Hand Claude a Folder: Analyzing Class Data to Plan Instruction

Every teacher is sitting on more data about their students than they can possibly hold in their head. The diagnostic from three weeks ago. The exit tickets stacking up in a drawer. Attendance patterns that mean something. The margin notes you scrawl after a rough small-group session. Each piece is a signal about where a student actually is — and each one, in isolation, is nearly useless, because the picture only emerges when you put them together. Nobody has the hours to put them together for thirty students, twice a semester.

That synthesis problem is what the Claude for Teachers "analyze class data to plan instruction" workflow exists to solve. You hand Claude a folder — roster, diagnostics, attendance, your own notes — and it builds a picture of where each student stands, so your planning starts from evidence instead of impression. The catch that makes this usable in a real classroom is the part most tools get wrong: you control exactly what's shared, and nothing you share is used to train the model. This is how the workflow runs, and why the privacy design is the load-bearing wall.

The synthesis you never have time for

The point of handing Claude a folder isn't storage; it's synthesis. On its own, a diagnostic tells you a score. Attendance tells you a percentage. Your notes tell you a vibe. Together, they tell you a story — this student's dip in the last unit tracks a run of absences, that student's diagnostic gap is specifically in the prerequisite skill for what you're teaching next week, this group of five all missed the same underlying competency and could be pulled for the same small-group reteach.

Claude does that cross-referencing at a scale you can't do by hand before Monday. And because Claude for Teachers is grounded in the Learning Commons Knowledge Graph — the standards for all 50 states and the learning progressions beneath each one — it doesn't just report that a student missed a question. It can place that miss on the progression: which prerequisite competency is shaky, and therefore what to reteach before moving on. A raw score becomes an instructional next step. That's the difference between data you have and data you can act on.

What goes in the folder

The workflow is deliberately concrete about its inputs. You assemble the evidence you already generate and give Claude the folder. Think of it as everything you'd want a thoughtful co-teacher to read before planning with you:

• The roster — who's in the room, groupings, any relevant flags like IEP goals or English-learner status that shape how you'd support each student.
• Diagnostics and assessments — the beginning-of-unit diagnostic, quiz results, benchmark data. This is the backbone of the "where is each student" picture.
• Attendance — patterns that explain gaps and flag students whose absences are quietly compounding into content holes.
• Your notes — the qualitative layer no data export captures: who's disengaged, who surprised you, what the last small group actually revealed.

Give Claude that folder and ask a real question. Not "analyze this," but "which students are not ready for next week's unit on ratios, and what's the specific gap for each?" The more your prompt looks like the instructional decision you're actually trying to make, the more the analysis reads like a plan instead of a report. Then take the picture Claude builds and feed it straight into the lesson-planning and differentiation workflows — now those plans are aimed at your real class, not an assumed one.

The privacy design is the whole point

Here's where most teachers rightly hesitate, and where the design earns trust. Student data is not something you sprinkle into a general-purpose chatbot and hope for the best. Claude for Teachers is built for this specifically.

You control exactly what's shared, and nothing you share is used to train the model. That single guarantee is what makes it responsible to hand over a folder of student information in the first place.

The protections are structural, not a checkbox. Claude for Teachers has its own teacher terms, written for K-12 privacy — distinct from consumer terms. Student information is covered by a K-12 Data Processing Addendum written to comply with FERPA, the federal law governing student records. And the product is educators-only, consistent with Claude's 18-and-over policy, which means the person operating it is the professional already entrusted with this data, not the student. None of that removes your judgment from the loop — you still decide what belongs in the folder and what stays out. But it means the workflow is designed to hold student data the way a school system is legally required to, rather than the way a free web tool treats whatever you paste in.

The practical upshot: share what you need for the instructional question at hand, and no more. De-identify where you can — student initials often carry enough signal for planning. The control is yours, and exercising it thoughtfully is itself good practice. The broader habits here are worth internalizing; the Claude for Teachers series treats data hygiene as a core teacher skill, not an afterthought.

From picture to instruction

A picture of the class is a means, not an end. The workflow pays off when the analysis drives the next move.

Ask Claude to turn its analysis into groupings for small-group instruction — five students who share the same gap, pulled together for a targeted reteach. Ask it to flag the two or three students whose data suggests they'll be lost by Wednesday without intervention. Ask it which diagnostic results point to a whole-class misconception worth addressing before you advance. Each of these is a planning decision that used to require you to hold the entire class in your working memory at once, and now starts from a synthesis you can read in a minute.

This is also where connectors extend the loop. Eedi generates diagnostic questions designed to reveal student thinking — the misconceptions beneath a wrong answer, in English and Spanish — which is exactly the higher-quality input that makes the class-data analysis sharper. Better diagnostics in, better instructional picture out. The Eedi + Claude workflow is the natural upstream companion to this one, and teachers experimenting outside Claude for Teachers will find adjacent tooling in the marketplace education agent skills collection.

Make it a rhythm, not a rescue

The teachers who get the most from this don't run it once in a panic before a benchmark. They make it periodic. Refresh the folder at each unit boundary, ask the same standing question — where is each student relative to what's next — and let the picture update as the semester moves. Over time, the running analysis becomes an early-warning system: it surfaces the student sliding quietly before the slide becomes a crisis.

And it composes with the rest of the toolkit. The class-data picture feeds planning; planning feeds differentiation; and a scheduled task can keep the loop turning on its own, reviewing each day's exit tickets and adjusting tomorrow — the 4pm exit-ticket review is that same instinct, automated. Data-driven instruction has always been the right idea and the wrong workload. This is the version where the workload finally fits the week. Verification is free through June 30, 2027, and the folder you already have is all you need to start.

Part of the Claude for Teachers series. Related: Eedi + Claude: Diagnostic Questions That Reveal Student Thinking · The 4pm Exit-Ticket Review That Runs Itself. Browse AI tutoring skills or more builder insights.