Product Engineering for a Sustainability Leader's Field Data Platform
AUS-based sustainability leader engaged Brainstack Technologies to build a field data-collection platform for smallholder farmers across rural East Africa, Latin America, and South Asia. The engagement spanned the full product engineering arc — discovery, proof of concept, MVP, multi-region rollout, and handover — under constraints that ruled out off-the-shelf solutions and demanded careful technical decisions at every phase.
The platform feeds an annual sustainability assessment programme used by global commodity buyers and certification bodies. Every farmer interview entered into the system needs to reach the cloud accurately and unedited, despite field conditions that include intermittent network coverage, shared low-end devices, frequent app uninstalls, and field officers working with farmers who have never used a smartphone before. Off-the-shelf survey tools fail one or more of those constraints. We were brought in to engineer a platform that handles all of them.
The Problem Space
Building a data-collection product for sustainability assessments at this scale is fundamentally different from building a typical SaaS application. The constraints below shaped every technical decision in the engagement:
Connectivity Is The Hardest Constraint
In rural East Africa and parts of Latin America, network coverage depends on towers that occasionally run out of fuel. Field officers can spend hours interviewing farmers in areas with no usable signal. The platform cannot lose any data captured during these periods, and field officers need confidence that their work has been recorded — even before sync. A naive cloud-first form-builder would lose data, frustrate users, and erode trust in the assessment programme itself.
Devices Are Shared And Low-End
Field officers often work from shared smartphones with limited storage and older Android versions. Apps get uninstalled accidentally when storage runs low. Personal mobile data is rare, so app installation through Play Store is friction-heavy and unreliable. Native apps are an operational liability in this environment. The platform needed to install instantly without a Play Store visit, recover gracefully from accidental uninstalls, and run on hardware that pre-dates many modern web APIs.
Low-Literacy End Users
The farmers being interviewed have varying literacy levels and often do not speak the field officer's primary language. Assessment workflows that depend on text-heavy UI, drop-downs, or precise typed input fail in these conditions. The platform needed image-driven interactions, iconographic prompts, and clear visual confirmation that data had been captured, with translation support for the languages of each region.
Programme Integrity Is Non-Negotiable
The assessment data feeds sustainability benchmarks used by global commodity buyers in coffee and cocoa supply chains. Buyers and certification bodies make procurement and audit decisions on this data. Any platform handling it needs to give the sustainability leader confidence that field data is being captured fully, that no records are silently dropped on sync, and that field officer engagement can be monitored over time so coverage gaps can be identified and addressed.
How We Approached the Engagement
We delivered the product across four distinct phases, each with a clear exit criterion before the next began. The discipline of phase gates — and the willingness to invest in PoC before MVP — is what made the multi-region rollout possible.
Discovery: Understanding The Field Reality
Discovery: Understanding The Field RealityBefore writing any code, we spent time with the sustainability leader's field operations team understanding what actually happens during a farmer assessment. The artefact from this phase was a written technical assessment that mapped each constraint to a candidate technical approach, with the sustainability leader's programme owners signing off on the trade-offs before we moved to PoC.
The biggest decision settled in this phase: Progressive Web App over native. A PWA installs from a URL without a Play Store visit, runs on older Android versions through a Chrome shim, can be re-installed in seconds after accidental uninstalls, and gives us a single codebase across regions instead of one per region. Native gave better device integration but cost too much in distribution friction for this user group. We documented the rationale so it would not need re-litigating later.
Proof of Concept: Prove Offline-First Works
The PoC was deliberately narrow: one assessment workflow, one country, ten field officers, one month of real use in conditions that mimicked the worst-case network reality. Its purpose was not to look like the final product. Its purpose was to retire risk on the single most expensive technical assumption — that an IndexedDB-backed offline queue could capture data reliably across days of disconnected use and sync everything cleanly when devices returned to coverage.
Every record captured during the PoC carried metadata: device clock time, network signal at capture, time of sync attempt, sync result. The PoC retrospective examined every record that had any anomaly — clock skew, late sync, partial submission — and either fixed the underlying behaviour or documented an accepted trade-off. Going into MVP, we knew exactly what the platform did under field conditions, not what it did in a developer's emulator.
MVP: Image-Driven UX And First Pilot Cohort
The MVP added the production-grade UX. Instead of typed input fields, we built image-based prompts: photos of common livestock, family-member icons used in place of name typing, iconographic indicators for activity types. Every form interaction had a visual confirmation when data was captured, and the field officer could review the day's captured records before leaving the village — even with no signal — so they had confidence the work was recorded.
We also added lightweight engagement telemetry: when did each device last sync, how many farmers did each field officer assess this week, and which regions had coverage gaps. The sustainability leader's programme operations team used this data to identify field officers who had stopped working — sometimes because of broken devices, sometimes because of life changes that meant they had left the role — and re-allocate coverage before assessment quotas were missed. The MVP went into pilot use with the first cohort of farmers in the original PoC country and ran for a full assessment cycle before we moved to multi-region.
Multi-Region Rollout And Handover
With the MVP validated against a full assessment cycle in the pilot country, we expanded to additional regions across Latin America and South Asia. The architecture was designed to make this expansion mostly a configuration exercise rather than a re-build: per-region locale and translation packs, per-region commodity-specific question sets, per-region field-officer training material. Multi-region rollout was completed without re-architecting the platform.
Handover was structured rather than transactional. The sustainability leader's in-house engineering and data teams received written architecture documentation with rationale (especially around the offline-first decision), a runbook covering deployment, sync monitoring, and incident response, and pair-working sessions where their first hires worked through the codebase with our lead engineer. Sync-pipeline ownership transferred to their data team, application development to their engineering team, and we provided a defined support window after the formal engagement ended.
Technology Choices
The stack was chosen for three properties: low device-side friction (PWA, no Play Store dependency), offline reliability (IndexedDB-backed queue with deterministic sync semantics), and operational simplicity for the sustainability leader's in-house team to inherit at handover.
Device-Side Application
Cloud Backend
Operations
Internationalisation
What This Engagement Taught Us
An offline-first PWA on shared low-end devices in low-connectivity regions is not a problem you can derisk in a developer's emulator. The PoC, narrow as it was, retired the single most expensive technical assumption before we committed to MVP. If we had skipped it, we would have rebuilt sync logic mid-MVP at twice the cost.
Device clock time, network signal at capture, sync attempt time, sync result — all attached to every assessment record. This metadata was invisible to field officers but indispensable to the programme team. It is what let us identify field officers who had silently stopped working, regions with degrading coverage, and devices misbehaving in ways that data integrity reviews would otherwise have missed.
For users where reading and typing are an active barrier, image-driven interactions are a correctness requirement, not a polish item. The MVP's adoption rate among low-literacy users was a direct function of how aggressively we removed text from the assessment workflow. Drop-downs and free-text fields are reserved for the small number of fields where they are unavoidable.
Locale, translation packs, and per-region commodity question sets were structured from MVP, even though the pilot ran in only one country. This made multi-region expansion a configuration exercise rather than a re-architecture. Trying to retro-fit i18n into a monolingual MVP would have cost more than building it correctly the first time.
Conclusion
What separated this engagement from a typical SaaS build was discipline at every phase boundary. Discovery produced a written assessment of constraints and trade-offs, signed off before any code. PoC retired offline-sync risk before MVP. MVP shipped image-based UX and engagement telemetry with the first cohort of farmers, not after. Multi-region rollout was a configuration exercise because the architecture was set up for it from MVP. Handover was structured, not transactional — with documentation, runbooks, pair-working, and a defined support window.
The platform now runs production assessment cycles for the sustainability leader's programme across three continents, feeding the sustainability benchmarks that global commodity buyers and certification bodies rely on. Both the device-side application and the cloud sync pipeline are owned and operated by the sustainability leader's in-house teams.
Have a Product That Has To Work In The Real World?
If your product needs to work in the field environment your users actually have — not the one a developer's emulator simulates — we can take you from discovery through structured handover. We do PoC work where the constraints are unfamiliar, image-driven UX where literacy varies, and offline-first when connectivity is part of the problem.
Technologies
React
Node.js
Python
PostgreSQL
Odoo
Docker
Kubernetes
TypeScript
MongoDB
More Success Stories
Custom Software Development
Mobile Development
Microservices Architecture
