Understanding NFIQ Scores for EFT

Every fingerprint that goes into an EFT file has a quality score attached to it. That score is called NFIQ (NIST Fingerprint Image Quality) and it is not a marketing rating. It is an algorithm developed by the National Institute of Standards and Technology that predicts how likely a given fingerprint image is to match correctly when the FBI's biometric system processes it. NFIQ does not decide whether a submission is accepted, but it does tell you where the weak prints on your card are while you still have time to act on them. This guide explains the scoring system, what each score means in practice, and how the scores are used.

What NFIQ actually is

NFIQ is open-source software, published by NIST, that takes a fingerprint image and outputs a single number describing how usable that image is. The original version, NFIQ 1, was developed in 2004 to "produce a quality value from a fingerprint image that is directly predictive of expected matching performance" (NIST Biometric Quality program). The current version, NFIQ 2, is the reference implementation of the international biometric image-quality standard ISO/IEC 29794-4.

The score is not a measure of how the print looks to a human. It is a measure of how well the algorithm thinks an automated matcher will be able to extract minutiae (the ridge endings and bifurcations the FBI's CJIS system uses to identify a print) from the image. A print can look reasonable to your eye and still score poorly because it is smudged at the cores, missing ridge edges, or scanned below the required resolution.

NFIQ runs the same way regardless of who captured the print. A livescan kiosk, a wet-ink card scanned at home, a print rolled in 1995 and digitized last week: they all get scored on the same scale, by the same algorithm, against the same model.

The NFIQ 1 scale: 1 to 5, lower is better

The original NFIQ algorithm assigns one of five integer classes to a print:

Score	Class	What it means
`1`	Excellent	High confidence the print will match.
`2`	Very Good	High confidence; minor imperfections.
`3`	Good	Adequate for matching; some ridge detail loss.
`4`	Acceptable	Usable but degraded; matching error rates rise.
`5`	Poor	Not recommended for enrollment.

Lower is better. A 1 is the best possible score; a 5 is the worst. This is the inverse of the more common "higher number is better" convention, which is the source of half the confusion around NFIQ on forums.

NFIQ ships as part of NBIS, the NIST Biometric Image Software package. NIST originally developed this package for the FBI to support fingerprint capture, segmentation, classification, and quality scoring. NBIS is the same toolkit that underlies the FBI's own internal quality and minutiae-extraction tooling. NFIQ 1 is the quality module inside it.

NFIQ 1 scores are derived from an 11-feature analysis of the fingerprint image, fed into a small neural network trained against matcher performance on a labeled dataset. The five-class output is intentionally coarse: the algorithm was designed to give a clear, fast, integer verdict that capture stations and downstream systems could act on without ambiguity. That coarseness is also why it has held up: a 1, 2, or 3 means the same thing in 2004 and in 2026, on a livescan capture and on a wet-ink card scan.

The NFIQ 1 score is written directly into your EFT file. The FBI's biometric pipeline reads this score, and the NBIS reference implementation of NFIQ 1 is still the algorithm most EFT-generating tools, including SLAP & ROLL, produce scores from.

NFIQ 2: The newer 0 to 100 scale

NFIQ 2 is a newer NIST quality algorithm, released as a separate package from NBIS. NIST maintains it as open source on GitHub and publishes the technical specification in NISTIR 8382. It does not replace NFIQ 1 in the FBI's biometric pipeline. Both algorithms are accepted in EFT submissions, and both serve the same fundamental purpose. NFIQ 2 is best understood as a higher-resolution alternative for systems that need finer-grained scoring.

Three things differ from NFIQ 1:

Score range. NFIQ 2 outputs a value from 0 to 100, where 100 is the best possible quality and 0 is the worst. This is the opposite direction from NFIQ 1, where lower is better. Always check which version a score belongs to before interpreting it.
Standardization. NFIQ 2 is the reference implementation of ISO/IEC 29794-4, which means a score produced by any conformant implementation should be comparable to a score produced by another. NFIQ 1 predates that standard.
Resolution and constraints. NFIQ 2 was developed specifically for 500 PPI plain-impression prints captured by optical sensors or scanned from inked cards. That constraint is built into the algorithm. Running NFIQ 2 on a print scanned at 300 PPI or 1200 PPI produces a score, but it's not a score the algorithm was calibrated to.

NFIQ 2 scores and NFIQ 1 classes are not directly interchangeable. They are produced by different algorithms calibrated against different reference sets, and NIST does not publish an official conversion between the two. A high NFIQ 2 score and a low (good) NFIQ 1 score generally agree on which prints are good and which are not, but the specific numbers don't map one-to-one.

For ATF eForm submissions, the NFIQ 1 score is widely accepted and understood. NFIQ 2 adds resolution, but it does not change the underlying question (will this print match?) that NFIQ was designed to answer.

What NFIQ score you actually need for an ATF eForm submission

The honest answer: there is no published numeric threshold. The ATF eForms portal does not reject a file because it carries an NFIQ score of 4 instead of 3. The matcher on the FBI side either resolves the prints against its database or it doesn't, and an NFIQ score is a prediction of how likely that resolution is, rather than a hard pass/fail gate.

How to think about each score in practice:

NFIQ 1 scores of 1–3 are reliably matchable. A card where every finger scores 3 or better is in good shape and there is nothing to act on.
NFIQ 1 score of 4 is fine. A 4 here and there on an otherwise clean card is normal and not a reason to re-roll. It is worth being aware of, especially if a lot of fingers come back as 4s, but it is not a problem in itself.
NFIQ 1 score of 5 is the lowest class on the scale. We have seen submissions with multiple 5s routinely accepted, but we encourage applicants to try and get a reasonably clean card completed. If you can easily re-roll a finger that returned a 5, doing so is good practice in case stricter quality requirements are enforced in the future.

The reason there's no hard cutoff is that the FBI's matcher looks at the prints, not the per-finger NFIQ value. A 5 on a small-area finger like a pinky can still carry enough usable minutiae to match. A 1 on a thumb that was rolled cleanly will almost always match. NFIQ is a useful predictor; it isn't the decision.

Aim for high quality. Don't fret over a few low scores. The goal of showing the score before you pay is to give you the option to act on the ones that are easy to fix, not to send you back to the rolling pad over a single 5.

For the FBI's own guidance on what makes a recordable print, see Recording Legible Fingerprints. For a practical walkthrough of evaluating your card before submitting, see How to tell if your fingerprint rolls are good enough for ATF.

How SLAP & ROLL computes and shows your scores

SLAP & ROLL computes an NFIQ score on every fingerprint we extract from your scanned FD-258 card, and shows the score before you pay.

The algorithm is the same one the FBI uses. We run NIST's NBIS library, the open-source biometric image software NIST developed for the FBI, and use its NFIQ 1 implementation directly. We do not invent a proprietary "quality score" and dress it up with a thumbs-up icon. The number you see on the dashboard is the same number that gets written directly into your EFT file and read by the FBI's downstream systems.

We use NFIQ 1 specifically because the five-class scale is the right resolution for the decision the applicant is actually making. The question is "is this finger in the green, yellow, or red bucket" rather than "does this print score a 73 or a 76." A precise 0–100 number on every finger sounds more rigorous than five integer classes, but for the purpose of deciding whether to re-roll a finger, integer classes are more useful: they put each print in a bucket the reader can act on at a glance.

The dashboard renders scores per finger, color-coded:

Green: NFIQ score of 1, 2, or 3. Reliably matchable. Nothing to act on.
Amber: NFIQ score of 4. Acceptable. Worth being aware of, but not a reason to re-roll on its own.
Red: NFIQ score of 5. The lowest bucket. We routinely see submissions with 5s accepted, but if you can easily re-roll a red finger, doing so is good practice to ensure future compatibility.

Hover over any fingerprint on the processed card image to see the exact score and the finger position it corresponds to.

You see all of this before paying. If a few fingers come back red and re-rolling is easy, it's worth doing. If your card looks generally good with an outlier 5 on a pinky, that's normal. The score is a tool, not a verdict.

What to do if a finger comes back as a 5

A 5 is the lowest NFIQ class, but it is not a rejection notice. Submissions with multiple 5s are routinely accepted. However, we encourage trying to get a reasonably good card completed to ensure future compatibility. If a finger can easily be re-rolled, doing so is good practice.

Two practical paths:

Re-scan at a higher resolution

If the original scan was below 500 PPI, the score may be reflecting scan quality more than print quality. Re-scan at 600 PPI or higher and re-upload. SLAP & ROLL will recompute scores on the new scan, and an under-resolved scan often clears up several fingers at once.

Roll a fresh card

Fingerprints on an FD-258 cannot be re-rolled in place; once a print is on the card, it is there. If you want to redo a finger, the practical move is to roll a fresh card. If the original card is mostly good but has one or two 5s on fingers you can clearly do better, a second card with those fingers re-done is the cleanest fix. If the whole card is scoring poorly, the issue is usually ink quantity, pad type, or rolling technique: see the rolling guide.

If you have a generally clean card with a 5 on a pinky or another small-area finger, the practical answer is usually to submit it. Pinkies score 5 for a lot of people because there's less ridge surface to read, and the FBI matcher takes that into account.

Aim for high quality. Don't let a single low score become a multi-week delay you imposed on yourself.

Why most services don't show you the score

This is the part most readers find surprising. NFIQ scoring is a NIST-published, open-source algorithm. Any service that processes fingerprint images can run it. The reason most kiosks, mail-in services, and even some livescan operators do not show you the score is structural rather than technical:

Kiosks are designed to capture, package, and submit in a single transaction. Showing a quality score creates a decision point at the counter that vendors generally want to avoid. If the applicant sees a 5, they ask why, and the kiosk operator becomes responsible for re-rolling on the spot.
Mail-in services receive your card, generate the EFT, and email you the file. By the time a quality score would be useful (before paying), the card is already in their possession and the transaction is already complete.
Livescan operators typically see the score on their internal capture software but do not surface it to the applicant. The trade is convenience for transparency.

We made the opposite trade. Quality scores are computed and shown before the conversion fee is charged. Re-roll, re-scan, or re-upload until the scores look right. Pay once, when you're confident the file is going to clear.