Data for Turbocharging Management Decisions
Improving cost-effectiveness of personalized learning, staff schedules and academic return on investment
By Nathan Levenson/School Administrator, January 2016
“Co-teaching is great! It’s the best of both worlds!”
“No, it’s expensive and I don’t think it works for most students.”
It was an old debate with predictable opponents and proponents. The discussion ended like it had in years past. Hard feelings, uncertainty as to who was right, and the continuation of the status quo. Adding Big Data changed everything — the discussion, the ambiguity, the budget and the services for students.
|Nathan Levenson, president of the District Management Council in Boston, Mass., believes schools can improve reading outcomes by creating detailed profiles of students’ strengths and shortcomings. (Photo by Richard Viard.)
Big Data, finding insights from large amounts of information drawn from multiple sources and incorporating many types of data, can help district and school administrators be more effective leaders and make better decisions.
Big Data in K-12 is less of a new idea and more of an expansion (in a big way) of an old idea. School districts are awash in data, schools are data-driven, teachers are often members of data teams and district leaders are constantly reviewing reports and analyses. Despite the sea of information, this is lots of data, but not Big Data.
The next evolution of managing schools and districts will be powered by management data
, which differs from today’s ocean of numbers in two ways:
It’s the information leaders need, not the data they have. Its primary purpose is to help school and district leaders, not the state department of education, grant administrators, teachers or others.
It’s much more fine-grained. Rather than, for example, looking at a special education teacher’s caseload at the start of the school year (one data point), it’s every 30 minutes of the teacher’s day, updated every week — more than 2,500 data points for a single teacher!
Management data incorporating Big Data has many uses. These include:
- Cost effectively personalizing learning on a systemic basis;
- Building student-centered, cost-effective schedules for reading teachers, special educators, paraprofessionals, English language learning teachers, etc.; and
- Conducting rigorous academic return on investment analysis of what works, cost effectively, for which students.
As schools turn to iPads, Chromebooks and adaptive curricula to personalize learning, they might consider a less-costly approach made possible by Big Data. Great teachers long have personalized instruction without hardware or software. As superintendent in Arlington, Mass., I recall asking one elementary teacher how, year after year, she achieved such outstanding reading results. Nearly every child in her class made more than a year’s growth, far better results than her colleagues in the same school or across the district.
“I personalize my instruction,” she answered simply.
This surprised me because there was no hardware or software to be seen. She went on to explain that she kept very detailed notes on each student, as she pointed to a box of neatly arranged file folders. She knew the strengths, interests, needs and learning style of each student. She knew that Mark struggled in phonics, consonant sounds in particular. He liked sports and found comfort with clearly defined tasks. She adapted her teaching style to meet his needs and placed Mark in a small group with other students in her class who also struggled with consonant sounds. This great teacher managed a lot of data in her head (and in the file folders). This was a case of small data leading to big results.
When the principal heard the story, he wondered how to scale these practices across the school. Using nothing more sophisticated than Post-it notes and lots of manual labor, more teachers began tracking student needs in fine detail, creating intervention groups across classrooms, finding the students with similar needs no matter who was their classroom teacher, and flexibly pairing students with teachers based on learning and teaching style. Every Sunday, some teachers would plan groups and lessons as they shifted color-coded Post-its with key facts about their students.
This worked well, sort of. For teacher teams willing to scribe notes in detail and plan over the weekend, students benefitted greatly. One such group of teachers was able to increase the number of students scoring advanced in math from 38 percent to 72 percent. But not every teacher had the time, skills or inclination to do this. Big Data would have helped.
Imagine a school district that creates a data stew with the following ingredients: a detailed profile of student strengths and needs, down to which specific phonetic sounds have been mastered; a student’s preferred learning style; data on every teachers’ past success in teaching specific strands (phonics vs. fluency); and their dominant teaching style.
Finally, add a scheduling algorithm that quickly creates new intervention groups each week that match students with like needs and learning styles to teachers with the best fit of expertise and approach. This is Big Data helping to personalize instruction in a low-cost, scalable way.
Special Education Spending
Big Data also can help manage special education costs more effectively, but this requires really Big Data, and unfortunately few school districts have even 1/100th the management information they need. In most districts, each special education teacher (and English language learning teacher or reading interventionist) makes his or her own schedule, decides when to serve which students and determines which students to group.
Why do these staff members have such “freedom” when most other teachers are handed schedules and class lists? In part, it’s just too hard for a centralized scheduler to know which students can be pulled from class, which students have similar needs, which individual education plans allow small groups and so on. Each teacher spends days collecting, often on lined paper, the required information by visiting classroom teachers, reviewing the IEP files and talking to colleagues. They then sketch out their schedule by hand. While doing the best they can, the results typically are not optimal. In my firm’s studies of more than 75 school districts big and small, the typical special education teacher’s schedule includes:
- Less than 50 percent of the school day working with students;
- Pullout sessions during math or reading, thus denying struggling students core instruction, unintentionally dragging them further behind; and
- Much more one-on-one instruction than required by the IEP.
This less-than-desirable situation isn’t anyone’s fault; it’s simply too hard to build great schedules by hand. But Big Data can change that. To build special education schedules that are good for students, teachers and taxpayers requires a lot of data. In a small elementary school, for example, that could be more than 25,000 data points. That’s a lot for paper and pen, even Excel, but it’s a modest task for computer software.
These data come from dozens of sources, a common attribute of Big Data, such as classroom schedules (when does Mrs. Smith teach reading), lunch schedules, specials schedules (when does Sarah go to physical education), the IEP database, a student’s reading scores and so on.
I have worked with districts using Big Data, enabled by web-based technology, to manage and build these schedules, which eliminated pulling students from core instruction, grouped students with like needs and reduced workloads for many staff. It also reduced staffing requirements, without changing a single IEP, in selected roles by 40 percent in one suburban district with a few thousand students, by 20 percent in a midsized urban district of 10,000 students and similarly in larger districts of 30,000-plus students. The algorithms and data needed to efficiently build optimal schedules exceeds most individuals’ ability. Using Big Data requires specialized tools.
Knowing What Works
Big Data also can help settle lots of long-running disagreements. Knowing in most districts what programs, strategies or efforts are achieving the intended results is mostly a matter of professional judgment because the data to rigorously evaluate effectiveness and cost-effectiveness are lacking and overwhelming to collect. Differing opinions are hard to reconcile and often lead to inaction or hard feelings or both.
One midsized suburban district was all too typical. Some staff had assumed (for seven years) that co-teaching at the middle school was effective. The principal, many teachers and some administrators strongly supported the effort. Others had their doubts. Each year, typically during budgeting or when lackluster test scores were reviewed, the debate repeated itself. Persuasion and majority rule carried the day.
Despite a mass of test scores, the district lacked the fine-grained data to distinguish success or failure of co-teaching itself from the impact of the teacher or other factors. Big Data helped settle the debate. With some outside help, the district built a data model that analyzed data to separate cause and effect.
Again, data from multiple sources needed to be combined, including enrollment by course and teacher, IEP data, class assignments, semester grades, past performance, growth in other subjects such as English, achievement of students not on IEPs, costs and teacher growth scores for past years. In total, there were tens of thousands of data points for a single school.
The insights gained were definitive and surprising to many. The upshot: No gains resulted from co-teaching despite seven years of strong support based on professional judgment. Moreover, the findings revealed that the skills of the general education teacher (with or without a co-teacher) mattered most. Simply pairing struggling students with a handpicked set of teachers was quite beneficial. It also showed that an alternative intervention that cost about 75 percent less than co-teaching was a bit more effective. The debate ended, and better decisions were possible.
Gathering the data you need, combining multiple data sources and using sophisticated algorithms to match students to teachers, build schedules or find useful insights can be a lighthouse in a sea of data. It can light the path to better outcomes at lower cost by helping school district leaders make even better decisions.
, a former superintendent, is president of the District Management Council in Boston, Mass. E-mail: firstname.lastname@example.org