How we measure, exactly

The free reading check finds a starting step, not a grade or a label. It runs in parts and adapts as the child reads:

• What is checked. Letter sounds, real word reading, nonsense word reading (we call them robot words, and every word is explicitly tagged real or robot, never guessed by position), then an adaptive survey of pattern layers: blends, digraphs, silent e and soft c and g, endings and two syllable words, tch and dge and y and consonant le, r controlled vowels, vowel teams, diphthongs, and affixes, five words per layer.
• Correct means read correctly on the child's own attempt. The adult marks every item; a near miss counts as a miss.
• Stop rules. The check stops early the moment the signal is clear: fewer than ten letter sounds places the child at the very first step; under 60 percent on real short words stops the check and places at the first unknown letter sound; a pattern layer passes at 4 of 5, and two consecutive failed layers end the survey. Older readers who read fluently skip ahead through an advanced gate rather than grinding through material below them.
• Tired is respected. Once enough signal exists, every screen carries a stop here and use a safe starting level escape. An early stop still produces a valid, conservative placement.
• Placement is conservative by design. The child is placed at the first wobbly layer, and any layer the survey did not reach is treated as untested, never assumed known. When two starting points are defensible we take the easier one on purpose, and if the first two weeks are perfect the program offers to accelerate.
• Age and starting level are recorded separately. Age sets the story register and session shape. What the child actually read sets the starting step. The two never contaminate each other.
• Reuse window. A completed check is honored for 30 days, so a family who checks tonight and signs up Thursday never repeats it. Children under five skip the check entirely and route to the Littles track, no tests at three and four.

• Cadence. A baseline probe in the very first session, then every fifth session, roughly weekly at the recommended dose. Probes never run on a level mastery day.
• Task. Sixty seconds of connected text reading from the current level's exit passage, which is decodable at the child's level, so the score reflects taught skills rather than guessing.
• Scoring. Words correct per minute is computed as words correct divided by actual elapsed seconds, times sixty. Errors are words the adult marked, and the adult's judgment is always the last word: in tap mode the adult marks the last word reached and any misses directly, and in listen mode every automatic mark can be overridden by the adult, token by token.
• Self corrections. In tap mode the adult can unmark a word the child fixed. In listen mode a flagged word stays flagged. The protocol treats the adult's mark as the record.
• What is stored. Each probe stores the session number, the date, words correct per minute, and accuracy. The child sees their own growth line and never a percentile.

• The norm table. Hasbrouck and Tindal 2017 oral reading fluency national norms, grades 2 through 6, by season (fall, winter, spring). Sources are on the evidence page.
• The goal is the 50th percentile. Fast enough. We do not chase speed past it, and the paced reread feature is capped at the grade level 50th percentile by design, so a child is never pushed to race.
• Accuracy comes first. The reporting bar is 96 percent of words read correctly. Below that bar, the weekly note tells the parent plainly that we feed accuracy, not speed, and growth in rate is not celebrated until accuracy holds.
• Young readers are protected. Grade 1 and below are never norm compared, anywhere. Readers above grade 6 are compared against the grade 6 ceiling, with a plain reading level label instead of a pretend norm.
• Norms are parent facing only. A struggling reader sees their own progress, never a national comparison.
• The validity limit, stated plainly. National norms only fully apply when a child reads an unseen passage at their actual grade level. Our weekly check uses a decodable passage at the child's instructional level inside the program, which is the right text for measuring growth, and the wrong text for declaring a national percentile. So the growth trend is the number we stand behind, the norm band is context, and for a formal grade-level fluency score we point families to the school's own benchmark screener, run three times a year at no cost.

• Step mastery. A session's mastery percentage pools word work, story reading, and dictation. A step is mastered only after two consecutive sessions at 90 percent or better, and any session below 90 percent resets the streak. Nothing advances on one good day.
• Level exit gates. Finishing a level's last step triggers a mastery check day: real words at 90 percent, nonsense words at 80 percent, dictation at 80 percent, and the exit passage at 95 percent, all four at once, passed twice in a row, before the next level opens.
• When mastery stalls, the program acts instead of repeating. Four practice days without mastery triggers focus sessions (a shorter set, 80 percent known material). Three focus sessions without mastery drops back one step. Still stuck triggers a phonemic awareness micro unit. Beyond that, the escalation flag below.
• Missed words never disappear. Every miss enters the spaced review queue and returns until it is beaten twice.

A home program owes families the truth about when home practice is not enough. The program raises a professional evaluation flag, in writing, in the session summary and the weekly note, when any of these fire:

• Flat growth under full support. The child has at least eight fluency probes and has already been through the full stall ladder, and the average of the last three probes is no more than one word per minute above the average of the first three.
• Age to placement gap. A child eight or older still working in Level 1, or eleven or older still at Level 3 or below.
• Persistent phonemic awareness difficulty. A child flagged for intensive sound work whose ear work stays under 60 percent across six or more of the last ten sessions.

The message recommends a free, formal school evaluation (every U.S. family's right under IDEA Child Find), in plain words, with no diagnosis and no upsell. That is a firewall rule: the escalation message never sells anything.

• The adult's judgment is the source of truth. Every reading mark in this program is a human decision. No machine ever grades the child's voice into the record.
• Speech recognition is a convenience, never a judge. Where listening assistance exists in the fluency probe, every automatic mark is adult overridable, and the whole feature degrades to manual marking when unavailable. We removed machine listening from story reading entirely, because browser speech recognition autocorrects a child's misreading into the right word, which makes it flattering and useless as a judge.
• Answers are concealed from the child mechanically. In ear work (phonemic awareness), the expected answer is blurred on screen until the adult taps to peek, so a shared tablet cannot leak answers. In dictation the protocol is screen faces the adult, and in the teen's solo mode the dictation target is mechanically hidden behind a blur until the reader chooses to check.
• Practice and measurement are kept distinct. Teaching moments carry scaffolds on purpose: the warm up models a missed sound only after the miss, never before, and word work may show the day's pattern hint while the child practices it. The clean measures are the fluency probes and the level exit gates, which carry no scaffolds.

• Dictation. Marked right or needs a fix per item by the adult, with every miss recycled into spaced review, and a fix it beat immediately after any miss. On probe nights a three word spelling snapshot is recorded separately and never counts toward mastery.
• Retell and comprehension. After the story, a retell moment and a one tap comprehension note (yes, almost, not yet), logged over time in the weekly note as an at home indicator, and labeled exactly that: a simple note, not a formal comprehension test.
• Confidence. A five item reading self concept check in, at most every 60 days, because a child who stops hiding from books is a real outcome too.