Why did a short take get no score instead of a low one?

sayit will sometimes show your accuracy, fluency, and every per-word verdict, and still leave the headline overall score blank. That's not a bug — it's a deliberate rule: below a minimum sample of scored words on a passage you didn't finish, sayit declines to publish an overall verdict at all, rather than hand you a number that's technically correct and practically misleading.

30-second version: If you scored fewer than 12 words on a reference-mode passage and completed less than 80% of it, sayit withholds the overall score and band, and says why. Everything else — accuracy, fluency, completeness, and every word's individual verdict — is still shown exactly as measured. Only the single blended number that would otherwise summarize "the whole take" in one figure gets held back, because at that sample size it doesn't actually summarize the whole take.

What actually triggered this rule?

A real case: a take that read nine words out of an 1,832-word passage and stopped. Accuracy on those nine words happened to be low, and blended with fluency the standard way, it produced a headline of 48 out of 100 — technically correct arithmetic, and a genuinely dishonest headline. A number built on half a percent of a passage was being presented exactly the same way as a number built on the whole thing, with nothing on screen to mark the difference. The fix isn't to silently adjust the math; it's to recognize when the sample is too thin to summarize at all and say so plainly.

What are the actual thresholds?

Two conditions have to both be true for the headline to be withheld: fewer than 12 words were actually scored, and you completed less than 80% of the passage. Both matter — a one-word flashcard drill or a short shadowing line is complete by design and still gets a full score, because there's nothing unfinished about it. It's specifically the combination of thin and abandoned that triggers withholding the number: a short exercise done in full is honest; a long passage barely started is not.

What still shows up when the headline is withheld?

Everything except the single blended number. Your accuracy on the words you did read, your fluency reading, your completeness percentage, and every individual word's verdict remain exactly as measured — you still see precisely what those nine words sounded like. What disappears is the one figure that would otherwise imply "here is how the whole take went," because at that sample size, it can't honestly say that.

Why not just show the number with a warning label instead of hiding it?

Because headline numbers get shared, screenshotted, and remembered without their footnotes. A 48 next to your name travels further than the caveat underneath it. sayit's own shared result links enforce this same rule on the public-facing side — a share built on too little never displays a fabricated-looking score, whether the sharer is the same person who took it or someone who clicked a link.

Does this apply outside read-aloud passages?

No — it's specific to reference mode, because it depends on comparing words scored against words expected in a known passage. Freeform speech has no fixed passage length to measure completeness against, so an empty or near-empty freeform take is instead reported with its own "no speech" reason rather than this rule.

When the rule fires, at a glance

Try it

Read a longer passage in sayit and stop a few words in on purpose — you'll see per-word results without a headline number, with a short note explaining why. Finish a passage, or a short drill, and the headline returns.