# nairarate.dev — methodology

How these rates are derived, so you can decide whether to trust them.

## Sources

Official   Central Bank of Nigeria's own JSON endpoints. USD uses the NFEM
           volume-weighted average, CBN's stated official rate for the day, and
           carries high/low/close and market turnover.
Parallel   Every changer quoting the pair via Monierate, aggregated here rather
           than taken pre-blended.

## Screening

Raw provider lists cannot be averaged. Unfiltered, USD returns a mid of 1181
against a true ~1392. Four exclusions:

1. A quote of 0 means "no quote on that side", not a rate of zero. Remittance
   corridors publish only a send rate — 19 of 65 USD providers. Averaging the
   zero halves the mid.
2. The central bank itself quotes inside the provider list. Including it drags
   the parallel rate toward the official one and understates the spread, which is
   the number this API exists to publish.
3. Reference and institutional feeds are not street rates: one quotes bid equal
   to ask, another a mechanical 0.50% band around a reference mid.
4. Outright bad data. One feed quoted GBP at 130 against a ~1850 market.

Survivors are aggregated by median, not mean — robust to the long tail of
outlying changers. A quote more than 3x from the peer median is discarded.
provider_count reports the surviving two-sided quotes, so a rate resting on a
single provider is visible as such.

Official history is screened against neighbouring days, which catches genuine
CBN data-entry errors (a Danish krone printed at 198,024 against a ~200
baseline). Two-sided bands are reconciled against the mid, which CBN publishes
directly; where a band is incoherent the offending side is dropped and the mid
kept.

## Validation

Our screened USD/NGN parallel mid, 1392.47, against Monierate's own independently
computed composite, 1397.82 — a 0.4% variance between different methodologies on
the same market.

Full detail: https://nairarate.dev/llms.txt