Low-inference observation notes record what happened in a lesson without judging it: what the teacher said, what students did, how many, and when. "Students were disengaged" is an inference. "9:14 - 6 of 24 students had heads down during independent practice" is a low-inference note. The notes come first, and the rating and feedback come after, built on evidence the teacher can't dispute.
What are low-inference notes?
Low-inference notes are a running record of observable facts: direct quotes, actions, counts, and times. They leave out adjectives, interpretations, and guesses about intent. TNTP's Core Observation Guide describes the goal as objective statements that tell the story of the lesson without reflecting what the observer thought of the lesson.
The "low" matters. All note-taking involves some choice about what to record, so no notes are inference-free. The aim is to push the judging to after the lesson, when you have the whole record in front of you and the rubric open beside it.
What is the difference between observation and inference?
An observation is something a camera and a microphone would have captured. An inference is what you concluded from it. "The teacher asked, 'Why do you think the author chose that word?'" is an observation. "The teacher asked higher-order questions" is an inference, even when it's a fair one.
Inference tends to slip into notes in four predictable ways:
- Adjectives. Good, strong, effective, confusing, slow. Every one of them is a rating in disguise.
- Vague quantities. "A lot of students" or "most of the class." TNTP's guide names vague quantifiers as one of three pitfalls of running records, alongside opinion statements and education jargon.
- Mind-reading. "The teacher was frustrated" or "students didn't care." You saw a tone of voice or a slumped posture. Record that.
- Jargon as shorthand. "Good scaffolding" or "strong differentiation" names a rubric category without the evidence that would place the lesson in it.
Low-inference observation notes examples
Each pair below shows the same moment written two ways. The first version is what most principals write under time pressure. The second is what a teacher can't argue with and a rubric can actually score.
- Instead of: "Students were disengaged during independent work." Write: "9:14 - 6 of 24 students heads down. 4 of those 6 had a blank page at 9:20."
- Instead of: "Great rapport with students." Write: "Greeted 19 of 22 students by name at the door. Asked one student, 'How did the game go Saturday?'"
- Instead of: "The directions were confusing." Write: "10:02 - Directions given verbally, not posted. 10:04 - 5 students asked a neighbor what to do. 10:05 - Teacher repeated directions to the whole class."
- Instead of: "Students understood the concept." Write: "Exit ticket: 17 of 23 solved the two-step equation correctly. 4 of the 6 errors were the same sign mistake."
- Instead of: "The teacher was frustrated with the back table." Write: "Teacher said, 'Back table, I've asked three times,' louder than prior directions. Back table quiet for 2 minutes."
- Instead of: "Good use of wait time." Write: "Asked, 'What would happen if we doubled the mass?' Paused about 6 seconds. 4 hands went up after the pause, 1 before it."
- Instead of: "The lesson moved too slowly." Write: "Warm-up 9:00-9:22. Mini-lesson 9:22-9:31. Independent practice 9:31-9:40. Class ended 9:40."
- Instead of: "Strong differentiation." Write: "Three versions of the reading packet, labeled by color. The group with the shortest text finished at 10:15 and started the extension task without prompting."
Notice that none of the low-inference versions says whether the moment was good or bad. Some clearly point one way, but the judgment is left to the rubric and the conversation. For four more pairs focused on rating disagreements, see the evidence-versus-inference section of our guide to calibrating observation ratings.
What should a low-inference notes template include?
A template doesn't need columns or boxes to work. It needs a consistent order, so you capture the same kinds of evidence in every visit. This one works on paper, in a document, or in a notes app:
- Header. Date, teacher, grade, subject, number of students, lesson topic, and anything the teacher told you in the pre-conference about what to watch for.
- Running record. One line per moment, each starting with a timestamp. Write what the teacher said or did, then what students said or did. Quote directly whenever you can.
- Snapshots. Every five minutes or so, a quick count: how many students are writing, talking about the task, waiting, or off task.
- Tallies. Simple counts for anything you agreed to watch: questions asked, students called on, student-to-student exchanges.
- Artifacts. What was on the board, the handout, the exit ticket question. A photo works.
- After the lesson. A separate section, filled in later, where you match evidence to rubric components and write your questions for the teacher.
Keeping the last section separate is the habit that makes the rest work. If interpretation has its own place, it's less likely to leak into the running record.
A few lines of a running record in this format look like this:
- 9:05 - T: "Turn to your partner and explain step two." 22 of 24 turned to partners. 2 students at back table did not.
- 9:07 - T circulates to back table, asks student A, "What did you get for step one?" Student A points to paper, says nothing.
- 9:10 - Snapshot: 20 writing, 3 talking about the problem, 1 head down.
Do low-inference notes look different in elementary and high school?
The method is the same. What you count changes. In elementary classrooms, transitions, routines, and small-group rotations take up more of the lesson, so times and student counts carry most of the evidence. In high school classrooms, more of the evidence is in the talk: the exact wording of questions, who answers, and whether students respond to each other or only to the teacher.
High school discussion is also where handwritten notes fall furthest behind. Two students debating a claim can say more in 30 seconds than you can write down, which is the case for recording audio.
Why do low-inference notes lead to fairer ratings?
Ratings built on inference depend on who walked in. Two administrators who both write "questioning was weak" may mean entirely different things, and neither can point to anything the other can check. Ratings built on low-inference evidence can be compared, argued about, and corrected.
Observation ratings are also less reliable than most principals assume. The Measures of Effective Teaching study found that a single observation is a weak signal and that reliability improves substantially when a teacher is observed more than once by more than one observer. Low-inference notes are what make several observations add up, because each visit's evidence can be laid next to the others instead of being reduced to an impression.
They also change the post-observation conversation. A teacher can push back on "your directions were confusing." It's much harder to push back on "five students asked a neighbor what to do within two minutes." The conversation moves from defending the lesson to working out what to do about the evidence, which is the conversation you wanted.
How does audio transcription help with low-inference notes?
A transcript is the most complete low-inference record of classroom talk you can get. It captures every question as it was actually worded, every student answer, and the order they came in, with no inference at all. What it can't capture is anything visual: who had their head down, what was on the board, how the room was arranged.
That points to a practical split. Let the recording capture the talk, and spend your own attention on student counts, movement, and artifacts. Our guide to recording classroom observations covers the consent laws and the practical setup.
Observation Copilot has been a game changer for me. In the past, I would go in, type verbatim the things I'm seeing, hearing, what teachers are doing, what students are doing. Then I would take the time, often 30 to 40 minutes post observation time, to analyze the data and determine the dominant Marzano elements that are within the framework we're using.
How do you turn low-inference notes into feedback?
The notes are the evidence. Feedback adds three things: the rubric component the evidence speaks to, a judgment about the level, and one next step. Our 50 teacher observation feedback examples show that finished form across the major frameworks, and writing better observation notes covers more habits for capturing evidence in real time.
Sorting evidence into rubric components is where most of the write-up time goes. Observation Copilot takes a timestamped running record, plus an audio transcript if you have one, sorts the evidence into your framework's domains and indicators, and drafts strengths, growth areas, and next steps for you to edit. Low-inference notes give it the most to work with, because it can only cite evidence you actually recorded.
Frequently Asked Questions
What are low inference notes?
Low-inference notes are an objective running record of a lesson: direct quotes, observable actions, student counts, and timestamps, written without adjectives or judgments. Observers use them to collect evidence first and decide ratings later, so that feedback rests on facts both the observer and the teacher can see.
What is the difference between observation and inference?
An observation is what a camera and microphone would capture, such as "6 of 24 students had heads down at 9:14." An inference is a conclusion drawn from it, such as "students were disengaged." Low-inference notes record observations during the lesson and save inferences for the analysis afterward.
Can you give me some examples of informal observations?
Informal observations, often called walkthroughs, are short unannounced visits of 5 to 15 minutes. Low-inference notes from one might read: "10:20 - Learning target posted. 14 of 18 students working in pairs on the lab. Teacher conferring with one pair about their hypothesis." Notes are shorter than in a formal observation but just as factual.
Should you share low-inference notes with the teacher?
Many principals do, either before the post-observation conference or at the start of it. Sharing the raw notes lets the teacher see the evidence before any rating, which keeps the conversation focused on what happened. Check your district's evaluation procedures, since some specify what must be shared and when.
How do you take low-inference notes without missing the lesson?
Use a consistent template, write in short timestamped lines, take a student count every few minutes, and develop shorthand for teacher (T) and students (S). Recording audio lets you stop scripting the talk and focus on what you can see.
Paste your low-inference notes and get rubric-aligned feedback in minutes.