Our Learning-Flow Data: 80% Correct, and Still 80% a Day Later

Every study app says its method works. Few show the numbers. This is our first report from NurseSavvy’s own student data, and it answers two narrow questions about learning flows, the short card sessions that teach one nursing topic at a time: are the cards at the right difficulty, and do students still know the material later?

The short answers: students answer about 80% of learning-flow cards correctly, which is the range we design for, and they answer about 80% correctly when a card returns a day or more later. The data is as of October 14, 2026. The method, sample sizes and limits are below and on our research page.

Finding 1: cards land in the productive difficulty range

Across 20,919 answers from 183 students, learning-flow cards were answered correctly 80.1% of the time. On a student’s first try at a card, the rate was 77.3%.

Why that number is the goal and not 100%: practice that is almost always right feels good but asks little of memory, and practice that is mostly wrong teaches frustration. Learning researchers call the useful middle “desirable difficulty”: hard enough to need real effort to recall, easy enough that the effort usually pays off. We aim learning flows at roughly 75–85% correct.

That target is our design choice, not a proven threshold for nursing students. The often cited “Eighty Five Percent Rule” came from a computational model, not a nursing trial, and we explain what it does and does not support here. What the data shows is that the cards hit the range they were built for.

One note on counting. Some cards only ask the student to read a short explanation and tap continue, so they are always “correct”. We left those out. Counting them would raise the headline to 82.9%, which flatters the cards without measuring anything.

Where we are off target

The same data shows where cards miss the range, and we are fixing them. Two card types land well below it: cards where students place a value on a sliding scale (56% correct, about 500 answers) and cards where students build an answer from pieces (62% correct, about 600 answers). Some of that is the format, not the nursing: placing a lab value precisely on a slider is harder than picking it from a list. Either way, a card below the range is a card to rewrite, and the next report will say whether the rewrites worked. At the other end, simple two-choice cards run near 87%, a little easier than they need to be.

Finding 2: students still know it a day later

Learning flows bring cards back on a spacing schedule: a few hours later, then a day, then longer gaps as the student keeps getting them right. The test of learning is what happens when the card returns after the student has had time to forget.

When a card returned more than a day after the student’s last attempt, they answered it correctly 79.8% of the time (2,202 answers from 43 students). That is slightly above the 77.3% first-try rate. The facts held up after the delay instead of fading.

Misses teach, too. After a miss, the student sees a short explanation and the card comes back. On that return, students answered correctly 91.2% of the time (3,609 answers from 155 students).

The caveat: cards come back for review on a schedule that depends on how the student did before, so this measures how well reviewed cards were remembered, not every card a student ever saw. Only 43 students have reached day-plus reviews so far, and we will report the number again as that grows.

What we are not claiming

• This is internal product data, not a peer-reviewed study or a randomized trial. Students chose to use NurseSavvy and chose how much.
• Card accuracy measures recall on our own cards. It is not an NCLEX score and does not predict passing.
• We have not yet published whether learning flows raise question-bank scores or readiness exam scores. We are measuring both, and we will publish them when there is enough data for the result to be statistically significant, whichever way it comes out.

Why publish this

Because “it works” means nothing without the numbers, and because publishing them keeps us honest. We will re-run these figures on a fixed schedule (next: November 2026) and keep every metric on the research page whether it goes up or down. Our question-difficulty data is published the same way.