Geo-tagged surveys
Sessions at beaches, rivers, markets, landing sites and dumpsites, with coordinates, weights and photo evidence.
Data Platform · Geospatial · Full Stack
Data infrastructure that turns field surveys into decisions
A shared civic dataset of geo-tagged waste surveys for Tanzania and East Africa, with trust scoring, brand audits and published benchmarks.
takadata.org- Industry
- Environmental data
- Timeline
- Ongoing
- Year
- 2026
T
What it supports
Item-level breakdown
Counts against a full litter taxonomy, plus material, product and recyclability classification.
Brand audits
Which producers' packaging was found where — the evidence base for extended producer responsibility.
Trust scoring
Automatic 0–100 score and A–D grade per entry, with peer verification restricted to other organisations.
Cleanliness benchmarks
Density graded against the Clean-Coast Index and OSPAR/EU thresholds, each badged with what kind of standard it is.
Multi-organisation access
Admin, collector and viewer roles per organisation over one shared cross-org dataset.
The challenge
Waste data in the region is collected constantly and trusted rarely. Every NGO, community group and municipality keeps its own spreadsheet in its own shape, so nothing can be compared across sites or years — and nobody can say whether a beach is clean, or which producers' packaging keeps turning up on it. The hard part was never storage. It was making numbers other people are willing to act on.
How we built it
The decisions that shaped the product, in the order we made them.
- 01
Fixed the protocol before the schema
Adopted an established item-level litter taxonomy and brand-audit method, so an entry means the same thing across organisations and can be compared with work published elsewhere.
- 02
Made trust computable
Every survey scores 0–100 on evidence quality and grades A–D; only entries above the bar become public. Verification is cross-organisation by design — you cannot approve your own data.
- 03
Built the attribution engine
Maps each litter item to the human activity that produced it, the product category, a behaviour class and a material code — with the honesty rules written into the code: a fragment attributes to 'unknown', never to a guess.
- 04
Graded sites against published scales
Shoreline sites use the Clean-Coast Index and OSPAR/EU thresholds; everything else uses provisional bands that say so on the page. Every grade states its source and its status.
- 05
Published it
Public per-site pages and a map anyone can read without an account, because a dataset that needs a login to consult does not inform anybody's decision.
What it's built with
Named in full, because the question behind it is usually whether your own team could take this over.
- Django + DRFAPI and analysis layer
- PostgreSQLUUID keys throughout
- QGISgeospatial preparation and validation
- Pythonattribution and benchmark engines
- Leafletthe public map
- Rechartsdashboards and breakdowns
- React + TanStack Startserver-rendered
- Docker + Caddydeployed on one box
Results
A–D
Every survey graded before it goes public
Cross-org
Verification you cannot do on your own data
Public
Site pages readable without an account