Identifying Good Data Sources (ROCCC Framework)#

📦 Data Preparation ⚖️ Bias & Data Ethics Lesson 011

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

A checklist for trust#

Before building on a dataset, an analyst must judge whether the source is trustworthy — and a memorable checklist captures what “trustworthy” means. Good data sources are ROCCC: **R**eliable, **O**riginal, **C**omprehensive, **C**urrent, and **C**ited. Running a candidate source through the five letters turns a vague sense of trust into specific, checkable questions.

The five criteria#

Reliable — is the data accurate, complete, and unbiased, from a source with a track record of getting things right? Reliable data comes from well-regarded, vetted origins, not anonymous or casual ones.
Original — can you get to the source of the data, rather than a copy of a copy? Original data comes from the party that actually collected it; each hand it passes through is a chance for distortion, so validating against the origin matters. (This echoes the first/second/third-party gradient from earlier.)
Comprehensive — does the data contain all the information you need to answer the question? A source missing critical fields or covering only part of the population, however accurate on what it has, cannot fully answer the question.
Current — is the data recent enough to be relevant? Data ages; a source from years ago may no longer reflect present reality, and the right recency depends on how fast the subject changes.
Cited — is the source of the data documented — who created it, when, where, and how? Cited data can be traced and vetted; uncited data asks you to trust it blindly.

A source that satisfies all five is a strong foundation; the more it satisfies, the more weight your conclusions can bear.

Using ROCCC in practice#

ROCCC is a pre-flight check for the Prepare phase: before committing to a dataset, walk the five letters and note where it is strong and weak. A source need not be perfect on every criterion — few are — but the framework makes the trade-offs explicit. “This data is comprehensive and current but not fully original, since it’s a third-party aggregation” is a clear-eyed basis for proceeding with appropriate caution, and for stating that caution in the final analysis.

Why a checklist beats a gut feeling#

The value of ROCCC is that it makes source evaluation systematic rather than intuitive. Under time pressure, it is tempting to grab whatever data is at hand and start analysing — and the resulting problems (unreliable, stale, or uncited data) surface only much later, after conclusions have been built on sand. Five questions at the start are cheap insurance against that.

The caveat#

ROCCC evaluates the source, not the fit to your specific question. Data can be impeccably reliable, original, comprehensive, current, and cited — and still be the wrong data for your problem, or carry sampling bias the framework does not directly test. Use ROCCC together with the relevance and coverage judgements from the choosing-data lesson: source quality and question-fit are separate checks, and a good analysis passes both. The next lesson applies ROCCC in reverse, to recognise the sources you should not trust.

Hint