# Run Chart Rules: How to Tell If a SIP Project Worked

How to read a run chart for a School Improvement Plan project: the four rules for a real change, weekly attendance and reading data, and Steps 9 to 11.

This is the Markdown copy of [Run Chart Rules: How to Tell If a SIP Project Worked](https://nash.com.ph/blog/run-chart-rules-sip-projects), an article on the NASH Reviewer Blog by Mark Anthony Llego, published Nov 20, 2026. Topics: School leadership, Aspiring school heads. NASH Reviewer is an independent practice platform not affiliated with DepEd, BHROD, or NEAP.

A run chart tells a school whether a School Improvement Plan (SIP) project changed anything. It is a line graph of one measure, such as a section's weekly attendance or the number of learners who read the week's passage at grade level, plotted in time order around its median. Four rules, set out by Perla, Provost and Murray in 2011, separate a real change from the ordinary ups and downs of the weeks: a shift, a trend, too few or too many runs, and an astronomical point.

The SIP guidelines, [DepEd Order (DO) No. 44, s. 2015](https://www.deped.gov.ph/2015/09/29/do-44-s-2015-guidelines-on-the-enhanced-school-improvement-planning-sip-process-and-the-school-report-card-src/), ask a project team to test a solution in a small group, roll it out, and check progress. For the test they give one instruction: "Compare your data before and after testing the solution" (SIP Guidebook, Step 9). A pair of averages cannot say whether the change caused the difference, whether the measure was already moving, or whether the gain lasted. Weekly data in a run chart can.

The rules below are those of the published method, applied to the steps of DO 44 as they stand on November 20, 2026.

## Key findings

- **DO 44 sets no rule for deciding whether a test worked.** Step 9 says to "Compare your data before and after testing the solution", and Step 11 has the planning team monitor "at the middle and end of the implementation period". As of November 20, 2026, no DepEd order or memorandum describes run charts or rules for reading data over time.
- **A shift is six or more points in a row on one side of the median.** "Values that fall on the median do not add to nor break a shift" (Perla, Provost and Murray, 2011). Six points above the median by chance alone are unlikely enough to read as a signal.
- **A trend is five or more points in a row all going up or all going down.** A simulation study found the trend rule "virtually useless" for detecting a gradual drift (Anhøj and Olesen, 2014, p. 5). Read a trend alone as a reason to keep plotting, with a shift as the stronger signal.
- **Too few or too many runs is a signal too.** A run is a stretch of points on one side of the median; the number to expect depends on how many points are not on the median, and a published table gives the limits (Perla, Provost and Murray, 2011, p. 48 and Table 1).
- **The rules need enough weeks.** "The shift and run rules require more than 10 points before they are applicable" (Perla, Provost and Murray, 2011, p. 48). For a SIP pilot, that means plotting ten or more weeks before the change and enough weeks after it.
- **The data already exist.** School Form 2 gives each section's attendance every day. In Key Stages 2 and 3 (Grades 4 to 10), the Academic Recovery and Accessible Learning (ARAL) Program's reading tutors give weekly formative assessments ([DepEd Memorandum No. 064, s. 2025](https://www.deped.gov.ph/2025/08/01/august-1-2025-dm-064-s-2025-implementing-guidelines-for-the-academic-recovery-and-accessible-learning-program-in-reading-for-key-stages-1-to-3/), Encl. 3, item 34). DepEd's monitoring framework allows monitoring "on a weekly, monthly, or quarterly basis depending on the M&E Plan" ([DepEd Order No. 029, s. 2022](https://www.deped.gov.ph/2022/06/22/june-20-2022-do-029-s-2022-adoption-of-the-basic-education-monitoring-and-evaluation-framework/), Sec. VI).
- **In one Philippine improvement project, tests were often measured only once.** In a 2021-2022 project with 19 school leaders in Lanao del Sur, "there was often just one measurement point during the 'study' phase" (Wolfenden and others, 2023, p. 34).

## What a run chart is

A run chart is a graph of one measure over time: the weeks along the bottom, the measure up the side, one point for each week, joined by a line, with a horizontal line at the median. In Perla, Provost and Murray's words, "A run chart is a graphical display of data plotted in some type of order", and "Usually, the median is calculated and used as the chart's centreline" (2011, p. 47).

Two words carry the rules. A **run** is one or more points in a row on the same side of the median; a point on the median belongs to no run and is skipped. The runs rule counts only the points that are not on the median; Anhøj and Olesen call them useful observations (2014, p. 3). With twenty weeks and two of them exactly on the median, a chart has eighteen.

The line is the median, not the average. The median "provides the point at which half the observations are expected to be above and below the centerline", and it "is not influenced by extreme values in the data" (Perla, Provost and Murray, 2011, p. 47). A week when a typhoon kept half a section at home drags the average down; the median stays where most weeks are.

A run chart is not a test of statistical significance, and it needs no software beyond a spreadsheet. It answers a practical question: is this week's pattern still the old one, or has something changed?

## Why comparing before and after is not enough

A before-and-after comparison reduces weeks of data to two numbers. Three different stories can give the same two numbers. In the figure, three invented sections tried the same change after week 10; each averaged 88.0 percent attendance before and 91.5 percent after. Only one of them shows a change that worked.

**Same before, same after, three different stories.** Three invented Grade 6 sections tried the same change after week 10. Each averaged 88.0% attendance in weeks 1 to 10 and 91.5% in weeks 11 to 20, so a before-and-after comparison cannot tell them apart. Their run charts can (rules of Perla, Provost and Murray, 2011).

- **Section Mabini.** Attendance held steady, then rose and stayed up after the change. Every week after the change is above the median of the ten weeks before it: a shift of ten points, well past the six the rule needs. The change likely worked. Weekly attendance (% present): 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.6, 89.4, 86.6, 91.1, 92.2, 90.6, 91.7, 92.8, 90.3, 91.9, 92.5, 90.8, 91.1.
- **Section Rizal.** Attendance was already climbing before the change, and kept on. The rise began before the change: the first ten weeks climb week after week, a trend far past the five points the rule needs, and the points sit above the median from week 6 on. The higher average after the change cannot be credited to it. Weekly attendance (% present): 84.4, 85.3, 86.4, 87.2, 88.1, 88.6, 89.4, 89.7, 90.3, 90.6, 89.7, 90.8, 91.1, 91.4, 91.9, 91.7, 92.2, 91.4, 92.5, 92.3.
- **Section Bonifacio.** Attendance jumped for the first weeks of the change, then slid back. Four high weeks, then six back at the old level: no shift of six above the median, and a trend of five points going down from week 12 as the gain fades. The average after the change hides a gain that did not last. Weekly attendance (% present): 87.6, 88.4, 89.1, 87.2, 86.8, 88.7, 89.3, 87.9, 88.5, 86.5, 96.4, 97.5, 96.1, 95.8, 88.6, 87.5, 89.2, 87.8, 88.6, 87.5.

The second section's attendance was rising before the change began, so the change cannot take the credit. The third gained for four weeks and slid back; its average after the change hides the slide. A team that compared two averages would report all three as successes.

The same trap affects a pretest and a posttest. Learners grow over a school year with ordinary teaching, and a group chosen for its low scores tends to score higher on a second test by chance alone. The [guide to the pretest-posttest design](https://nash.com.ph/blog/pretest-posttest-design-regression-to-the-mean) explains both traps; measuring every week, before and after the change, is one of its remedies.

## The four run chart rules

Three of the rules are counts, and the fourth is a judgment. The shift and runs rules are set so that the pattern each looks for has less than a 5 percent chance of appearing when nothing has changed: they rest on "an α error of p\<0.05" (Perla, Provost and Murray, 2011, p. 47), or in the data guide's words, "a probability of less than 5% chance (or p = 0.05) of occurring just by chance when there is no real change" (Provost and Murray, 2011, p. 77). The trend rule rests on experience, and the fourth rule is not a count. Over a whole chart, the chance that some rule signals by chance is higher, as the simulations below show.

| Rule | What to look for | How to count |
|---|---|---|
| 1. Shift | "Six or more consecutive points *either* all above or all below the median" | "Values that fall on the median do not add to nor break a shift. Skip all values that fall on the median and continue counting." |
| 2. Trend | "Five or more consecutive points all going up or all going down" | "If the value of two or more consecutive points is the same, only count the first point and ignore the repeating values; like values do not make or break a trend." |
| 3. Runs | "too few or too many runs, or crossings of the median line" | "count the number of times the line connecting the data points crosses the median and add one", then compare with the table below |
| 4. Astronomical point | "one that is obviously, even blatantly, different from the rest of the points; all studying the chart would agree the point is unusual" | No count: "Rule 4 is subjective". Every chart has a highest and a lowest week; they are not astronomical for that reason. |

*Rules as printed by Perla, Provost and Murray (2011), pp. 47-48.*

The six of the shift rule comes from coin tosses. If nothing has changed, each week has an even chance of falling above or below the median, and six in a row on one side happens with a probability of about 0.03 (Provost and Murray, 2011, p. 78). The five of the trend rule is a judgment: Provost and Murray wrote: "We have based our decision rule of 5 for a trend on our experience" (p. 79).

The rules have critics. A simulation study found the trend rule "virtually useless for detection of linear drift over time, the purpose it was intended for", and recommended "that the use of the trend rule should be discontinued" (Anhøj and Olesen, 2014, pp. 5 and 12). Applied together, the rules signal more often than any one alone. In a later simulation of 24-week charts with no real change, the Perla rules signaled in 466 of 1,000 charts, against 73 for the stricter rules Anhøj proposed (Anhøj, 2015, Table 2). Those stricter rules call a shift only when the longest run passes log2(n) + 3, rounded, where n is the number of useful observations. They also look for too few crossings of the median. With 20 points, that is a run of more than 7, or fewer than 6 crossings (Anhøj and Olesen, 2014, p. 3).

For a school, two habits follow. Decide before the data come in which rules the team will use: "the choice of run chart rules should be decided before data collection begins" (Anhøj, 2015, p. 8). And treat a trend alone as a reason to keep plotting, not as proof; a shift after the change is the stronger signal.

### How many runs to expect

If only chance moves the measure, the line should cross the median at a fairly regular rate. Too few runs means the weeks have bunched on one side, as they do when the level has moved. Too many means the weeks alternate up and down more than chance allows, which most often "represents data in need of stratification rather than data that truly reflect improvement" (Provost and Murray, 2011, p. 81): two groups mixed in one measure, such as two sections reported on alternate weeks.

Count the points that are not on the median, then read the limits:

| Points not on the median | Too few: fewer than | Too many: more than |
|---|---|---|
| 10 | 3 | 9 |
| 12 | 3 | 11 |
| 14 | 4 | 12 |
| 16 | 5 | 13 |
| 18 | 6 | 14 |
| 20 | 6 | 16 |
| 22 | 7 | 17 |
| 24 | 8 | 18 |
| 26 | 9 | 19 |
| 28 | 10 | 20 |
| 30 | 11 | 21 |

*Selected rows of Perla, Provost and Murray (2011), Table 1, p. 49, which runs from 10 to 60 points and is "based on about a 5% risk of falling [sic] the run test for random patterns of data", adapted from Swed and Eisenhart (1943).*

"Given 20 data points that do not fall on the median, as few as 6 or as many as 16 runs are expected" (Provost and Murray, 2011, p. 81).

### Which median to draw

To test a change, take the median from the weeks before it, then extend it, unchanged, across the weeks after: "If the baseline data come from a process exhibiting no signals (shift, trend, runs, astronomical data point), extend or 'freeze' this initial median into the future. By using the baseline data and extending the median into the future, new data are not allowed to influence the initial median" (Perla, Provost and Murray, 2011, p. 48).

A median taken from all the weeks, before and after, sits between the old level and the new one and can hide the change it is meant to show. Provost and Murray put it plainly about one of their examples: "If they had allowed all the data to be used in the determination of the median rather than solely the baseline data, the shift would not have been detected" (2011, p. 88).

### How many weeks you need

The sources differ a little. Perla, Provost and Murray say the shift and run rules "require more than 10 points before they are applicable" (2011, p. 48). Provost and Murray's data guide says the median "should be created using 10 or more data points" (2011, p. 87). Anhøj and Olesen advise adding the median "After at least 12, preferably 20 or more, data points" (2014, p. 12).

For a SIP project, aim for ten to twelve weeks before the change. They need not be new weeks: attendance for the weeks already past is in School Form 2, so a team can build its baseline from records on the day it plans the test. After the change, a shift needs at least six weeks to appear.

## Plot your own weeks

The lab below draws a run chart from twenty weeks you type, drag or load. Move the week your change began: the median is taken from the weeks before it, and the rules mark every shift and trend.

**Plot twenty weeks and read the signals.** On the page, the reader enters twenty weeks of a measure and the week a change started; the chart is drawn around the median of the weeks before the change, and the rules of Perla, Provost and Murray (2011) are applied: a shift is 6 or more points in a row on one side of the median, a trend 5 or more in a row all going up or all going down, and too few or too many runs is a signal. The invented series, each with the change after week 10:

- **Attendance, a real shift** (learners present, percent of the week's attendance): 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.6, 89.4, 86.6, 91.1, 92.2, 90.6, 91.7, 92.8, 90.3, 91.9, 92.5, 90.8, 91.1. Shift: weeks 11 to 20, 10 points above the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 9 runs in 20 useful points; between 6 and 16 is ordinary. A signal after the change, in the direction you want, against a baseline median of 87.95%. The change likely made a difference: adopt it, or test it more widely.
- **Attendance, only noise** (learners present, percent of the week's attendance): 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.6, 89.4, 86.6, 88.4, 89.1, 87.3, 86.8, 88.2, 87.5, 89, 88.7, 87.2, 88.6. Shift: none. No 6 points in a row on one side of the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 15 runs in 20 useful points; between 6 and 16 is ordinary. No signal: the weeks vary around the baseline median of 87.95% as they did before. Keep plotting, or change the change. A single week far from the rest is the fourth rule's question, and only you can judge it.
- **Attendance, a slow climb** (learners present, percent of the week's attendance): 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.6, 89.4, 86.6, 88.1, 88.9, 89.4, 90.3, 91.1, 89.7, 90.6, 91.4, 90.8, 91.7. Shift: weeks 11 to 20, 10 points above the median. Trend: weeks 10 to 15, 6 points going up. Runs: 9 runs in 20 useful points; between 6 and 16 is ordinary. A signal after the change, in the direction you want, against a baseline median of 87.95%. The change likely made a difference: adopt it, or test it more widely.
- **Attendance, one unusual week** (learners present, percent of the week's attendance): 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.6, 89.4, 86.6, 88.4, 89.1, 87.3, 97.8, 88.2, 87.5, 89, 88.7, 87.2, 88.6. Shift: none. No 6 points in a row on one side of the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 15 runs in 20 useful points; between 6 and 16 is ordinary. No signal: the weeks vary around the baseline median of 87.95% as they did before. Keep plotting, or change the change. A single week far from the rest is the fourth rule's question, and only you can judge it.
- **Reading, a real shift** (learners who read the week's passage at grade level, out of 30): 9, 12, 10, 8, 13, 11, 9, 12, 11, 8, 13, 14, 12, 15, 14, 16, 15, 17, 16, 18. Shift: weeks 11 to 20, 10 points above the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 8 runs in 20 useful points; between 6 and 16 is ordinary. A signal after the change, in the direction you want, against a baseline median of 10.5 of 30. The change likely made a difference: adopt it, or test it more widely.
- **Reading, only noise** (learners who read the week's passage at grade level, out of 30): 9, 12, 10, 8, 13, 11, 9, 12, 11, 8, 11, 9, 10, 12, 13, 9, 8, 11, 12, 10. Shift: none. No 6 points in a row on one side of the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 13 runs in 20 useful points; between 6 and 16 is ordinary. No signal: the weeks vary around the baseline median of 10.5 of 30 as they did before. Keep plotting, or change the change. A single week far from the rest is the fourth rule's question, and only you can judge it.
- **Reading, a slow climb** (learners who read the week's passage at grade level, out of 30): 9, 12, 10, 8, 13, 11, 9, 12, 11, 8, 10, 11, 12, 13, 14, 13, 15, 16, 15, 17. Shift: weeks 12 to 20, 9 points above the median. Trend: weeks 10 to 15, 6 points going up. Runs: 8 runs in 20 useful points; between 6 and 16 is ordinary. A signal after the change, in the direction you want, against a baseline median of 10.5 of 30. The change likely made a difference: adopt it, or test it more widely.
- **Reading, one unusual week** (learners who read the week's passage at grade level, out of 30): 9, 12, 10, 8, 13, 11, 9, 12, 11, 8, 11, 9, 10, 23, 13, 9, 8, 11, 12, 10. Shift: none. No 6 points in a row on one side of the median. Trend: none. No 5 points in a row all going up or all going down. Runs: 13 runs in 20 useful points; between 6 and 16 is ordinary. No signal: the weeks vary around the baseline median of 10.5 of 30 as they did before. Keep plotting, or change the change. A single week far from the rest is the fourth rule's question, and only you can judge it.

The four series show the cases a project team meets most often. In the first, the weeks after the change sit above the old median and a shift appears. In the second, the weeks wander around the median as before: the change has made no difference yet. In the third, the climb after the change is gradual; a trend appears, and a shift follows. In the fourth, one week stands far from the rest, which the counted rules do not flag and the fourth rule leaves to your judgment.

Now load the first series and set the change to start after week 12. The weeks are the same, but the rise now begins in week 11, before the change, so the lab no longer gives the change the credit: something else moved the measure first, and the team has to find out what.

## Choosing a weekly measure for a SIP project

A run chart needs a measure that is taken the same way every week. The best ones are already recorded.

| Measure | Where it comes from | How often | Rule |
|---|---|---|---|
| A section's attendance rate | School Form 2, the Daily Attendance Report of Learners | Daily, summed each month | [DepEd Order No. 4, s. 2014](https://www.deped.gov.ph/2014/01/30/do-4-s-2014-adoption-of-the-modified-school-forms-sfs-for-public-elementary-and-secondary-schools-eeffective-end-of-school-year-2013-2014/); [DepEd Memorandum No. 014, s. 2021](https://www.deped.gov.ph/2021/03/26/march-26-2021-dm-014-s-2021-interim-guidelines-on-the-preparation-submission-and-checking-of-school-forms-for-the-sy-2020-2021/), Annex 1 |
| Learners absent five days in a row | School Form 2 | Monthly count, "which can be used as basis for home visitation" | DO 4, s. 2014, SF2 data element 18 |
| Reading progress in the ARAL Program, Grades 4 to 10 | Tutors' weekly formative assessments and Individualized Reading Intervention Plans | Weekly | DM 064, s. 2025, Encl. 3, items 32 and 34 |
| Reading profiles from the Comprehensive Rapid Literacy Assessment (CRLA) and the Philippine Informal Reading Inventory (Phil-IRI) | DepEd's reading assessments | In the beginning-, middle- and end-of-year testing windows | [DepEd Order No. 009, s. 2026](https://www.deped.gov.ph/2026/04/16/april-16-2026-do-009-s-2026-guidelines-on-the-implementation-of-the-three-term-school-calendar-in-basic-education/), Annex B; [DepEd Order No. 018, s. 2025](https://www.deped.gov.ph/2025/06/13/june-13-2025-do-018-s-2025-implementing-guidelines-of-the-academic-recovery-and-accessible-learning-program/), Table 1 |

The CRLA and the Phil-IRI give at most three points a school year, too few for a run chart within the year. They remain the measures of the result. A weekly chart follows something that moves faster: attendance, the share of a group that reads the week's passage, or how often the change itself was carried out, such as the number of first-day calls home made each week. In Grades 4 to 10, the [ARAL Program](https://nash.com.ph/blog/deped-aral-program-guide) already has reading tutors check their learners every week.

Bryk and his co-authors call this "practical measurement": "measures to inform improvement must be embedded in the regular work of teaching and learning", and "data need to be collected frequently" (2015, p. 100). Their Carnegie Foundation colleagues set such measures apart from "the measures commonly used by schools for accountability or by researchers for theory development" (Yeager and others, 2013, p. 2). A measure for improvement, they write, must "embed within the constraints of everyday school practice"; a survey given routinely in class, for example, may get "no more than 3 minutes". It must also be "sensitive to changes in the short term and quickly accessible" (Yeager and others, 2013, pp. 12 and 13). Attendance already meets both tests. A reading check meets them when it is short, such as one passage a learner reads aloud to a tutor.

### Computing a weekly attendance rate

School Form 2 prints a monthly formula in its guidelines: "Percentage of Attendance for the month = (Average daily attendance / Registered Learners as of end of the month) x 100", where "Average Daily Attendance = Total Daily Attendance / Number of School Days in reporting month" (SF2, as printed in DM 014, s. 2021, Annex 1). A week works the same way: add the learners present on each class day of the week, divide by the number of class days, then by the learners enrolled, and multiply by 100.

Count only class days. A week with a holiday has four class days, not five, and a suspended day is not a class day. If the divisor is always five, a holiday week looks like a week of absences.

### The school calendar breaks the weeks

Under the three-term calendar for School Year (SY) 2026-2027, each term ends with "a two-week End-of-Term Block" (DO 009, s. 2026, item 15), and the year has holidays, assessment weeks and a year-end break (Annex B). Leave out the weeks without regular classes, and keep the week numbers of the weeks plotted, so the gap shows on the chart. Plot weeks in which the measure was taken the same way; do not join a reading check done in the End-of-Term Block to the weekly checks of the Instructional Block.

## From the test to the progress report: Steps 9 to 11

The SIP Guidebook's own example of a test is a Grade 6 attendance project: "if the project team made improvements to the Grade 6 attendance checking process, the concerned stakeholders or process owners would be the Grade 6 class advisers" (Step 9). The figure follows such a project through the three steps on two run charts.

**From a test in one section to the progress report.** One invented school's Grade 6 attendance project, followed through the SIP Guidebook's Steps 9 to 11 (DepEd Order No. 44, s. 2015) on two run charts.

- **Before Step 9: Plot the weeks before any change.** Ten weeks of Section B's attendance from School Form 2, before anything changes. The points wander between 86 and 90 percent around a median of 88.0: no shift, no trend, and an ordinary number of runs. This is how the process behaves now. Decision: Freeze the median of these ten weeks and extend it across the chart as the line the change must beat.
- **Step 9, Test the solutions: Test the change in one section.** From week 11, Section B's adviser calls home on the first day a learner is absent. Every week since sits above the frozen median; the sixth in a row, week 16, makes a shift. The grade's chart, where three sections have not changed, shows nothing yet. Decision: Adopt: the change worked where it was tried. Hand it to the other Grade 6 advisers, the process owners, for week 19.
- **Step 10, Roll out the solutions: Roll it out and keep plotting.** All four advisers make the calls from week 19. The grade's chart shifts above its own frozen median by week 24. Week 25 drops to 63 percent: a typhoon struck that Monday, and when classes resumed many learners stayed home to help clean up. It is an astronomical point with a known cause, marked on the chart, not a sign that the change failed. Decision: Keep the calls in every section, note the storm week on the chart, and keep plotting.
- **Step 11, Check progress of AIP: Write the progress report from the chart.** The Project Monitoring Report Form asks for the project's objectives and targets, its accomplishments or status to date, its issues and its recommendations. The run chart supplies the status to date and the dates of any shift. Decision: Status to date: Grade 6 attendance above the baseline median of 88.5 percent in every week since the roll-out except the typhoon week. Issue: the typhoon week (week 25), when attendance fell to 63 percent. Recommendation: keep the calls; set the target from the new level.

- Grade 6, Section B (the pilot), weekly attendance (% present), weeks 1 to 28: 88.6, 89.2, 87.4, 88.3, 87.1, 86.9, 88.9, 87.7, 89.4, 86.6, 90.0, 91.4, 90.8, 92.2, 91.7, 92.5, 91.1, 92.8, 91.9, 92.2, 92.5, 91.7, 93.1, 92.2, 61.4, 92.5, 91.9, 93.3. Baseline median from weeks 1 to 10.
- Grade 6, all four sections, weekly attendance (% present), weeks 1 to 28: 88.4, 87.9, 88.8, 87.6, 88.6, 88.1, 89.0, 87.4, 88.5, 88.2, 88.9, 88.7, 88.3, 89.1, 88.6, 89.2, 88.5, 88.2, 90.1, 90.8, 91.2, 90.6, 91.5, 91.0, 63.2, 91.4, 91.8, 91.3. Baseline median from weeks 1 to 18.

**Step 9, test the solutions.** The guidebook asks for a test "on a small population to reduce the risk of failure", such as "a section out of an entire grade level". Plot that section's measure for ten or more weeks before the test, freeze the median, and keep plotting. The test has worked when a shift appears after the change, in the direction the project wants. The [guide to root cause analysis](https://nash.com.ph/blog/root-cause-analysis-school-improvement-plan) covers what to test and how to choose the section. The improvement method behind run charts asks three questions, the second being "How will we know that a change is an improvement?", and "stresses learning by testing changes on a small scale rather than by studying the problem before any changes are attempted" (Langley and others, 1996, p. 4). Bryk and his co-authors tell the story of a hospital team that started a change with one doctor and one unit. When the change went hospital-wide, the team reviewed new records every week and "tracked the data over time in what is known as a run chart" (2015, p. 116). A section is a school's one unit.

**Step 10, roll out the solutions.** The solution is rolled out "to the rest of the school, not by the Project Team, but by the concerned stakeholders (process owners)" (Step 10). Plot each group's measure as it takes the change up, starting with the whole grade, with its own baseline median from the weeks before its roll-out. The roll-out has taken hold when the grade's chart shifts; a test section that improves while the grade does not means the other sections are not yet doing the change.

**Step 11, check progress of the Annual Implementation Plan (AIP).** The School Planning Team (SPT) monitors "at the middle and end of the implementation period. For longer projects, the SPT will have to monitor more than twice" (Step 11). The Project Team reports on the guidebook's suggested Project Monitoring Report Form: the project's objectives and targets, its "Accomplishments/Status to Date", its issues, and its recommendations, which the two teams discuss together. The [sample SIP](https://nash.com.ph/blog/school-improvement-plan-sample-sip-aip-template) shows the form filled in. The run chart gives the status in a sentence a parent can follow: attendance above the old median every week since the roll-out, except the week a typhoon struck.

[DepEd's monitoring framework](https://nash.com.ph/blog/deped-bemef-monitoring-evaluation-framework) asks every school head to "Develop M&E plan for their respective School improvement plan" and to hold a [quarterly program implementation review](https://nash.com.ph/blog/deped-smea-school-monitoring-evaluation-adjustment) (DO 029, s. 2022, Sec. VIII, School M&E System, items h and f). A run chart kept every week gives the quarterly review something to look at besides two numbers.

### Mark what happened on the chart

Write the change and its date on the chart, where it starts, and note any week that is out of the ordinary: a typhoon, a class suspension, a fiesta, an examination week. Perla and colleagues note that "annotations of changes and other events can also be added to the run chart" (2011, p. 47). The Institute for Healthcare Improvement's run chart guide is more direct: "Annotate the chart, indicating when tests of change were initiated, so that it is easy to see the effect of changes on the measure. Also indicate any external events that may have affected the performance of the process" (IHI, 2017, step 10). A week far below the rest with a typhoon written beside it is an astronomical point with a known cause, not a failed project.

### When to draw a new median

After a shift has held, the old median no longer describes the process. Anhøj and Olesen's advice is to "lock the median and continue to collect and plot data while improving the process", and "If, or when, the target has been reached, calculate the new median or define the target as the new median" (2014, p. 12). On a chart with two medians, the rules are "applied separately to the data surrounding each median" (Provost and Murray, 2011, p. 90). The new median is the baseline for the next change, and the level the project has to hold.

## Making a run chart in Excel or Google Sheets

A spreadsheet is enough:

1. **List the weeks.** Column A holds the week number or the Monday's date, column B the measure for that week.
2. **Compute the baseline median.** In a cell of its own, take the median of the weeks before the change, for example `=MEDIAN(B2:B11)` for ten baseline weeks.
3. **Make a median column.** In column C, put the median cell in every row (`=$E$1` if the median is in E1), so the chart draws it as a flat line across all the weeks.
4. **Draw a line chart** of columns B and C, with markers on the points.
5. **Count the runs.** In column D, mark each week "above", "below" or "on" the median: `=IF(B2>C2,"above",IF(B2<C2,"below","on"))`. Read down the column for six "above" or six "below" in a row, skipping the "on" weeks.
6. **Label the change** with a text box at the week it started.

Keep one file per project and add each week's point as it comes in. The chart is for the team's weekly look, not only for the report.

## Common mistakes

- **Too few points.** Two or three points after a change cannot show a shift, which needs six. Wait, and keep plotting.
- **Moving the median after the change.** A median taken from all the weeks, including those after the change, sits halfway between the old level and the new one, and can hide a shift. Freeze the baseline median and extend it.
- **Changing the measure partway.** A new attendance rule, a harder reading passage or a new divisor makes the weeks after it a different measure. Note the change, and start a new baseline if it is large.
- **Reading every bump.** Week-to-week ups and downs are the process as it is. Only the rules' signals are evidence of change.
- **Choosing the start week after seeing the data.** Decide which weeks are the baseline before the change begins.
- **A small group.** One section of 30 or 40 learners moves several points in a week when a few learners are sick. The rules allow for this; a single good or bad week does not.
- **Treating the chart as proof of cause.** A shift says the measure changed. That the project changed it is a judgment the team makes from what else happened in those weeks, which is why the chart is annotated.

## Run chart or control chart?

A run chart answers whether the measure changed. It cannot say whether the process is stable: "To determine if a process or system is in a stable state, a Shewhart (control) chart is needed" (Perla, Provost and Murray, 2011, p. 49). A control chart adds limits computed from the data's own spread, and it catches a large change that lasts only a week or two, which a run chart can miss (Anhøj and Olesen, 2014, p. 10).

Two cases call for a control chart, or at least for caution with the run chart rules:

- **Group sizes that vary.** The rules assume each point stands on about the same number of learners: "each data point has a roughly equal denominator, + or − 25% of the average denominator size" (Provost and Murray, 2011, pp. 93-94). A section's enrollment barely changes over a term; an ARAL group that grows from 8 to 20 learners does not meet the condition.
- **Points that carry over.** The rules assume each week is independent of the last. A running total, such as the number of learners who have reached grade level so far, can only rise, so it shows a trend whatever the project does. Plot each week's own value, not the total to date. Provost and Murray say that when points depend on one another, the probability rules are "invalid" and the team should "Rely on only the visual analysis" (2011, pp. 98-99).

Anhøj and Olesen suggest "always studying process data with run charts before applying control charts" (2014, p. 10). For a school's first projects, a run chart is enough.

## What this means for school heads

1. **Ask every SIP project team for one weekly measure** of its result or of the change itself, taken the same way each week.
2. **Plot before the change.** Ten or more weeks of baseline from records the school already keeps, such as School Form 2.
3. **Test small, read the rules.** In Step 9, call a test a success when a shift appears after the change in the direction the project wants, not when an average rises.
4. **Keep plotting through the roll-out** and bring the charts to the quarterly program implementation review and to the mid-year and year-end checks of Step 11.
5. **Write the Project Monitoring Report from the chart**, with the weeks and the dates of any shift.

Records to keep: the weekly data file, the chart with its notes, the date each change started, and the decision taken at each check.

Reading a run chart is part of monitoring and evaluation, PPSSH Strand 1.7: indicator 1.7.2 asks a school head to "Utilize available monitoring and evaluation processes and tools to promote learner achievement". It also serves Strand 1.4, research and innovation: indicator 1.4.2 is to "Utilize relevant research findings from reliable sources in facilitating data-driven and evidence-based innovations to improve school performance." The [guide to PPSSH Domain 1](https://nash.com.ph/blog/ppssh-domain-1-leading-strategically) explains both strands.

## Frequently asked questions

### What is a run chart?

A run chart is a line graph of one measure over time, with a horizontal line at the median. Four rules, a shift, a trend, too few or too many runs, and an astronomical point, tell a real change from ordinary variation (Perla, Provost and Murray, 2011).

### What are the rules of a run chart?

A shift is six or more points in a row on one side of the median. A trend is five or more points in a row all going up or all going down. Too few or too many runs for the number of useful points is a signal, and so is an astronomical point, one week blatantly different from all the others. Points on the median are skipped (Perla, Provost and Murray, 2011, pp. 47-48).

### How many data points does a run chart need?

Perla, Provost and Murray say the shift and run rules "require more than 10 points before they are applicable" (2011, p. 48), and Anhøj and Olesen advise at least 12 and preferably 20 (2014, p. 12). For a SIP test, plot ten to twelve weeks before the change, from School Form 2 or other records, and at least six after it.

### What is the difference between a run chart and a control chart?

A run chart draws the median and applies counting rules. A control chart, or Shewhart chart, adds upper and lower limits computed from the data, and is the tool for judging whether a process is stable (Perla, Provost and Murray, 2011, p. 49). Use a control chart when the number of learners behind each point varies by more than about a quarter (Provost and Murray, 2011, pp. 93-95).

### Does DepEd require run charts for SIP projects?

No. DO 44, s. 2015 asks a team to compare data before and after a test and to monitor progress at mid-year and year-end, and DO 029, s. 2022 allows weekly monitoring, but as of November 20, 2026 no DepEd issuance names run charts. They are a way of doing what the SIP steps ask.

## Sources

**DepEd issuances**

- [DepEd Order No. 44, s. 2015, Guidelines on the Enhanced School Improvement Planning (SIP) Process and the School Report Card (SRC)](https://www.deped.gov.ph/2015/09/29/do-44-s-2015-guidelines-on-the-enhanced-school-improvement-planning-sip-process-and-the-school-report-card-src/), September 29, 2015: SIP Guidebook, Steps 9 to 11 and the Project Monitoring Report Form
- [DepEd Order No. 029, s. 2022, Adoption of the Basic Education Monitoring and Evaluation Framework](https://www.deped.gov.ph/2022/06/22/june-20-2022-do-029-s-2022-adoption-of-the-basic-education-monitoring-and-evaluation-framework/), June 22, 2022: Sec. VI, Types of M&E; Sec. VIII, School M&E System
- [DepEd Order No. 4, s. 2014, Adoption of the Modified School Forms (SFs) for Public Elementary and Secondary Schools Effective End of School Year 2013-2014](https://www.deped.gov.ph/2014/01/30/do-4-s-2014-adoption-of-the-modified-school-forms-sfs-for-public-elementary-and-secondary-schools-eeffective-end-of-school-year-2013-2014/), January 30, 2014: School Form 2, data element 18
- [DepEd Memorandum No. 014, s. 2021, Interim Guidelines on the Preparation, Submission, and Checking of School Forms for the SY 2020-2021](https://www.deped.gov.ph/2021/03/26/march-26-2021-dm-014-s-2021-interim-guidelines-on-the-preparation-submission-and-checking-of-school-forms-for-the-sy-2020-2021/), March 26, 2021: Annex 1, the guidelines printed on School Form 2
- [DepEd Order No. 009, s. 2026, Guidelines on the Implementation of the Three-Term School Calendar in Basic Education](https://www.deped.gov.ph/2026/04/16/april-16-2026-do-009-s-2026-guidelines-on-the-implementation-of-the-three-term-school-calendar-in-basic-education/), April 16, 2026: item 15 and Annex B
- [DepEd Order No. 018, s. 2025, Implementing Guidelines of the Academic Recovery and Accessible Learning Program](https://www.deped.gov.ph/2025/06/13/june-13-2025-do-018-s-2025-implementing-guidelines-of-the-academic-recovery-and-accessible-learning-program/), June 13, 2025: Table 1
- [DepEd Memorandum No. 064, s. 2025, Implementing Guidelines for the Academic Recovery and Accessible Learning Program in Reading for Key Stages 1 to 3](https://www.deped.gov.ph/2025/08/01/august-1-2025-dm-064-s-2025-implementing-guidelines-for-the-academic-recovery-and-accessible-learning-program-in-reading-for-key-stages-1-to-3/), August 1, 2025: Encl. 3, items 32 and 34

**The run chart method**

- Perla, R. J., Provost, L. P., and Murray, S. K. (2011). [The run chart: a simple analytical tool for learning from variation in healthcare processes](https://doi.org/10.1136/bmjqs.2009.037895). *BMJ Quality & Safety*, 20(1), 46-51: pp. 47 to 49 and Table 1
- Provost, L. P., and Murray, S. K. (2011). *The Health Care Data Guide: Learning from Data for Improvement*. Jossey-Bass: pp. 77 to 99
- Anhøj, J., and Olesen, A. V. (2014). [Run charts revisited: a simulation study of run chart rules for detection of non-random variation in health care processes](https://doi.org/10.1371/journal.pone.0113825). *PLOS ONE*, 9(11), e113825: pp. 3, 5, 10 and 12
- Anhøj, J. (2015). [Diagnostic value of run chart analysis: using likelihood ratios to compare run chart rules on simulated data series](https://doi.org/10.1371/journal.pone.0121349). *PLOS ONE*, 10(3), e0121349: Table 2 and p. 8
- Institute for Healthcare Improvement (2017). [*QI Essentials Toolkit: Run Chart*](https://www.ihi.org/sites/default/files/QIToolkit_RunChart.pdf): step 10
- Langley, G. J., Nolan, K. M., Nolan, T. W., Norman, C. L., and Provost, L. P. (1996). *The Improvement Guide: A Practical Approach to Enhancing Organizational Performance*. Jossey-Bass: p. 4

**Improvement in schools**

- Bryk, A. S., Gomez, L. M., Grunow, A., and LeMahieu, P. G. (2015). *Learning to Improve: How America's Schools Can Get Better at Getting Better*. Harvard Education Press: pp. 100 and 116
- Yeager, D., Bryk, A., Muhich, J., Hausman, H., and Morales, L. (2013). [*Practical Measurement*](https://www.carnegiefoundation.org/practical-measurement/) (working paper, draft). Carnegie Foundation for the Advancement of Teaching: pp. 2, 12 and 13
- Wolfenden, F., Walley, P., Agbaire, J., and Hartley, J. (2023). [*Strengthening School Leadership Towards Improving School Resiliency: An Improvement Science Approach*](https://oro.open.ac.uk/92173/1/Strengthening-School-Leadership-Towards-Improving-School-Resiliency-Final-Technical-Report.pdf), final technical report. FIT-ED: p. 34

*Written by Mark Anthony Llego for NASH Reviewer. Sources: DepEd Orders No. 44, s. 2015, 029, s. 2022, 4, s. 2014, 018, s. 2025, and 009, s. 2026; DepEd Memoranda No. 014, s. 2021, and 064, s. 2025; Perla, Provost and Murray (2011); Provost and Murray (2011); Anhøj and Olesen (2014); Anhøj (2015); Langley and others (1996); Bryk and others (2015); Yeager and others (2013).*

---

© 2026 Mark Anthony Llego, NASH Reviewer. All rights reserved. You may quote up to 100 words from this article with the author's name, its title and a link. In research, cite it as: Llego, M. A. (2026, November 20). *Run Chart Rules: How to Tell If a SIP Project Worked*. NASH Reviewer. https://nash.com.ph/blog/run-chart-rules-sip-projects. Any other reuse of the article, its charts or its data needs written permission: ask through our [contact form](https://nash.com.ph/contact). Terms of Service: https://nash.com.ph/terms#website-content-and-data
