From Question to Answer: The PPDAC Cycle
In Reading 1, we established that data without a question is just storage. Now we are going to trace exactly what happens between the moment a question is posed and the moment an answer is communicated.
Researchers and statisticians have been thinking carefully about this process for decades. One of the most widely used frameworks internationally is the PPDAC cycle, developed by Wild and Pfannkuch in 1999 and now embedded in school curricula and data science workflows around the world. PPDAC stands for:
P — Problem — What question are you trying to answer?
P — Plan — How will you go about getting the data you need?
D — Data — Collect, locate, and prepare your data.
A — Analysis — What patterns, relationships, or answers emerge?
C — Conclusion — What did you find, and what does it mean?
We will use PPDAC as our framework throughout this reading. But we want to be honest about two places where the CS and school data context calls for a slightly different emphasis than the original framework assumes. We will flag those departures clearly as they arise.
It is also called a cycle for a reason. In practice, investigators rarely march cleanly from P to C. Exploring the data often reveals that the original question was too vague, which sends you back to refine it. Cleaning the data often surfaces quality problems that change what analyses are possible. Communicating findings often generates new questions. The cycle repeats.
But for learning purposes, it helps to walk through the stages in order first. We are going to do that using a single extended example. Read through the whole example before attempting the "Now You Try" at the end.
Worked Example: Attendance and Assessment Scores
Imagine you are a curriculum coordinator at a mid-sized elementary school. Your principal has a hunch that chronic absenteeism — missing 10 percent or more of the school year — is affecting how students perform on your state's spring reading assessment. She asks you to look into it.
Problem — Pose a Question
Your principal's hunch is a starting point, but it is not yet a precise question. Before you touch any data, you need to sharpen it. Ask yourself: what exactly do you want to compare? Over what time period? For which students?
After some thought, you settle on this question:
Among third, fourth, and fifth graders at our school this past year, did students who were chronically absent (missing 18 or more days) score lower on the spring reading assessment than students who were not chronically absent?
Notice what this question specifies: a grade range, a time period, a definition of "chronically absent" (18 or more days out of a 180-day year = 10%), and a specific outcome measure (spring reading assessment scores). That precision will guide every step that follows.
Plan and Data — Locate and Prepare Your Data
Your question requires two pieces of data for each student in grades 3-5: their total number of absences for the year, and their spring reading assessment score. Fortunately, your school already collects both. Your attendance system has daily records for every student, and the state sends back reading assessment results each June.
You do not need to go out and collect new data — you need to locate and extract existing data. You pull an attendance report and an assessment results file from your school's data systems. After some work, you have a spreadsheet with one row per student containing: student ID, grade level, total absences, and spring reading score.
This is a good moment to notice something important: your data has attributes (the columns: student ID, grade, absences, score) and records (the rows: one per student). We will use this vocabulary extensively in Topics 5B and 5C when we discuss databases.
About the Plan stage in PPDAC: In the original PPDAC framework, Plan is a dedicated stage focused on designing how data will be collected — choosing a sample, designing a survey, deciding on measurement methods. When you are collecting new data, planning that collection carefully is essential, and PPDAC rightly gives it its own stage.
In CS classrooms and school data work, however, the data often already exists. We have combined Plan and Data into a single stage here, but planning still matters — the questions just shift. Before touching any data, a careful investigator asks: Does the available data actually match what my question requires? What are its known limitations? Was it collected consistently? Who was included and who might have been left out? This kind of planning — evaluating existing data before committing to it — is just as important as planning a collection, even if it looks different.
Data — Clean and Organize
Raw data almost never arrives in perfect condition. When you open your spreadsheet, you discover several problems:
- Fourteen students are missing assessment scores. (They were absent on test day and never made it up, or transferred out before testing.)
- One student has an absence count of 847, which is clearly an entry error in a 180-day school year.
- Grade levels are recorded inconsistently — some rows say "3", others say "Grade 3", others say "3rd".
You have to make decisions about each of these. For the missing scores, you decide to exclude those students from the analysis — you cannot compare scores if there are no scores. For the entry error, you flag it and leave it out. For the inconsistent grade labels, you standardize them all to "3", "4", and "5".
Analysis — Analyze and Explore
With clean, organized data, you are ready to look for an answer. You calculate the average spring reading score for students who missed 18 or more days and compare it to the average for students who missed fewer than 18 days. You also break this down by grade level, since your question specifically asked about grades 3-5.
Your results look something like this:
| Grade | Avg. Score (Not Chronically Absent) | Avg. Score (Chronically Absent) | Difference |
|---|---|---|---|
| 3rd | 218 | 197 | −21 points |
| 4th | 231 | 208 | −23 points |
| 5th | 244 | 225 | −19 points |
A consistent pattern emerges: chronically absent students score lower across all three grade levels, with gaps ranging from 19 to 23 points. The answer to your question appears to be yes.
But a careful investigator pauses here. The data shows a correlation — the two things tend to move together — but it does not prove that absence caused the lower scores. Students who are frequently absent may also face other challenges (illness, housing instability, family circumstances) that affect test performance independently. The data cannot tell you which factor is the driver. This is one of the most important lessons in data investigation: correlation is not causation, and a good investigator states findings carefully.
Conclusion — Analyze and Communicate
Our refinement of PPDAC here: The original PPDAC framework combines what we call Analysis and Conclusion into two stages, but treats Conclusion as encompassing both drawing inferences from data and communicating findings to an audience. We separate these deliberately. Reaching a sound conclusion from your analysis and presenting that conclusion clearly to an audience are genuinely different skills — and people often need explicit support with both. A thoughtful analyst who presents findings poorly has not finished the job. By naming Communicate Findings as a distinct step, we make that expectation visible.
You prepare a short report for your principal. It includes a clear statement of the question you investigated, a description of the data you used and the cleaning decisions you made, the table above, and a brief interpretation that notes both the pattern found and the important caveat about correlation vs. causation.
You also include a bar chart visualizing the score differences by grade level, because your principal will share these findings with staff at a faculty meeting and a clear visual will communicate the pattern more quickly than a table alone.
Good communication of data findings means being honest about what the data shows and what it does not show. Overstating findings — claiming the data proves something it only suggests — is one of the most common and most consequential errors in real-world data work.
The Cycle Across Grade Bands
Each stage of the PPDAC cycle appears explicitly in Iowa's Data and Analysis standards, scaled to the appropriate grade level. A kindergartner posing the question "What is the most popular pet in our class?" and counting responses on a tally chart is working through Problem, Data, and Conclusion in a form appropriate for a five-year-old. A high school student using a spreadsheet to clean a dataset, apply a sorting algorithm, and create a multivariate visualization is working through all five stages with considerably more sophistication.
As a teacher, recognizing where in the cycle your students are working helps you target your instruction. If students are struggling to interpret their results, that is an Analysis or Conclusion problem. If they are generating questions that cannot actually be answered with available data, that is a Problem stage issue. The PPDAC cycle gives you a diagnostic framework.
Now You Try
Read the following scenario, then answer the questions below. A suggested response is provided in the answer key.
Scenario: A middle school librarian notices that the library's collection of books published before 2010 rarely gets checked out anymore. She wonders whether updating the collection would increase overall circulation. She has access to a checkout log going back five years that records: student ID, book ID, publication year of the book, and checkout date.
- Write a precise, answerable question the librarian could investigate using her available data.
- What data would she need, and does she already have it?
- Describe one data quality issue she might encounter when she opens her checkout log, and explain how she might handle it.
- What would a useful analysis look like? What comparison would she make?
- What is one thing she should be careful not to claim when she communicates her findings?
Show Answer Key
1. A precise question: "In the past three years, what percentage of total checkouts were books published before 2010, compared to books published in 2010 or later? And has that percentage changed year over year?"
2. Data needed: She needs the publication year of each book and the checkout date for each transaction. Both appear to be in her existing checkout log, so no new data collection is required.
3. A likely data quality issue: Some book records may have missing or incorrect publication years if they were entered manually when books were added to the catalog. She might handle this by checking a subset of records against the actual books, or by flagging and excluding records with obviously wrong years (e.g., publication year listed as 1899 for a clearly modern book).
4. A useful analysis: She could calculate the percentage of total checkouts each year that came from pre-2010 vs. 2010-and-later books, then look at whether that ratio has been trending in one direction over time. She might also compare checkout rates (checkouts per book) rather than raw counts, since the library likely has more older books than newer ones.
5. What not to claim: She should not claim that updating the collection will increase circulation. Her data shows how old books have been circulating — it cannot tell her how students would respond to new books. She can say the data suggests older books are underutilized, which provides a rationale for the change, but the outcome of the change itself is not something the current data can predict with certainty.