A remedial-learning programme, a teacher-coaching initiative, a scholarship fund and a state partnership need different ways of working. This playbook helps you identify your model, the information your team needs and the smallest useful system to begin with.
| Meera K. | 10 | not enrolled?? |
| Meera Kumari | 9 | dup? |
Five places, one child. Nobody can see she's about to leave.
Many organisations combine two or three. Start with the closest match, then borrow only the parts that solve a real programme problem.
Your team runs classes, camps or remedial groups directly with children and young people.
You train, observe and coach teachers, fellows or community educators over time.
You work with school leaders and staff to improve several parts of a school.
You support applications, disbursements, mentoring and movement into the next stage.
Volunteers identify learners, run sessions or support enrolment and retention.
You support government staff, schools or districts to adopt and sustain a programme.
Not sure which one you are? Bring one tracker and describe a normal week. The model usually becomes clear quickly.
These come up in almost every first conversation we have with education teams.
"We found out she'd dropped out when the endline came."
Attendance is tallied monthly, so a child who stops coming in week two shows up as a number in week eight.
"Enrolment, learning and second chance sit in three systems."
Nobody can say which girls we enrolled last year are now reading at grade level without a month of matching names by hand.
"Twelve facilitators, twelve versions of the sheet."
Names spelt three ways. Duplicate children. Nobody's sure which file is current.
"The funder report takes three weeks."
Chasing Excel files, reconciling totals, and still unable to say how many children reached Class 6.
The people and screens above are illustrative. These are real systems we've built with partners, and each write-up says what we built and how long it took.
Compass MIS runs state-wide teacher training, classroom observations and multilingual assessments. The first week of a new-state pilot logged 2 million interactions.
Read the case study →TAP India Foundation16,000+out-of-school children trackedIdentification, enrolment, attendance and retention in one system, with separate access for field associates, partners and managers.
Read the case study →Avanti Fellows8,000students on prioritised call queuesA telecalling system that pulls students from the main database and shares calls fairly across 10 to 15 telecallers. Live in four months.
Read the case study →Recognise your programme? Tell us where the work gets stuck and we'll suggest a useful first version.
This is one operating model, not a template for every education organisation. Use it to see how everyday work, follow-up and management decisions can connect.
Stage 01 · Find every child
Seasonal migrants, girls kept home after Class 5, children working in the fields. If the survey stops at the main road, so does your enrolment drive.
| ID | Child | Habitation | Status |
|---|---|---|---|
| ED-1042 | Meera Kumari | Brick Kiln Cluster | Visit due today |
| ED-1029 | Pooja Devi | New Colony | Overdue |
| ED-1038 | Rafiq Ansari | Bhata Tola | Visit booked |
| ED-1016 | Arjun Singh | Kalyanpur | Enrolled |
Stage 02 · Group by level
A grade label hides the learning level. A Class 5 child reading at letter level needs a different room from one reading stories.
| ID | Child | Reading | Group |
|---|---|---|---|
| ED-1029 | Pooja Devi | Letter | Vision referral |
| ED-1042 | Meera Kumari | Word | Reading B |
| ED-1038 | Rafiq Ansari | Paragraph | Reading C |
| ED-1016 | Arjun Singh | Story | Reading C |
Stage 03 · Catch who slips
A monthly attendance total arrives too late. By then, the family has already left for the kiln.
| Group | Facilitator | Attendance | Action |
|---|---|---|---|
| Reading C | Rina Das | 76% | Priority visit |
| Reading A | Asha Sharma | 82% | Mentor visit |
| Maths B | Kabir Ali | 88% | Parent calls |
| Reading B | Joseph Mathew | 91% | On track |
Stage 04 · Prove it & follow through
Endline scores only count if they stay connected to the child. Keep the case open until she's on the next school's register.
| ID | Child | Reading gain | Next school |
|---|---|---|---|
| ED-1029 | Pooja Devi | +2 levels | Not traced |
| ED-1038 | Rafiq Ansari | +1 level | Admission pending |
| ED-1042 | Meera Kumari | +2 levels | Confirmed |
| ED-1016 | Arjun Singh | No change | Confirmed |
Shown for one block. The same screens roll up to district and state. Run teacher development? The record follows the teacher: every training, observation and coaching cycle, with their class's learning results alongside. See how Compass MIS does it for 20,000+ teachers →
A funnel usually shows where you lost children. This one shows who's now responsible for each of them.
Illustrative numbers. Learning gain is always reported with the attrition beside it.
Choose the parts that match how your education programme actually works. Most organisations need a focused combination, not every module.
Set up the places and groups where learning happens, including multi-year movement.
Track recruitment, training, placement, observation, support and alumni journeys.
Keep assessment tools, curriculum versions and learning evidence connected.
Turn visits and feedback into clear coaching actions and follow-up.
Manage eligibility, documents, disbursements, mentoring, renewal and progression.
Use official codes, geography and reporting formats, with a clear handover trail.
The stages show the story. This is the checklist your procurement, finance and M&E teams will look for. We'll show each one working in the walkthrough.
| Capability | What it covers | Where it runs today |
|---|---|---|
| Enrolment and admissions | Household surveys, out-of-school reasons, admission confirmed by the school, duplicate checks across villages. | Live at TAP India |
| Attendance | Daily or session attendance offline, with absence streaks that create a home-visit task. | Configured to you |
| Learning assessments | ASER-style or your own tools, baseline to endline matched per child, with attrition shown beside gain. | Configured to you |
| Groups, sessions and facilitators | Grouping by level, session logs, facilitator and volunteer records. | Configured to you |
| Teacher training and observations | Training cohorts, observation rubrics, coaching cycles and classroom visits. | Live at Compass MIS |
| Transitions and second chance | Class 5 to 6 and 8 to 9 tracking, open-school registrations and results day. | Configured to you |
| Scholarships and fees | Applications, eligibility checks, disbursal tranches and renewals, where your programme needs them. | Configured to you |
| Government IDs | UDISE+ school codes, APAAR or state portal IDs stored alongside yours. | Configured to you |
| Donor reporting and audit trail | Funder-wise cohorts and indicators, with a change log on every record. | Configured to you |
| Integrations | Survey apps, BigQuery, WhatsApp, LMS and portal exports, connected rather than replaced. | Live at Avanti Fellows |
| Roles and permissions | Volunteers, field associates, partners and managers each see only what they need. | Live at TAP India |
| Offline and languages | Phone forms that work without signal, in Hindi, Marathi, Kannada or your language. | Configured to you |
'Live at' names a real system with that capability. Everything else is configured per programme, so not every build includes every row.
Most AI projects stall because the records underneath are messy. So we work in a fixed order.
Every field has to earn its place by helping someone decide something. If nobody uses it, we don't collect it.
A system people avoid is worse than a spreadsheet. We design around shared phones, patchy signal and WhatsApp.
On clean, connected records, AI can read paperwork, draft reports and flag who needs attention. A person always checks.
Where you start depends on where you are. If your records are already in good shape, we can move to AI quickly. If they're in notebooks, we start there, and we don't start from zero.
Offline forms, local languages, links to government IDs and AI features are set up to suit each programme's needs and consent rules, so not every build includes all of them.
We picked the jobs that save education teams the most hours. Each starts with a problem we hear often, and each keeps a person in charge of the decision. How we think about AI →
“We have thousands of volunteer notes on why a girl isn't in school, and nobody has time to read them.”
Your team still decides: The coordinator confirms every tag before a visit is assigned.
“Government schools keep paper registers, and we can't ask teachers to enter attendance twice.”
Your team still decides: A volunteer checks every extracted tick.
“Assessing every child one-on-one takes weeks before teaching can start.”
Your team still decides: The facilitator listens too and can override the level.
“By the time district numbers reach us, the girls who stopped coming have already left.”
Your team still decides: It shows which groups haven't reported, before any trend.
“We find out a girl didn't make it to Class 6 when the next survey runs.”
Your team still decides: The coordinator adds children the data missed.
“Our funders and evaluators want matched baseline to endline, and building it takes a month.”
Your team still decides: Your M&E lead puts attrition on page one.
Children's data needs care. These lines don't move, whoever's asking.
As children get enrolled, they drop off the list. Progress you can see on your own phone.
Children by level, today's absences and the mentor's last feedback, on one screen.
Groups ranked by falling attendance and time since the last visit.
Learning gain by baseline level, facilitator and cluster, with support budget alongside.
Matched baseline to endline, externally checked, with dropout by gender and grade.
If 30% of children left before the endline, your average gain looks better than your programme is. We report the gain and the attrition, side by side.
Funders notice. It's the kind of honesty that gets you renewed.Most systems launch in three to four months. The timeline moves with the number of programmes and roles, how messy the old data is, and how many tools need connecting.
Tell us where the work gets stuck today. Spreadsheets, notebooks, WhatsApp groups, all of it.
You show us how the programme runs. We'll tell you straight if you need a new system, a better workflow or nothing at all.
We map how your programme works, talk to the people who'll use it, review the tools you use today and hand you a System Blueprint, a clear plan of what we'll build.
Working screens every 2 to 4 weeks, tested by your field team in your own terms. We clean up your old spreadsheets and move them across.
Field training and launch, then 30 to 60 days of close support. After that, the system keeps changing as your programmes do.
We connect to your survey app (SurveyCTO, ODK or Kobo), BigQuery or Postgres, WhatsApp, your LMS and UDISE+ exports, rather than replacing what already works. Zoho licences sit in your organisation's name, and everything exports whenever you want it.
The questions your finance head and IT lead will ask, answered before the call.
Zoho licences are bought in your organisation's name and hosted in the Zoho data centre region you choose, including India. We work inside your account.
Every record exports to Excel or CSV whenever you want, including the day you stop working with us. You're not locked in to us for support either.
Each person sees only the records their role needs. Children's records follow the rules for minors in India's DPDP data protection law, with guardian consent recorded.
Edits are logged with who made them and when, so an audit or a funder query doesn't start from memory.
Consent is recorded for each purpose, and we agree up front how long data is kept and when it's deleted. ZolaShield, our DPDP platform, shows where you stand.
AI features run on your records under your account's controls. Beneficiary records never go into public AI tools.
Use these focused implementation guides alongside the sector playbook.
We’ll map what your team does now, identify the smallest useful first version and show where follow-up, learning and reporting can share the same information, without forcing your programme into a generic template.
Four short questions. Plain language is perfect; you do not need to have the solution worked out.
The same approach, shaped to each kind of work.
