Chapter Two: About the dataset

A real supply chain dataset, and the facts in one row

The data

Trase is a nonprofit initiative that maps the supply chains of commodities linked to deforestation, and publishes the results as open data. This course uses an excerpt of its Indonesian palm oil supply chain dataset:

Benedict, J. J., Biddle, H., Gollnow, F., Heilmayr, R., Mueller, C., Ribeiro, V., & Suavet, C. (2024). Indonesia palm oil supply chain (2018–2022) (Version 1.2) [Data set]. Trase. https://doi.org/10.48650/X83N-7M36. Licensed under CC BY 4.0.

One row of the dataset is one trade flow: palm oil grown in a district (a kabupaten), processed at a mill, sometimes refined at a refinery, exported by one company and imported by another into a country of first import.

The excerpt contains:

81

trade flows: 71 of refined palm oil and 10 of crude palm oil

41

mills

6

refineries

37

companies, as exporters, importers and parent companies

6

districts of production, all in the province of East Kalimantan

6

EU countries of first import

5,505 ha

total 10-year deforestation exposure

How the excerpt was made
  • Rows: from about 1.7 million rows, only flows from 2022, from East Kalimantan to the EU, with every actor known; then 81 of those, chosen so that every modeling case in this course appears. No values were invented or edited.

  • Columns: 16 of about 35. Dropped: the year, Trase identifiers and country codes, the port of export, economic blocs, sustainability scores and commitments, and emissions metrics.

  • Caveat: Trase flows are modeled from customs, production and facility data. They are estimates, not observed shipments.

You can download the excerpt as a CSV file: palm_kaltim_eu_2022_demo.csv.

The columns

Group Columns

Places

country_of_production, province_of_production, kabupaten_of_production, country_of_first_import

Companies and facilities

mill, refinery, exporter, importer, each with a *_group column naming its parent company where there is one

The flow itself

product_type, volume (tonnes), fob (value in US dollars), palm_oil_deforestation_10_year_total_exposure (hectares)

One row, as elementary facts

Here is one row, transposed, with the numbers rounded:

Column Value

country_of_production

INDONESIA

province_of_production

KALIMANTAN TIMUR

kabupaten_of_production

BERAU

mill

JABONTARA EKA KARSA (JABON)

mill_group

KUALA LUMPUR KEPONG (KLK)

refinery

LDC EAST INDONESIA

refinery_group

LOUIS DREYFUS

exporter

LDC EAST INDONESIA

exporter_group

LOUIS DREYFUS

importer

LOUIS DREYFUS

importer_group

LOUIS DREYFUS

country_of_first_import

SPAIN

product_type

REFINED PALM OIL

volume

3750.036

fob

1311345.23

palm_oil_deforestation_10_year_total_exposure

916.7547

Before modeling, let’s list the facts this row states:

  1. The kabupaten Berau lies in the province Kalimantan Timur.

  2. The province Kalimantan Timur lies in the country Indonesia.

  3. The mill Jabontara Eka Karsa belongs to the group Kuala Lumpur Kepong.

  4. The refinery LDC East Indonesia belongs to the group Louis Dreyfus.

  5. The exporter LDC East Indonesia belongs to the group Louis Dreyfus.

  6. Palm oil from Berau, processed at the mill Jabontara Eka Karsa and refined at the refinery LDC East Indonesia, was exported by LDC East Indonesia and imported by Louis Dreyfus into Spain as refined palm oil.

  7. That flow had a volume of 3,750.036 tonnes, a value of 1,311,345.23 US dollars, and a deforestation exposure of 916.7547 hectares.

Most of these facts are binary. The sixth is not: it relates seven things at once, and it cannot be split into binary facts without losing information. Knowing which mill supplied which refinery, and which refinery supplied which exporter, does not tell you which mill’s oil reached which importer. Chapter Five shows what is lost if you try.

Some facts need interpretation. LDC EAST INDONESIA appears as both the refinery and the exporter, so is it one thing or two? And NOT REFINED, used in other rows, is not a refinery at all, but the absence of one. Chapter Four covers these decisions.