Data-Driven Leadership That Changes School Culture

It is amazing to me how many people I “meet” in these interviews that just feel like friends. Deborah may be talking data-driven but she’s really talking people-driven and the data just helps. She’s an award-winning NASSP middle school principal, yes, but in the end, she just loves the students and staff she works with and treats them with respect. When you put people first and you use data to surface people issues, then people thrive.

So, in this show, she talks about how she uses data with her assistant principals, her teachers, and with attendance data. She also shares some of her strategies she uses to help both students and teachers care about improvement. And her personal story is also encouraging. Dr. Deborah Dennie is going to take us into data-driven school success that is really just people success. Let’s learn!

Listen to the Show

Watch on YouTube and subscribe for new episodes every week.

Key Takeaways for Educators from Dr. Deborah Dennie

• Her first year, she changed nothing on purpose. “I’m just going to look and I’m going to listen,” she says — because her predecessor had built something over eight or nine years and she did not want to be the principal who walks in and takes it away. That restraint is the leadership move most of us skip.
• Data is not just the gradebook. Every two weeks she sits down with her assistant principals and her resource teacher and reads instructional data, attendance data, discipline data, and climate data together — because the failing grades often trace straight back to empty seats.
• She turns the data into something students want to win. Improve the discipline numbers by the end of October and the whole school gets a classic car show, judged by the students, with community members bringing their cars. The kids know exactly why they earned it.
• When it’s hard, she calls her people. Her word to tired teachers is not a strategy — it’s to find the one person you can call at the end of an awful day. “They have your back.”

Resources Mentioned in This Episode

About Dr. Deborah Dennie

Deborah Dennie, EdD, has served as the principal of Leonardtown Middle School in St. Mary’s County, MD, for 10 years, providing steady, visionary leadership grounded in high expectations and genuine care. During her tenure, she has strengthened instructional practice through data-driven decision-making, elevated student accountability, and cultivated a culture of continuous professional growth among educators.

Widely recognized as a mentor and advocate, she empowers staff to pursue leadership opportunities and expand their professional capacity, contributing to improved teaching and learning outcomes schoolwide while prioritizing the emotional and physical well-being of students and staff. She ensures instructional time is purposeful, distractions are minimized, and collaborative planning is both funded and prioritized. This shared focus has resulted in rising proficiency, greater equity in classrooms, and stronger student readiness for high school and beyond. Dr. Dennie is a NASSP 2026 Middle Level National Principal of the Year finalist.

Connect with Dr. Dennie: LinkedIn

Data Driven Resources for Teachers and Administrators

AI Prompts To Get You Started

I wanted to give you some samples that work to help you as you get started. As always, make sure your AI tools are approved by your schools.

What she did: Dr. Dennie reviews her building’s data every two weeks — and not just grades. Attendance, discipline, climate — all of it. AI can help you start a similar habit, whether you lead a classroom or a building. Below are two prompts you can copy and use this week, plus what you need to know before any student data goes near an AI tool. Also, administrators can build data dashboards. AJ Juliani talked about this recently on Cool Cat Teacher Talk.

STEP 1: Protect Student Privacy

Remove ALL of this before any data goes into an AI tool: student names, student/state ID numbers, birthdates, addresses and contact info, IEP/504 or services status, discipline details tied to a student, and any demographic detail that could identify a student in a small group. Rename students “Student 1, Student 2…” and keep your key on paper — never in the chat. If a subgroup has fewer than 10 students, report it as a group total, not individuals.

The teacher versions are safer — but anonymize anyway. ChatGPT for Teachers (free for verified U.S. K–12 educators through June 2027) does not use what you share to train models by default and is built with FERPA requirements in mind. Claude for Teachers runs under K–12 terms with a FERPA-aligned data processing agreement, and Anthropic states it will not train on your inputs or outputs. I’ve walked through exactly how to sign up for both in Free AI for Teachers: How to Get Claude and ChatGPT Free. Those are real protections a personal free account does not have — but privacy experts still urge caution. You are the one responsible for student records under FERPA, so de-identify first, every time.

And check your district: use a tool your district has approved, and follow your district’s AI and student-data policies.

STEP 2: Get Question Level Data from Your Tools

One of the most useful things you can do after you give any kind of assessment is to look at the question data. A class average can’t tell you what to reteach. You need per-question (or per-standard) data. Most assessment tools will hand it to you:

• Wayground (formerly Quizizz): open the activity’s report and download it — you get a spreadsheet with question-by-question accuracy for every student.
• Kahoot!: download the spreadsheet report from any game — it includes per-question performance.
• ExamView (the Assessment Suite many textbook programs use): the Test Manager’s reports include an item analysis showing how the class performed on each question, which you can export.
• Google Forms quizzes, Canvas, Schoology, and most LMS gradebooks export results as a spreadsheet or CSV too — and if you use Blooket, Gimkit, Formative, or Edulastic, look in the reports section: most will let you download a question-level spreadsheet as well. (Often called a CSV – see below.)

Then attach your files instead of pasting data into the chat — it’s cleaner, keeps the formatting intact, and works with the file you already have. De-identify your results export inside the file first: open it, delete the name and ID columns, replace them with Student 1, Student 2…, and save with a new name. Be careful to save or you may accidentally upload the data with the names in it.

What’s a CSV? Just a plain-text spreadsheet: one row per student, columns separated by commas. Every tool above can produce one (or an Excel file, which works the same way). A csv export might look like this. See the commas…

Student,Q1 (6.NS.1),Q2 (6.NS.1),Q3 (6.EE.2),Score
Student 1,correct,incorrect,correct,67%
Student 2,incorrect,incorrect,correct,33%

STEP 3: Build Your Prompts

How these prompts are built

A quick word about the prompts below, because the how matters as much as the copy-paste:

• Name the job, not a person. Each prompt opens with “You are a [role]” plus a job description. That’s not pretending the AI is human. Instead, it is a job description that focuses the output on the work you need done. If you have a job description of what you want AI to do, state the job. It unlocks capabilities you might not know about.
• Skip the small talk. No “please,” no “thank you” — an AI tool isn’t a person and doesn’t need politeness to perform. State the task, the data, and the format you want. And when you don’t get what you want, tell it clearly and what is wrong.
• Number what you want back. Numbered requests help produce better results.

SAMPLE PROMPT #1: Prompt for teachers — after any quiz, test, or unit assessment

Replace the words in the brackets that look like this [what you type there] with the word that is needed there. When you have a prompt you like, put it in your prompt library.

Attach: (1) your de-identified question-level results export  (2) the test you gave (with key)

You are an instructional data analyst for a [grade level] [subject] teacher. Your job is to find patterns in de-identified assessment data and recommend next instructional steps.
Attached are two files: (1) the question-level results export — de-identified, no student names or IDs, students listed as Student 1, Student 2, etc.; (2) the assessment itself, so you can see what each question actually asked.
1. Identify the weakest standards, concepts, and question types, and note which students show similar gaps.
2. Sort students into three groups — reteach, practice, extend — and suggest one small-group activity for each group.
3. State what this data cannot tell us and what to watch for in class.
4. Write three questions to bring to my collaborative planning meeting.
5. Write an improved, reusable version of this prompt for my next assessment, based on what you needed most from these files.

SAMPLE PROMPT #2: Prompt for principals and admin teams — modeled on Dr. Dennie’s two-week data review

Attach: your de-identified building reports. Data might be failure rates by course section vs. building average, attendance percentages, top discipline categories (anonymized – see above).

You are a school-improvement data analyst for a [school level] principal. Your job is to connect academic, attendance, and discipline data into a picture of what is happening in the building and what to investigate next.
Attached are de-identified building reports — no student names or IDs. The files are: [describe each file you attach — e.g., failure rates by course section vs. building average, attendance percentages, top discipline categories]
1. Flag the sections above the building-average failure rate and list contributing factors worth investigating — instructional and non-instructional.
2. Cross-reference: where do attendance or discipline patterns line up with academic gaps?
3. Write three pointed but supportive questions to send the relevant teachers, in the spirit of: What are we doing for the kids who are failing? Are we reteaching? Are we looking at the standards?
4. Recommend one incentive or culture move aimed at the biggest non-academic driver in this data.
5. Write an improved, reusable version of this prompt for my next data review, based on what you needed most from these files.

Pro Tips For Better Results

• Show it what “good” looks like. Attach two more files — a de-identified sample of your data and a report or result you liked. Name them clearly: sample-data and sample-report. Then add one line to the prompt: “Match the format of sample-report.” The AI builds on what already works for you, and the output gets noticeably better.
• Name every file you attach something you can reference in the prompt — “unit-4-results,” “unit-4-test” — instead of “export(3).xlsx.”

STEP 4: Make Your Data Task Repeatable

Dr. Dennie’s data review happens every two weeks — the power is in the routine, not the one-off. If this becomes your routine too, stop rebuilding the prompt from scratch: request #5 in each prompt above has the AI write your next, better prompt for you.

Save it in your own prompt library. And if you work in one tool regularly, build the routine right in — in Claude you can package it as a Skill, and in ChatGPT you can turn it into a custom GPT. Either way, next time it’s one click instead of twenty minutes.

For Spreadsheet Analysis: Excel and Google Sheet Resources

If your data lives in a spreadsheet, you may not need to attach it anywhere:

• Copilot in Excel — when you’re signed in with your school (work/school) account. If your school has an account (double check with them), Microsoft’s enterprise data protection applies: your prompts and data stay in your school’s tenant and are not used to train the underlying models. If you’re signed in with a personal account, those protections don’t apply — check which account you’re in before you touch student data.
• Gemini in Google Sheets — yes, Gemini does spreadsheets, right inside Sheets. With a Google Workspace for Education account, enterprise-grade data protection applies and your school’s content is not used to train the models. Same rule as Copilot: know which account you’re signed into.
• Claude for Excel — Anthropic’s add-in reads and works with your spreadsheet directly (available on paid Claude plans). Data handling follows your Claude account’s settings, so know what plan and settings you’re on before using it with school data.
• Whichever you use, the anonymize-first rule above still applies. The tool changes; your FERPA responsibility doesn’t.

One last thing: the AI’s answer is the starting point for your team’s conversation — not the decision. As Dr. Dennie reminds us, the numbers only matter because of the kids behind them.

MORE Free Resources on Data-Driven Leadership

Ready to dig into the kind of all-encompassing data work Dr. Dennie describes? Here are some more resources to get you started:

From my archive:

Data tools and research:

• Attendance Works — Data Tools (schools & districts; free self-calculators): Dr. Dennie’s 94% attendance push is exactly what these support — calculate chronic absence and spot the students who need you. This organization gives free quantitative analysis tools relating to attendance. If you’re focusing on attendance you might want to start here. attendanceworks.org/resources/data-tools
• Attendance Works — Principals Leading Attendance Toolkit (school leaders): free guidance for using chronic-absence data to decide which students need additional supports. View the toolkit
• IES Practice Guide: Using Student Achievement Data to Support Instructional Decision Making (K–12 teachers & leaders; free PDF): the What Works Clearinghouse’s research-based recommendations for making data an ongoing cycle of improvement — the research behind what Dr. Dennie practices. Download from IES

More free AI prompt libraries for educators:

Meet all of the NASSP 2026 Finalists.

• NASSP 2026 National Principal of the Year Finalists — meet Dr. Dennie and her fellow finalists; great examples to share with an aspiring leader you’re mentoring. See the finalists · Connect with Dr. Dennie on LinkedIn.

Other Shows for School Leaders

Listen and Subscribe

If this episode gave you something to try tomorrow, share it with a school leader who needs it.

Episode Transcript

This transcript was generated using AI and has been reviewed by humans for accuracy. Minor errors or artifacts may remain but I worked my best to find any issues with the transcript as I reviewed the show. – Vicki