NUSAP-lite: how we grade what we publish
The five-tag data-quality framework, and FAND's status on each tag.
Most data portals show you a number. FAND is built to show you how much to trust it as well.
NUSAP is an established framework for communicating data quality in policy-relevant science, developed by Silvio Funtowicz and Jerome Ravetz in 1990 (Uncertainty and Quality in Science for Policy). It addresses a long-standing gap in technical reporting: numerical results carry no native vocabulary for their own reliability, so consumers either treat all numbers as equal or fall back on ad-hoc footnotes. NUSAP closes the gap with five named tags carried alongside every value.
FAND calls its practice NUSAP-lite. The “lite” is deliberate: it names what has not been done yet, which is a spread on every value and the arithmetic that carries grades through sums and ratios. How NUSAP relates to the IMF’s and Open Data Watch’s assessments of official statistics is set out in Assessing National Statistics.
The five tags
| Letter | Stands for | What the user will see | Status in FAND today |
|---|---|---|---|
| N | Numeral | The value itself ($4,200 per capita) | In place |
| U | Unit | Current Atlas USD, tonnes, km², etc. | In place |
| S | Spread | Uncertainty range or confidence band (±$300) | Aggregate only; not yet per value |
| A | Assessment | Qualitative reliability rating | Recorded as Exceptions to Expectations; a trust-status grid is being built into the FAND app |
| P | Pedigree | Full provenance: source, tier, statistical adjustments | Tags on every branch, per cell on cross-border positions; 0–4 scores for cross-border positions and minerals |
N — Numeral
The point-value itself: the number FAND publishes. For a stock measure this is a level (e.g., 4.7 trillion USD of produced assets); for a flow it is a rate over a stated period. Zero and blank are different values: blank means no information, zero means measured or computed to be zero. The numeral alone tells you nothing about how trustworthy it is. That is the job of the other four tags.
U — Unit
The unit of account: currency, mass, area, count, ratio. FAND uses current Atlas USD as the default monetary unit for cross-country comparability, with PPP and constant-USD conversions available where appropriate (governed by Operational Guideline OG-004). Physical units follow agency conventions (tonnes for minerals, km² for land, persons for population). Every leaf value obeys V = Q × P, and units are checked against an independent source before a series is loaded.
S — Spread
A quantitative measure of uncertainty around the numeral: confidence interval, standard error, or range. Today FAND measures spread in aggregate only, as the gap between a source’s world total and the sum of FAND’s countries. Where a source agency publishes uncertainty, FAND will carry it through unmodified. Where FAND derives a value through statistical adjustment (e.g., ratio-splicing or the HR production function), the spread will reflect both source uncertainty and the adjustment. The proposed route, from pedigree scores to calibrated bands, is in Assessing National Statistics.
A — Assessment
A qualitative reliability rating that says whether the value should be used for cross-country comparison, country-level policy, or only as an order-of-magnitude reference. Assessment combines source tier, statistical-adjustment burden, and coverage. A Tier 1 unadjusted value will rate high; a Tier 3 modeled fallback will rate low. Completed and estimated values carry lower ratings.
P — Pedigree
The full provenance chain: source agency, table, year, source tier (Tier 1 authoritative → Tier 4 model fallback), any statistical adjustments applied, and the URL of the original data. Pedigree is the reviewer-grade audit trail that lets an institutional peer reproduce the value from the public source. FAND’s tags are being converted to 0–4 scores on a four-column pedigree matrix: source authority, temporal coverage, spatial coverage, and valuation method. It is the most detailed of the five tags and is surfaced last.
The target display
The five tags on a single value, in the order they will surface:
Default view: $4,200 / pc ← N + U
On hover: $4,200 ± $300 / pc ← + S
On click: Assessment: HIGH ← + A
Source detail: USGS MCS 2026 Table 1 → Tier 1 ← + P
(ratio-spliced to 2025 base year per OG-002)
https://www.usgs.gov/centers/...
Casual readers see only N and U. Institutional reviewers reach the full pedigree in one or two clicks. The figures above are illustrative.
Why it matters
For institutional outreach (IARIW reviewers, statistical-office peers, multilateral-bank analysts), NUSAP is a signal of rigor: opinions about uncertainty are written down, not left implicit. For policy users downstream, it is a signal of honesty about what the data community knows less well. Both audiences benefit; neither has to pay for the other’s needs.
Further reading
- Assessing National Statistics: NUSAP beside the IMF’s DQAF and Open Data Watch’s ODIN, and FAND’s next steps.
- Funtowicz, S. O., & Ravetz, J. R. (1990). Uncertainty and Quality in Science for Policy. Kluwer Academic Publishers.
- Costanza, R., Funtowicz, S. O., & Ravetz, J. R. (1992). “Assessing and communicating data quality in policy-relevant research.” Environmental Management 16(1): 121–131.
- van der Sluijs, J. P., et al. (2005). “Combining quantitative and qualitative measures of uncertainty in model-based environmental assessment: the NUSAP system.” Risk Analysis 25(2): 481–492.