Quickstart

Overview

This guide walks you through a complete Cardinal workflow using a real example. You’ll initialize settings, prepare OCDS data, and calculate procurement indicators.

Before you begin: Make sure you have Cardinal installed on your system.

Prerequisites

Cardinal installed via Cargo or pip
OCDS data in line-delimited JSON format (compiled releases)
Basic familiarity with the command line

Don’t have OCDS data yet? Download sample data from the OCP Data Registry or use the examples in this guide.

Complete Workflow

Cardinal’s workflow consists of three main steps: initialize settings, prepare data, and calculate indicators.

Step 1: Initialize Settings

Create a default configuration file that enables all indicators and documents all available options:

ocdscardinal init settings.ini

Output

Settings written to settings.ini.

This generates a settings.ini file with default values and extensive comments explaining each option.

View the settings file contents

You can preview settings without creating a file by using - as the filename:

ocdscardinal init -

This outputs the settings to your terminal:

; currency = USD
; no_price_comparison_procurement_methods = Random Selection
; price_comparison_procurement_methods = Reverse Auction

[output]
; info_currency_mismatches = true

; `prepare` command
;
; Read the documentation at:
; https://cardinal.readthedocs.io/en/latest/cli/prepare.html

[defaults]
; currency = USD
; item_classification_scheme = UNSPSC
; bid_status = valid
; award_status = active
; party_roles = true

[redactions]
; amount = 0
; organization_id = placeholder

; ... (additional configuration sections)

[R003]
; threshold = 15
; procurement_methods = open|selective|limited

[R018]
; procurement_methods = open|selective

[R024]
; threshold = 0.05

; ... (more indicator configurations)

Step 2: Prepare Your Data

The prepare command corrects quality issues in your OCDS data and reports problems that could affect indicator calculations.

Run the prepare command

Process your input data, writing corrected data to one file and quality issues to another:

ocdscardinal prepare input.jsonl \
  --settings settings.ini \
  --output prepared.jsonl \
  --errors issues.csv

Replace input.jsonl with your OCDS data file.

Command breakdown

input.jsonl: Your OCDS compiled releases (one JSON object per line)
--settings settings.ini: Configuration file from step 1
--output prepared.jsonl: Where to write corrected data
--errors issues.csv: Where to write quality issue reports

Review quality issues

Open issues.csv to see reported quality issues. The CSV has these columns:

Column	Description
line	Line number in the input file
ocid	OCID of the contracting process
path	JSON path to the problematic field
array indexes	Index position if the field is in an array
incorrect value	The problematic value (blank if not set)
error description	Type of issue: `not set`, `invalid`, or `is zero`

Example issue:

line,ocid,path,array indexes,incorrect value,error description
1,ocds-213czf-1,/bids/details[]/status,0,,not set

This indicates that bid #0 in contracting process ocds-213czf-1 is missing a status field.

Configure data corrections

Edit settings.ini to fix quality issues automatically. For example, if bids are missing status values, add:

[defaults]
bid_status = valid

Common configurations:

Fill in missing currency

[defaults]
currency = USD

Set default classification scheme

[defaults]
item_classification_scheme = UNSPSC

Redact placeholder amounts

[redactions]
amount = 0|99999999

Prefix organization IDs

[modifications]
prefix_tenderer_or_supplier_id = DO-RPE-

Re-map invalid bid statuses

[codelists.bid_status]
Qualified = valid
Disqualified = disqualified

Re-run and iterate

Run the prepare command again with your updated settings:

ocdscardinal prepare input.jsonl \
  --settings settings.ini \
  --output prepared.jsonl \
  --errors issues.csv

Review issues.csv again. Repeat this process until you’re satisfied with the data quality.

Most quality issues are repetitive and can be fixed with a single configuration change. For example, if 1,000 bids are missing statuses, adding bid_status = valid fixes all of them at once.

Step 3: Calculate Indicators

Run red flag analysis on your prepared data:

ocdscardinal indicators --settings settings.ini prepared.jsonl > results.json

Command options

Additional flags you can use:

ocdscardinal indicators \
  --settings settings.ini \
  --count \
  --map \
  prepared.jsonl > results.json

--count: Print the number of results per group to stderr
--map: Include mappings from contracting processes to organizations
--no-meta: Exclude statistical metadata from results

Step 4: Analyze Results

Open results.json to view your indicator results.

Example output structure

{
  "OCID": {
    "ocds-213czf-1": {
      "R036": 1.0,
      "R028": 1.0
    },
    "ocds-213czf-2": {
      "R018": 1.0,
      "R030": 1.0
    }
  },
  "Buyer": {
    "DO-UC-55216": {
      "R038": 0.85
    }
  },
  "Tenderer": {
    "DO-RPE-1422": {
      "R025": 0.12,
      "R048": 2.5
    }
  },
  "Meta": {
    "R024": {
      "q1": 66.67,
      "q3": 100.0,
      "lower_fence": 16.67
    },
    "R025": {
      "q1": 0.5,
      "q3": 1.0,
      "lower_fence": 0.0
    }
  }
}

Understanding the Output

Results are organized by group (OCID, Buyer, ProcuringEntity, Tenderer) and identifier:

OCID group

Results for individual contracting processes, keyed by /ocid.Example: "ocds-213czf-1" triggered indicators R036 and R028.

R036: Lowest bid disqualified (value: 1.0 = true)
R028: Identical bid prices (value: 1.0 = true)

Buyer group

Results aggregated by buyer organizations, keyed by /buyer/id.Example: Buyer "DO-UC-55216" has a high rate of disqualified bids (R038: 0.85).

ProcuringEntity group

Results aggregated by procuring entities, keyed by /tender/procuringEntity/id.

Tenderer group

Results aggregated by tenderers (bidders), keyed by /bids/details[]/tenderers[]/id.Example: Tenderer "DO-RPE-1422" has:

R025: Low win rate (0.12 = winning bid rate is a low outlier)
R048: High variety of items supplied (2.5 = heterogeneous supplier)

Meta section

Statistical information about quartiles and fences used for outlier detection:

"R024": {
  "q1": 66.67,      // First quartile
  "q3": 100.0,      // Third quartile
  "lower_fence": 16.67  // Outlier threshold
}

These values help you understand how indicators determined outliers.

Indicator Codes

Each code represents a specific red flag:

Code	Red Flag	Description
R003	Short submission period	Submission period is too short
R018	Single bid received	Only one tenderer submitted a bid
R024	Price close to winning bid	Second-lowest bid is suspiciously close to winner
R025	Excessive unsuccessful bids	Tenderer has low win rate (outlier)
R028	Identical bid prices	Multiple tenderers submitted same price
R030	Late bid won	Winning bid received after deadline
R035	All except winner disqualified	Only winner’s bids were valid
R036	Lowest bid disqualified	Lowest bid rejected despite price-only criterion
R038	Excessive disqualified bids	High disqualification rate (outlier)
R048	Heterogeneous supplier	Tenderer supplies unusually diverse items
R058	Heavily discounted bid	Winner’s price is much lower than competitors

Learn more about each indicator in the Indicators documentation.

Real-World Example

Let’s walk through a concrete example using sample data.

Sample Data

Create a file called example.jsonl with this OCDS compiled release:

{"ocid":"ocds-213czf-1","buyer":{"id":"DO-UC-001","name":"Ministry of Education"},"tender":{"id":"T-001","title":"Office Supplies","status":"complete","procuringEntity":{"id":"DO-UC-001"},"procurementMethod":"open","tenderPeriod":{"startDate":"2023-01-01T00:00:00Z","endDate":"2023-01-08T00:00:00Z"},"value":{"amount":100000,"currency":"USD"}},"bids":{"details":[{"id":"BID-001","status":"valid","tenderers":[{"id":"SUP-001","name":"Supplier A"}],"value":{"amount":95000,"currency":"USD"}},{"id":"BID-002","status":"disqualified","tenderers":[{"id":"SUP-002","name":"Supplier B"}],"value":{"amount":92000,"currency":"USD"}},{"id":"BID-003","status":"valid","tenderers":[{"id":"SUP-003","name":"Supplier C"}],"value":{"amount":98000,"currency":"USD"}}]},"awards":[{"id":"AWD-001","status":"active","suppliers":[{"id":"SUP-001","name":"Supplier A"}],"value":{"amount":95000,"currency":"USD"}}]}

This represents a procurement where the lowest bid (SUP-002 at

92,000) was disqualified, and the second-lowest bid (SUP-001 at

95,000) won.

Run the Analysis

Initialize settings

ocdscardinal init settings.ini

Prepare data

ocdscardinal prepare example.jsonl \
  --settings settings.ini \
  --output prepared.jsonl \
  --errors issues.csv

Calculate indicators

ocdscardinal indicators --settings settings.ini \
  prepared.jsonl > results.json

View results

cat results.json | jq .

You should see results indicating that R036 (Lowest bid disqualified) was triggered for this contracting process.

Common Workflows

Testing specific indicators

To test a single indicator, disable all others in your settings file by commenting out their sections:

; [R003]
; [R018]
[R036]  ; Only R036 is enabled
; [R038]

Excluding specific procurement methods

Some procedures shouldn’t be analyzed (e.g., emergency procurements). Exclude them:

[exclusions]
procurement_method_details = Emergency|Direct Award|Random Selection

Adjusting thresholds

If you’re getting too many false positives, increase indicator thresholds:

[R003]
threshold = 20  ; Increase from default 15 days

[R024]
threshold = 0.10  ; Increase from default 0.05 (5%)

Working with large datasets

Cardinal handles large files efficiently, but you can monitor progress:

# Increase verbosity to see progress
ocdscardinal indicators -vv --settings settings.ini \
  large_dataset.jsonl > results.json

Use -v for warnings, -vv for info, -vvv for debug output.

Troubleshooting

No indicators are triggered

Possible causes:

Settings file missing indicator sections: Check that indicators are enabled in settings.ini
Data quality issues: Review issues.csv - missing required fields prevent calculations
All data excluded: Check your [exclusions] configuration
Currency mismatches: Cardinal excludes processes with non-primary currencies

Run with increased verbosity to see more details:

ocdscardinal indicators -vv --settings settings.ini prepared.jsonl

Too many false positives

Adjust thresholds in your settings file for specific indicators
Add exclusions for procurement methods that shouldn’t be analyzed

Configure price comparison methods to exclude inappropriate procedures:

no_price_comparison_procurement_methods = Framework Agreement|Reverse Auction

Prepare command reports many issues

This is normal! Most issues are repetitive:

Sort the CSV by the path column to group similar issues
Add configuration to fix entire categories at once
Focus on not set errors first - these prevent indicator calculations
Iterate - run prepare → review → configure → repeat

Next Steps

Prepare Command

Deep dive into data preparation and quality corrections

Indicators Command

Learn about all indicator calculation options

Settings Configuration

Master Cardinal’s configuration system

Red Flags Reference

Understand each indicator’s methodology

Get Started

Core Concepts

Commands

Overview

Prerequisites

Complete Workflow

Step 1: Initialize Settings

Step 2: Prepare Your Data

Step 3: Calculate Indicators

Step 4: Analyze Results

Understanding the Output

Indicator Codes

Real-World Example

Sample Data

Run the Analysis

Common Workflows

Troubleshooting

Next Steps

Prepare Command

Indicators Command

Settings Configuration

Red Flags Reference

Build docs developers (and LLMs) love

Get Started

Core Concepts

Commands

Documentation Index

​Overview

​Prerequisites

​Complete Workflow

​Step 1: Initialize Settings

​Step 2: Prepare Your Data

​Step 3: Calculate Indicators

​Step 4: Analyze Results

​Understanding the Output

​Indicator Codes

​Real-World Example

​Sample Data

​Run the Analysis

​Common Workflows

​Troubleshooting

​Next Steps

Prepare Command

Indicators Command

Settings Configuration

Red Flags Reference

Build docs developers (and LLMs) love

Overview

Prerequisites

Complete Workflow

Step 1: Initialize Settings

Step 2: Prepare Your Data

Step 3: Calculate Indicators

Step 4: Analyze Results

Understanding the Output

Indicator Codes

Real-World Example

Sample Data

Run the Analysis

Common Workflows

Troubleshooting

Next Steps