What we measure, and what we don't.

By Yechiel · Founder, PD That Works. Yechiel builds PD That Works — the workshops, the daily AI coach, and the monthly faculty reports — and wrote every page on this site.

Most PD evidence fails in one of two directions: it measures the workshop instead of the classroom, or it measures the classroom with instruments that cannot reach it and dresses the result up as proof. This page is the inventory — every kind of evidence this product produces, labeled honestly, with the claims we will not make listed alongside.

Every number carries a type

Internally, every metric and report element is tagged with what kind of evidence it is. Those labels are now visible on the leader's dashboard too, next to the sections they describe, because a number whose provenance is invisible invites the reader to assume the strongest possible reading.

The bottom row is the important one. Nobody in this category of product can currently give you externally validated outcomes from a 30-day cycle, and we would rather say so on a public page than let the word "proof" do quiet work in a sales conversation.

What gets captured, and when

Day 1 — the baseline. One goal in the teacher's own words, how often they report having used the practice in the prior five teaching days, a confidence rating on a stable five-point scale, and a sentence of classroom context. Only what will actually be used later. Without this, nothing that follows is movement.

Days 2–29 — the record. Whether the check-in happened, what the teacher reported trying, whether they adapted the practice or set it down, whether they asked for help, an optional weekly confidence rating, and anything they chose to share.

Day 30 — the close. The same behavior-frequency and confidence questions asked at baseline, so the comparison is like-for-like; whether the practice was refined, retained, replaced, or abandoned; whether the coaching was useful; whether the teacher would continue; and an optional comment.

How the report is allowed to talk

The reporting language is constrained, and the constraint is enforced in the software rather than left to whoever writes the summary.

The right-hand column is not a list of things we find distasteful. Each one is a claim this data cannot support: an inference about a classroom nobody observed, a causal claim with no counterfactual, a judgment of teachers dressed as a finding. And low engagement is never reported as resistance — it may as easily be workload, absence, a poor fit, or a teacher quietly deciding the goal was wrong.

Where a number is withheld

Some sections of a report come back deliberately empty.

• Below five teachers, no themes and no confidence trend. A pattern drawn from three people is those three people, identifiable to anyone who knows the staff. Within a qualifying cohort, an individual theme additionally needs at least three teachers behind it.
• No direct quotation unless a teacher explicitly chose to share those words.
• Exact counts are softened to ranges where a precise number in a small cohort would re-identify someone.

Each of these makes the report less impressive and more trustworthy, which is the trade we would make every time. A blank section says the floor held.

The honest limits, stated plainly

• Self-report is not observation. A teacher telling you she used wait time is evidence that she believes she did, gathered daily rather than recalled at year-end. It is the same instrument a coach uses in a sit-down. It is not a rubric score and should never be presented as one.
• Confidence is direction, not attainment. A rating that moves from 2 to 4 says something changed in how a teacher reads her own practice. It does not convert into a skill level.
• Thirty days is thirty days. One cycle shows whether the practice took hold. A trend needs a second month; a defensible picture needs a third.
• Student outcomes are out of scope — not because they don't matter, but because a 30-day report claiming them would be lying, and the claim would be untraceable to anything in the data.

What this is good for

Within those limits the evidence answers the question a board actually asks, at the level it can honestly be answered in a semester: teachers engaged with follow-up at this rate, here is what they reported working on, here is how their confidence on their own goals moved, and here is what they chose to share about it. That is a real answer about the weeks after the workshop, which is more than an attendance sheet has ever offered.

Common questions

Does this measure whether teaching actually improved?

It measures what teachers report trying, how consistently they engaged, and how their confidence on their own goal moved. That is self-report and telemetry, not observation. A report claiming observed instructional improvement from this data would be overstating it, so ours does not.

Do your reports include student outcomes?

No, and each report says so in its own text. Those effects take longer than a cycle and are entangled with everything else a school does.

Can I add my own observation data?

A later release will allow one cohort-level measure a school already collects — a brief walkthrough rubric, an artifact count, an existing survey item — kept separate from private coaching data and labeled as leader-entered. It is not available today.

Why is part of my report blank?

Because the cohort was too small for that section to be shown without identifying someone. The floor is five teachers for themes and confidence trends. The report names the suppression rather than filling the space.