorise, remove. Log the weekly recurrence rate per category.

Trace the top category back — a repeating brand on wrappers, a specific vendor. The number becomes a lever.

Weigh your household's food waste every day for two weeks. Categorise: bought-and-forgot, over-cooked, plate-scraped. Which category dominates? What single change halves it?

A kitchen scale, a spare container, and a cooperative family that will not throw food into the main bin without weighing it first.

Weigh daily, log by category. In week two, try one deliberate intervention (a shopping list, smaller portions, a leftovers night). Measure the change.

Convert grams wasted to rupees wasted using average grocery prices. The economic frame lands harder than the environmental one.

Log the ambient temperature at 3pm on your rooftop and inside your top-floor room daily for four weeks. Compare to the same measurements from a shaded ground-floor courtyard. Quantify the urban-heat effect on your own home.

Two identical digital thermometers (or one, moved on a fixed rotation), a stopwatch, and a data sheet with weather-noted days.

Same time each day, same spots, same five-minute settle. Note cloud cover. Plot the daily gap between rooftop and courtyard.

Test one intervention on the rooftop — a white paint patch, a shade cloth, a potted plant cluster — and measure the temperature change under it.

If you have access to any water body: count visible plastic fragments per 500 mL at three points along it. If landlocked: sample water from the discharge drain of a fish market.

A 500 mL glass jar, a fine mesh (a fine tea strainer works), a magnifier, and a fixed counting protocol.

Same jar, same volume, same filter time. Photograph residue against grid paper. Count fragments above 1 mm. Three samples per site.

Classify fragments by likely source (fibre, film, hard). The mix tells you whether the load is upstream discharge or local litter.

Bird count. Same 20-minute window, same three points in your neighbourhood, twice a week for a month. Which species show up, when, and where?

Binoculars if you have them, an ID app (Merlin, eBird), and a notebook. Comfortable footwear.

Fixed points, fixed times (dawn works best). Log species, count, and behaviour. Upload to eBird so your data joins the global dataset.

Correlate your counts with tree cover, water availability, or human activity at each point. Bird diversity is almost always a proxy for something.

File three RTI applications on the same topic — say, school infrastructure spend — with three different local bodies. Log response time, quality, and completeness. How responsive is your local government, really?

The RTI Act 2005 basics, ₹10 per application, and a parent or teacher to co-sign if you are under 18. A dated log of every reply.

Same question, same wording, same date filed. Score each reply on time-to-response, information completeness, and whether re-appeal was needed.

Publish the anonymised comparison. The variance between institutions is usually the finding worth reporting.

Pick a local NGO working on one of the SDGs above. Volunteer for one month. Then interview them: what data do they wish they had? Design a small data-collection project that gives it to them.

One month of consistent hours, a parent's permission, and a willingness to do the boring work before the interesting work.

Volunteer first, ask second. After four weeks, sit with the person who runs the operations and list every data gap they mention. Pick one you could reasonably close.

Hand the NGO your dataset at the end, in a format they can actually use. That handover is the project.