Skip to content
ESSA evidence · Tier 4

How Zelvi is meant to help, and how we check.

The logic model behind Zelvi, the research it rests on, and our plan for studying whether it works. Published to meet ESSA Tier 4, “demonstrates a rationale”.
How a pilot runs

What Tier 4 means

Under the Every Student Succeeds Act, Tier 4 means a product has a well-specified logic model informed by research, and an ongoing effort to study its effects. It is not a claim that Zelvi has been shown to raise results. We will update this page when a study reaches a higher tier.

The problem

Mistakes that never get fixed.

When a student gets an idea wrong, they usually move on without fixing it, and the teacher finds out at the test. By then the class has moved on too.

  1. Inputs

    • The instructor’s own course material
    • A list of common mistakes for each idea, reviewed by teachers
    • Class lists from the LMS or school system
    • Zelvi: a tutor that explains with pictures and voice

  2. Activities

    • Students practise from the course material
    • On a wrong answer, Zelvi names the mistake and re-teaches it from the material
    • A hint comes before any answer, and the student tries again
    • Students still stuck are passed to a teacher or tutor, with their attempts attached
    • Teachers see who is stuck on what, and send a short re-teach to those students

  3. Outputs

    • Practice attempts and retries after a re-teach
    • Mistakes matched to the reviewed list
    • Hand-offs to people
    • Re-teach checks sent by teachers

  4. Short term

    Within weeks

    • More stuck students get it right on their own next try
    • Fewer students repeat the same mistake
    • Teachers know who needs help before the test

  5. Mid term

    Within a term

    • Higher scores on the unit tests for the targeted ideas
    • Fewer students needing remediation

  6. Long term

    Within a year

    • Higher course pass rates
    • Fewer students dropping, failing or withdrawing

What the model assumes

  • Students learn more when they still do the thinking, so Zelvi gives a hint first and keeps the answer hidden after a first miss.
  • Feedback works best when it is about the specific mistake and comes while the idea is still fresh.
  • Teachers act on a short, clear list of who is stuck and on what.

The research

What the model rests on.

The study plan

How we find out if it works.

Design
Each pilot compares a course with the same course in an earlier term, or with a section that did not use Zelvi. The comparison is agreed with the institution before day one.
Measures
The measures below, defined in writing before the pilot starts. Every number is shown with its sample size.
Timing
Class map at week 2, evidence report at week 6, course outcomes at the end of term.
Independence
Institutions can export their data for their own research office or an outside evaluator. We publish results whether or not they favour Zelvi.
Next level
Once enough pilots have finished, we will run a study with a comparison group and statistical controls, aimed at ESSA Tier 3 (“promising evidence”).

What every pilot measures

First-retry recovery
Share of struggles resolved on the student’s next attempt.
Eventual recovery
Share resolved within the term, and how long it took.
Repeat misconception
Students who made the same mistake again after help.
Hand-offs to people
How many reached a tutor or instructor, and what happened next.
Material gaps
Questions your material could not answer, by topic.
Course outcome
Assessment scores and drop / fail / withdraw rate against your baseline.

Want to see the report a pilot produces? See a sample week-6 report.