Swiftscore Docs

Evaluation Scoring

Swiftscore generates a complete evaluation by scoring every component of your district's framework — Danielson by default, or a custom framework your district has configured — and producing a rating, a narrative, and supporting evidence for each one.

What gets scored

When an evaluation is generated, Swiftscore scores every component in the framework you're using, not just a subset. Each component gets:

  • A rating on the framework's scale (Danielson uses a 1-4 scale; a district-specific custom framework may use a different scale).
  • A narrative explaining the score in plain language.
  • Evidence citations that point to the specific transcript moments, observation notes, or uploaded materials the rating is based on.

Component scores roll up into domain summaries, and domain summaries roll up into an overall evaluation score.

How scores connect to evidence

Every score is meant to be defensible, not just asserted. The narrative for a component is written to cite the specific moments in the lesson — things said or done during the observation, notes the evaluator added, or supporting materials that were uploaded — that justify the rating given. The goal is that an evaluator (or a teacher reviewing the result) can trace a score back to what actually happened in the classroom, rather than treating the number as a black box.

Overall (summative) scoring

For frameworks that define a holistic scoring approach — including Danielson and OTES — the overall evaluation score is calculated holistically rather than as a simple average of the individual component scores. This reflects how many evaluation frameworks are actually meant to be scored: the overall rating considers the full picture of a teacher's performance, not just an arithmetic mean of the parts. Each framework declares how it wants its summative score calculated, so the approach can vary depending on which framework your district uses.

If component scoring isn't complete for an evaluation, a holistic summative score can't be calculated yet — Swiftscore will surface a clear message asking the evaluator to finish scoring the remaining components first.

Calibration and consistency

Scoring is designed to reflect each evaluator's calibration settings — tone, depth, and the specific terminology your district or evaluator prefers — so the language in the narrative feels consistent with how that evaluator typically writes feedback.

Scoring also takes a teacher's historic evaluation data into account. Where relevant, the narrative can reference patterns across a teacher's recent evaluations (for example, noting continued growth in a particular area since a prior observation), rather than treating each evaluation as if it exists in isolation.

Because the same underlying evidence should lead to a similar score regardless of which teacher is being evaluated, calibration is also the mechanism that helps keep scoring consistent and reduces the risk of evaluator-to-evaluator bias. Swiftscore periodically reviews a sample of evaluations across teachers with similar profiles to check for unexplained outliers in scoring patterns.

District-specific frameworks

If your district uses a customized version of a framework, that customization is layered on top of the core scoring engine rather than forking it — meaning improvements and fixes to the underlying scoring logic apply across every district's version of a framework, including customized ones.

Scoreless ("informal") evaluations

Some evaluations are configured as scoreless — typically informal observations where a formal rating isn't appropriate. In that mode, components don't display a numeric rating, but they still receive a full feedback narrative, so evaluators and teachers get the same quality of written feedback without a formal score attached.

Regenerating scores

If an evaluator wants to regenerate part of an evaluation — a single component, a full domain, or just a comment — Swiftscore can re-run scoring for that slice independently, rather than requiring the entire evaluation to be regenerated from scratch. The same calibration and evidence-citation behavior applies to a regenerated slice as it does to the original generation.

What happens when something goes wrong

Evaluation scoring is built to fail gracefully rather than break the evaluation:

  • If a scoring call doesn't return a usable result, Swiftscore retries or falls back to a safe default rather than leaving the evaluation in a broken or incomplete state.
  • If the amount of context available for a component (transcript, notes, historic data) is very large, older historic data is trimmed first, prioritizing the most recent and relevant information.
  • If a generated score somehow falls outside the framework's valid scale, it's automatically corrected and logged for review rather than displayed incorrectly.
  • If your district updates a framework's template mid-cycle, evaluations that were already in progress continue using the framework version they started with, while new evaluations pick up the updated version. This avoids retroactively changing the rules an in-progress evaluation is being scored against.

Where scores show up

Scores appear on the evaluation's Evaluation tab, with the ability to drill into a modal for each component to see the full rating, narrative, and cited evidence. Teachers see their scores and narratives on their own view of the completed evaluation.