Data Science At Learnmetrics 1: The Recipe Builder

Julian Miller
edtechintersect
Published in
4 min readMay 23, 2018

To kick off our summer blog series, we first want to begin with the foundational principles that led to our recipe builder when collaborating with partner schools:

  • Each school is unique and has different workflows and measures for success. Cookie cutter dashboards inevitably fall short.
  • Each stakeholder needs access to different information, whether it be a teacher viewing metrics for their classroom, a student or guardian looking at their historical data, or a school principal comparing longitudinal attainment across grade level cohorts.
  • For almost every school, a tool that requires an IT department to write code or database queries is a non-starter.
  • Push notifications are valuable for staff that are pressed for time during the school day.

We need a tool that can be fully configured to a school’s needs, to provide customizability otherwise limited to writing code, without writing a single database query or line of code. This is the core of the recipe builder — users create a flowchart composed of action blocks that represent the workflow, and we compile it into the code that does the heavy lifting under the hood. Experimentation and rapid prototyping of new workflows is painless — add on new blocks or delete those no longer needed and we recompile it to reflect new changes.

Our partner schools have used this to streamline workflows for meal tracking, attendance submission, and RTI tier management, but we’ll consider the case of a student behavior referral:

When a teacher submits a new behavior referral, we want to trigger a series of actions for each stakeholder. If the incident was severe enough to require guardian contact, we want to contact the guardian and provide a link where they can log in to view their child’s behavior along with other pertinent information. If the student has an IEP, we want to pull up their case manager’s contact information from the data warehouse and notify them. The referral is sent to an administrator, who can finalize any disciplinary action and automatically send push notifications to stakeholders when it is complete. By combining a behavior referral with other student information all in a single view, staff can make well informed decisions without having to spend any extra time or effort searching across systems.

We’ll end with a common theme and the variations we’re able to support with a modular, composable recipe engine. Multi-tiered systems of support/response to intervention (MTSS/RTI) and personalized learning have grown in popularity to provide students additional resources for growth outside of the traditional classroom lecture method. But schools identify at-risk students, typically with a custom, in-house workflow. We highlight two cases below:

In our first example, a school district maintains an in-house rubric as a guideline to assign students to RTI tiers after testing windows close in the fall, winter, and spring. We integrate the data through PDF/spreadsheet reports provided by testing products, as well as teacher-level data entry through the app. We convert the rubric to a recipe that calculates the student’s “at-risk coefficient”, condensing several testing dimensions into a single value for math and language arts, for staff to review (along with attendance, gradebook, and demographic information) at a Data Day to consider students for tiered support. Once assigned, we can grant custom access to students in at-risk tiers to interventionist staff. Aggregate statistics are shared up to the district office to make sure the program is administered consistently across the schools.

In the second example, a private school uses an in-house rubric to assign students a skill level based on proficiency scores in NWEA MAP, so that a student from each skill level is randomly assigned to each project group. This is easily extensible, for example to guarantee groups maintain the same demographic diversity as the student population.

These recipes aren’t complex, but access to a data warehouse and the recipe builder allows us to turn around these requests in a matter of minutes, as soon as the reports are available to ingest. Next time, we’ll get into some deeper analytics and consider correlation matrices in the Recipe Builder, which can elucidate the relationship of key student metrics to one another in a single visualization.

In the meantime, if you’d like help mapping your school and district workflows, head to www.learnmetrics.com and send us a message. We’d love to hear from you!

--

--

Julian Miller
edtechintersect

Ex-high school teacher/co-founder @Learnmetrics where we work to rebalance the relationship between education, data and edtech.