Here is a number that should stop any education program designer in their tracks. Before the pandemic, roughly 56% of children in India could not read and understand a simple, age-appropriate text by the age of 10, according to the World Bank — a condition it calls "learning poverty." Globally, after COVID-19 school closures, that figure reached 70% in low- and middle-income countries.
The most recent ASER 2024 survey shows real, hard-won recovery — the share of Grade 3 children who can read a Grade 2-level text rose from 16.3% in 2022 to 23.4% in 2024 — but read that the other way: more than three in four Grade 3 children still cannot. This is the gap India's NIPUN Bharat mission is racing to close, with a national target of foundational literacy and numeracy for every child by the end of Grade 3.
So how does an NGO actually design a program to move that number in one district — and design it well enough to fund? This is an end-to-end case study of exactly that, using the Logical Framework Approach. The organization below is a composite, drawn from the many teams who do this work; the data, policy, and evidence are real.
The brief, and the trap
Picture Saksham Shiksha Trust (a composite), a mid-sized NGO invited to submit a two-year proposal to improve early-grade reading across 60 government primary schools in one district. The team's first instinct — a very human one — was to reach for a solution: "We'll run reading camps and donate storybooks." Good energy, wrong starting point. That is the classic logframe trap: committing to activities before you have defined the problem or checked whether your chosen response is the one the evidence supports.
The LFA insists on a different order. Its analysis stage — stakeholders, problems, objectives, strategy — comes first. Only then does the matrix get written. Here is how Saksham worked through it.
Stage 1 — Stakeholder analysis
Before touching the problem, the team mapped who is involved and where power and interest sit. The point is practical: it tells you whom to partner with closely, whom to keep informed, and whose buy-in the whole design depends on. In a government-school program, the District Education Officer's support is not a nice-to-have — it is a precondition.
The insight that changed the design came from this stage: teachers sat in the "high interest, moderate influence" band, but they were also the single point of failure. Whatever the program did, it would live or die on whether ordinary government teachers could and would deliver it.
Stage 2 — The problem tree
Next, the team built a problem tree: the core problem in the middle, its causes below, its effects above. This is where "we need reading camps" gets interrogated into "why can't these children read?"
The tree exposed the flaw in the first instinct. Reading camps and storybooks touch the last cause — support and exposure — but do almost nothing about the first and most powerful one: instruction aimed at the average of the grade leaves the children who are behind permanently behind. A program that ignored that root cause would be busy and beloved and ineffective.
Stage 3 — The objective tree
The objective tree is the problem tree flipped into positives. "Teaching is pegged to the grade, not the child's level" becomes "children are taught at their actual level." "No routine assessment" becomes "foundational skills are assessed regularly and used to group children." Read upward, these objectives form the results chain the program will commit to. The exercise is almost mechanical — and that is the point. A rigorous problem analysis hands you your objectives; you are not inventing them from hope.
Stage 4 — Strategy selection (and why evidence matters here)
Several branches of the objective tree could be tackled. This is the decision point where good NGOs consult the evidence rather than their preferences. For foundational learning, the most rigorously tested approach is Teaching at the Right Level (TaRL), developed by Pratham: children are grouped by their current reading level rather than by grade for part of the day, and taught with focused, activity-based methods until they progress.
The reason to choose it is not fashion. As J-PAL documents, six randomized evaluations across seven Indian states found TaRL "consistently effective when implemented systematically," producing some of the largest effect sizes in the education literature — and among the most cost-effective ways to raise learning that J-PAL has measured. It has reached more than 80 million children. For a team accountable to a funder, "we chose the approach with the strongest causal evidence" is a far stronger sentence than "we chose the approach we like."
Saksham selected TaRL, delivered by government teachers with mentoring support, aligned to NIPUN Bharat. Now — and only now — the matrix could be written.
Stage 5 — The logframe matrix
The results chain, from the ambition at the top to the work at the bottom:
Turned into the classic four-column matrix, with each level's indicators, means of verification, and assumptions:
| Level | Intervention logic | Indicators (SMARTI) | Means of verification | Assumptions |
|---|---|---|---|---|
| Goal | Children complete primary school as confident readers | Learning-poverty rate among Grade 5 children falls over time | State assessment; periodic ASER-style survey | FLN remains a policy priority; no major shock to schooling |
| Purpose | Grade 3 children read a Grade 2 text with comprehension | Share reading a Grade 2 text rises from 22% to 50% across 60 schools by end of Year 2 | Independent endline vs. baseline reading assessment (ASER-style tool) | Teachers apply TaRL; pupil attendance stays stable |
| Outputs | Teachers deliver daily TaRL; children grouped by level; assessment system running | 90% of 180 teachers deliver ≥30 min daily TaRL; all 60 schools regroup children each term | Mentor observation logs; termly assessment records | Teachers stay in post; materials arrive on time |
| Activities | Train, equip, assess, mentor | 180 teachers trained; materials in 60 schools; 3 assessment rounds; monthly mentoring visits | Training attendance; delivery notes; visit reports | Funds released on schedule; DEO permits school access |
Making the indicators SMARTI
Notice the purpose-level indicator. It is not "reading improves." It states a target group (Grade 3 children in 60 named schools), a baseline and target (22% to 50%), a deadline (end of Year 2), and — crucially — a means of verification that an outsider could trust (an independent assessment using a recognised tool, not teachers grading their own pupils). That is the SMARTI standard — Specific, Measurable, Available at acceptable cost, Relevant, Time-bound, and Independently verifiable. An indicator you cannot verify is only an intention.
The assumptions column: where the real risk lives
The blandest-looking column is the most important. Two assumptions here are "killing factors" — if they fail, the whole program fails, not just a part of it:
- Teacher retention. TaRL lives in the teacher's hands. High transfer or turnover would hollow it out. The design responded by building a mentoring layer and training block-level coordinators, so capability does not walk out the door with one teacher.
- Teachers actually applying the method. Training does not equal practice. The output indicator therefore measures delivery observed in classrooms, not just attendance at a workshop.
Naming these is not pessimism. In results-based management it is the discipline that separates a fundable plan from a wish — and it is exactly what a serious evaluator looks for.
Where design thinking and AI accelerated the work
None of this required a specialist economist. It required structure and a few hard questions asked at the right moments — which is precisely where an AI methodology coach earns its place. Working in LFA Studio, the team could restate the core problem several ways and watch each version reshape the objective tree in minutes, lowering the cost of a "prototype" the way design thinking prescribes. The AI checked the vertical logic tirelessly ("if teachers are trained but you have no assumption about whether they apply it, your output logic has a gap"), tested each indicator against SMARTI, and prompted for the retention risk nobody had written down. It did not supply the insight — a field coordinator did — but it created the structured, unhurried space in which that insight became a design decision.
What the evidence predicts (illustrative results)
Because this is a composite, there is no real endline to report — and it would be dishonest to invent one. But the target is not plucked from the air. Modeled on the effect sizes TaRL has produced in rigorous evaluations — where well-implemented programs have doubled the share of children able to read a paragraph — a jump from roughly a fifth to a half of Grade 3 children reading with comprehension over two years is ambitious but evidence-consistent. That is what a fundable target looks like: bold enough to matter, grounded enough to defend.
What a consultant takes from this
The order is the lesson. The team's first idea — reading camps and storybooks — survived into the final design only in a minor supporting role, because the problem tree showed it did not touch the root cause and the evidence pointed elsewhere. Had they started with the matrix, they would have written a tidy grid around the wrong intervention. The Logical Framework Approach, used properly, is not paperwork you produce for a funder. It is the thinking that stops you from confidently doing the wrong thing at scale.
Designing an FLN or education program of your own? Start free with LFA Studio and build it the right way round — analysis first, matrix last.
Sources
- World Bank — 70% of 10-year-olds in learning poverty (2022) and India learning-poverty data.
- ASER Centre / Ideas for India — ASER 2024 findings.
- Ministry of Education — NIPUN Bharat mission guidelines.
- J-PAL — Teaching at the Right Level: evidence; Pratham — TaRL approach.