TumbuhGuard Standardize — Validate the measurer, not just the measurement

Inspiration: The problem we want to solve

A trainee measures a child's height twice: 73.0 cm and 72.9 cm. The readings are almost identical, which looks reassuring. But a qualified reference measurer records approximately 74.2 cm.

The trainee is consistent, yet consistently inaccurate.

That's the problem TumbuhGuard addresses.

Child growth programmes rely on anthropometric data, but collecting measurements digitally doesn't guarantee that the measurements themselves are reliable. Supervisors need a structured way to evaluate whether workers can measure accurately before trusting their results.

We were motivated in part by reported quality concerns in community anthropometry programmes in Indonesia. We wanted to explore an approach that could also be useful in other underserved, connectivity-constrained settings.

What we built

TumbuhGuard is an offline-first web application that helps supervisors assess practical length and height measurement skills.

Instead of simply accepting measurements and calculating scores, it enforces an assessment workflow:

Setup → First measurement round → Lock → Blinded second round → Reference measurements → Evidence → Evaluation → Review → Remediation

The system evaluates measurement repeatability, agreement with a qualified reference, reference validity, and protocol compliance.

A built-in synthetic demonstration shows a trainee with excellent repeatability but poor reference agreement. The supervisor can inspect the evidence and initiate a linked re-standardization session without rewriting the original assessment.

The innovation is not the Technical Error of Measurement formula itself. It is the structured, reviewable process around that formula.

How we built it

The frontend uses React, TypeScript, and Vite. A deterministic domain engine handles calculations and enforces legal workflow transitions.

Dexie and IndexedDB support local persistence. Zod validates stored records, Web Crypto provides application-level integrity checks, and revision-based compare-and-swap guards against stale writes.

Workbox provides offline-first PWA functionality after the application has been loaded and cached.

We test calculations, invalid transitions, concurrency, persistence, offline behaviour, and browser interactions using Vitest and Playwright.

The current prototype does not require accounts, cloud databases, runtime AI APIs, or real child records.

Our biggest technical challenge

The difficult part was preserving assessment integrity.

We had to prevent the second measurement round from being influenced by the first, reject invalid workflow transitions, handle concurrent browser tabs, and preserve historical assessment evidence after remediation.

These challenges led us to design a protocol-controlled state machine instead of a conventional form or spreadsheet.

Who would use it?

Our proposed customers are organizations responsible for training and supervising field measurers, including public-health NGOs, community health worker training programmes, and research organizations.

Individual field workers would use the tool during supervised assessments, while training organizations would be the intended paying customers.

Business model and financial sustainability

Our initial business hypothesis is institutional licensing rather than charging individual workers.

A possible entry-level subscription is $49 per organization per month, with larger training programmes paying for onboarding, support, and future multi-team reporting capabilities.

At 100 paying organizations, this illustrative price would produce $4,900 in monthly revenue before operating expenses.

This is a pricing hypothesis, not an existing revenue stream. No paying customers or validated willingness-to-pay are claimed.

The current browser-local architecture avoids mandatory backend hosting for assessments, but real operating costs would include training, support, product maintenance, security review, and institutional onboarding.

Our proposed go-to-market approach is to begin with supervised pilots involving qualified trainers, validate the workflow and its usefulness, then test institutional licensing.

Scalability

The existing PWA can be distributed as a static application to many independent devices. After caching, the core assessment workflow can operate offline.

That makes it suitable for exploration in settings with inconsistent connectivity.

However, the present implementation is local-first, not a centralized multi-user SaaS product. Organization-wide reporting, authenticated team accounts, cross-device synchronization, and production data governance would require additional engineering.

A future institutional version could add optional encrypted synchronization and controlled supervisor reporting while retaining offline assessment capabilities.

Responsible deployment and limitations

TumbuhGuard currently operates with synthetic demonstration subjects and a working anthropometry standardization profile.

It is not a diagnostic tool, official certification system, or WHO/Kemenkes-approved product. Independent external-reference parity, qualified-trainer validation, and real-world institutional testing remain necessary before operational use.

This project was developed previously as a competition prototype and is being presented here with a startup-oriented business proposal. We do not claim it was newly built for Next Founders.

What we learned

Reliable health data starts before data entry.

A digital system can store a measurement perfectly even when that measurement is wrong. Building TumbuhGuard reinforced the importance of controlled processes, reproducible calculations, reviewable evidence, and clear scientific limitations.

Our goal is to help training organizations identify measurement problems earlier and improve the trustworthiness of the data they eventually collect.

Built With

• broadcastchannel
• dexie
• indexeddb
• javascript
• playwright
• progressive-web-app
• react
• sha256
• typescript
• vite
• vite-plugin-pwa
• vitest
• web-crypto-api
• workbox
• zod