Offline Reading Level Assessment App for Teaching at the Right Level Programmes

The problem: teaching to the textbook, not to the child

Large household surveys in Pakistan, such as ASER Pakistan, regularly find that many children in primary school cannot read a short story at the level expected for their grade. Teachers follow the textbook for the grade, so children who are behind fall further behind. Programmes that group children by their actual level for part of the day and teach them from there have shown strong results, but they depend on assessing every child quickly and consistently, then reassessing every few weeks.

On paper, assessments get lost, assessors judge differently, grouping takes hours and nobody can see progress across a district.

Consistency is the quiet problem. Two volunteers can listen to the same child and record different levels, especially at the boundary between reading a paragraph and reading a story. When that happens, children land in the wrong group and progress figures become unreliable. A tool that measures and reduces that disagreement makes the whole programme more trustworthy to teachers, parents and funders.

What already exists and the gap

• ASER Pakistan uses a simple oral reading tool, on one sheet of paper, with tasks for letters, words, a paragraph and a story, to measure children's basic reading in large surveys.
• General data collection apps such as ODK and KoboToolbox can record assessment results if a programme builds the forms itself.
• Spreadsheets are how many NGOs and schools group children today.

The gap is a purpose-built tool for running a level-based programme day to day: fast assessment, automatic grouping, progress tracking and, importantly, a way to check that different assessors are judging consistently.

Inspiration

Mainstreaming an Effective Intervention: Evidence from Randomized Evaluations of "Teaching at the Right Level" in India (Banerjee, Banerji, Berry, Duflo and colleagues, NBER Working Paper 22746, 2016) evaluates Pratham's approach of grouping children by learning level, assessed with a simple oral tool, and teaching them at that level. The papers stress that implementation quality matters as much as the idea. This project builds the tool that makes the assessment and grouping part reliable at scale.

What you build

• Assessors (teachers or volunteers) open a child's profile and run the assessment on the phone, starting at a level and moving up or down by the standard rules. The app records the highest level reached. With consent, it can record the child reading for moderation.
• The app groups the class or centre by level and produces lists for each group.
• After a few weeks, assessors reassess and the app shows each child's movement and each group's progress.
• Programme staff see a dashboard of level distributions across centres once phones sync.

The technically interesting part

• Equivalent forms. Repeated tests need different texts of the same difficulty. Build a card bank with teachers and track which card each child has seen.
• Measuring assessor consistency. Have two assessors rate the same child independently for a sample, then compute agreement (for example, Cohen's kappa) per assessor and flag those who differ systematically. This is real measurement work, not just an app.
• Audio moderation without AI. Store short recordings compressed on the device, sync them when possible and let a senior teacher review a random sample. Design storage and deletion rules carefully.
• Offline-first sync. Centres may be offline for weeks. Use stable identifiers and append-only assessment records so sync never loses or overwrites results.

How to evaluate it

• Speed: time per assessment compared with the paper process in a trial.
• Reliability: agreement between assessors with and without a short in-app training and calibration step.
• Grouping: time saved in producing level groups compared with manual grouping.
• Field trial: with permission, a small trial with a school or NGO over two assessment rounds; collect teacher feedback.

Suggested plan

• Weeks 1 to 3: study the ASER tool and TaRL guides; meet a school or NGO; design assessment cards with teachers.
• Weeks 4 to 7: assessment flow, card bank and child profiles.
• Weeks 8 to 10: grouping, progress tracking and sync.
• Weeks 11 to 12: reliability module and dashboard.
• Weeks 13 to 15: field trial, report and demo.

Path to a product

NGOs, education departments and low-cost private school chains running remedial programmes need dependable tools, and many are supported by funders who value good data. An open source app with paid setup and hosting for programmes is realistic. The project also shows mobile development, measurement and fieldwork, which are valued in education technology roles.

Scope warning

Use existing, tested assessment tools as the model and design cards with teachers; do not invent your own reading standard. Speech recognition is out of scope. Children's data and recordings need parental consent, minimal storage and clear deletion rules, agreed with the partner before any trial.

Learn what you need on Sefism

• Mobile development roadmap: offline apps, local databases and sync.
• Flutter Documentation: the app framework.
• UI/UX design roadmap: designing for teachers and volunteers in a busy classroom.
• OpenIntro Statistics: agreement measures and simple comparisons.
• Mobile development cheat sheet: background sync and testing on real devices.

Get help with this idea

The hard parts are equivalent assessment cards and measuring assessor consistency properly. Tauseef can help you find a partner, plan the trial and keep children's data safe. If you are new here, create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session