Evidence
How Insight measures, and how we know it works.
The framework behind the instrument, the process used to validate it, and what the evidence shows so far — including what's still being studied.
The instrument
3 dimensions
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
constructs
8 SEL skills
9 Learning environment
7 Personal wellbeing
24
constructs, three dimensions
7 steps
published validation protocol
50M
assessments a year
2024
independent external review
The instrument
24 constructs, three dimensions.
Each dimension captures a different part of a student's experience. Together they give a complete picture — not a single wellbeing score.
8
SEL skills
The internal capacities a student draws on to learn and relate.
9
Learning environment
The climate, relationships, and culture of the school around them.
7
Personal wellbeing
The emotional state, supports, and habits that sustain them.
Insight also assesses the socio-emotional disposition of teachers and school leaders, with the same instrument architecture.
Foundations
We didn't invent our own theory of what matters.
Insight's framework rests on the work that education systems already recognize: CASEL's competency model, UNESCO's four pillars of education, and the OECD's work on social and emotional skills.
That matters for a practical reason. A school or ministry adopting Insight isn't adopting a private theory — it's adopting an instrument built on frameworks their own policies already reference.
CASEL
Competency model
UNESCO
Four pillars of education
OECD
Social and emotional skills
Validation
A published process, followed item by item.
Insight builds and refines its scales using the process developed at the Harvard Graduate School of Education by Gehlbach and Brinkworth — a method designed specifically to reduce measurement error before a question ever reaches a student.
01
Literature review
Define each construct precisely, against existing research.
02
Interviews and focus groups
Test whether real students and educators think about it the same way.
03
Synthesis
Reconcile what the literature says with what the population says.
04
Item development
Write questions using the respondents' own vocabulary.
05
Expert review
Independent specialists rate each item for relevance and clarity.
06
Cognitive pre-testing
Students explain in their own words what each question is asking.
07
Pilot testing
Statistical analysis identifies and removes items that don't perform.
A question only stays in the instrument when it measures something distinct, measures it consistently, and measures something a school can act on.
Distinct
Consistent
Actionable
Gehlbach, H. & Brinkworth, M. (2011). Measure twice, cut down error: A process for enhancing the validity of survey scales. Review of General Psychology, 15(4), 380–387.
Calibration
Validated globally. Calibrated locally.
The frameworks Insight builds on were developed and validated largely in a handful of educational contexts. They're the right starting point. They aren't the finishing point.
In India
NEP 2020's framework for measuring the whole child
Calibrated for each context
Items re-piloted per context and age band · thresholds set against real peers · constructs added where data shows a gap
Internationally validated base
CASEL · UNESCO · OECD
A question that measures perseverance well in one country can measure something subtly different in another — or nothing at all. The same is true of belonging, of safety, of how students relate to teachers. Language, school culture, and family expectations all shape what an answer means.
So Insight keeps the internationally validated base intact and calibrates on top of it: items rewritten and re-piloted for each context and age band, thresholds set against real comparison groups rather than imported norms, and constructs added where the data shows something is being missed.
Calibration doesn't relax the rigor. It raises it.
Every added layer goes through the same validation protocol as the base — literature review, cognitive interviews with real students in that context, piloting, factor analysis, refinement. A construct only enters the instrument when it demonstrably measures something distinct, does so consistently, and measures something the school can act on.
This is also why a result means something. A score is read against genuine peers — same grade, same kind of school, same context — not against a population the student has nothing in common with.
In India
In India, that meant calibrating against NEP 2020's own framework for measuring the whole child — not adapting language alone, but validating against the standard schools are already required to meet.
Independent evaluation · 2024
We asked an outside body to examine the work.
In 2024, Insight commissioned the International Centre for EdTech Impact — an independent research organization that evaluates whether education technology is actually grounded in science — to review the platform, the research behind it, and the claims made for it.
Logic model
what Insight does and what it’s meant to change
Theory of change
each expected outcome tied to published research
Theory of action
the scientific pillars of the work, and the gaps still to close
We're publishing it in full, including the parts that describe what hasn't been established yet.
Independent review
Insight: logic model, theory of change and theory of action
PDF · 2024
Download the ICEI review (PDF) →
Research status
What's established, and what we're still studying.
Insight's expected outcomes rest on a substantial body of published research on social-emotional learning — including meta-analyses covering hundreds of studies across dozens of countries, with consistent findings on academic achievement, school climate, peer relationships, and engagement.
What that literature establishes is that SEL works. What it doesn't establish is the specific effect of Insight's own components — the assessment, the intervention library, the AI assistant, the goal-setting. That's what we're studying now.
Working from the research plan developed in the ICEI review, correlational studies are currently underway, with higher-tier study designs planned to follow.
We'd rather say that plainly than claim more than we can show.
Where the evidence stands
Published SEL research
Meta-analyses across hundreds of studies
Established
Insight's own components
Correlational studies, from the ICEI research plan
Underway
Higher-tier study designs
To follow the correlational work
Planned
Findings
What we're seeing across the assessments.
Insight runs more than 50 million assessments a year. Some patterns repeat often enough, across enough schools and contexts, to change how a school or a system plans.
Signal on a cohort
Sense of Belonging
4° Primaria · 48 students
2.4
▼ 0.4 since Oct
2.8
2.7
2.6
2.5
Oct 2.8
Feb 2.6
May 2.4
Need range
Belonging comes before disengagement.
Students who leave school before graduating almost always show low sense of belonging first. The signal matters, but the root matters more — belonging is built in who students choose for group work, in what happens at recess, in whether they're part of anything beyond the school day. Knowing where it's missing is what makes an intervention fit the student.
Skill up, results follow
+13%
Academic performance
Delay gratification segment · one term
+0%
+1%
+2%
+3%
+4%
+5%
+6%
+7%
+8%
+9%
+10%
+11%
+12%
Oct
Feb
May
Self-management
Student–teacher connection
Academic results
Delayed gratification tracks with academic performance.
High-achieving students consistently share the ability to delay gratification. Which raises a better question than "how do we raise grades?" — what if that skill were built deliberately, from the first years of school, rather than left to appear on its own?
+13%
average academic improvement
−17%
average reduction in dropout
100%
client retention since launch
These are patterns in our own data, not universal laws. They're the starting point for the research described above, not a substitute for it.
See what it finds in your own schools.
The instrument is best judged on your own students. Schools can start with a free trial; districts and systems can start with a pilot.