Always ready to pivot
David has been the person who builds the database for as long as he has had a job, and he has pivoted more than tables along the way. At an internet startup during college, that meant learning Microsoft Access. As a teacher of English as an additional language at Lee High School in Houston, it meant writing randomized math worksheets in Excel so every student got a fresh set of problems, and a master's thesis on computer-adaptive testing for English language proficiency. At TNTP, where he led the NYC Teaching Fellows program that trained one in four of New York City's math teachers and one in six of its special education teachers, and then launched teacher pipeline and principal training programs in Newark, Pittsburgh, and Washington, D.C., it meant living in Access, SQL, and SPSS every day, because the questions the programs were asking could not be answered any other way.
That habit of reaching for the data is the through-line. The tools changed, from Access to PostgreSQL, from SPSS to Python and R, from Excel macros to Apps Script and agents. The instinct did not.
Insight: leading indicators of school culture
At TNTP, coaching principals in the D.C. charter sector in partnership with the DC Schools Fund, David's team kept running into the same problem. Leaders could see test scores and teacher turnover, but only after the fact. What they needed were leading indicators: valid, reliable signals of school culture that a principal could act on in October rather than regret in June.
David designed and led the development of Instructional Culture Insight to be that signal. The constructs and item bank went through a two-year independent validation with the American Institutes for Research. The result, the Insight Index, proved to be a reliable predictor of both student growth and teacher retention.
Just as important was what happened after the survey closed. David led the development of the norm-referenced reports and the school-leader trainings that turned them into action plans, each initiative pegged to a specific construct. The reports were simple dashboards by domain and item. A principal could scan for the domains where their school differed from its region, district, or network, and see how it stacked up against the top-quartile schools in its city. Open-ended comments, scrubbed of identifiers, sat beside the numbers so leaders could unpack and challenge their own interpretations with a depth the quantitative metrics alone could not give them.
Insight went on to be used by more than 3,300 schools. What the data showed became Greenhouse Schools: How Schools Can Build Cultures Where Teachers and Students Thrive (TNTP, 2012), which David co-authored. He also co-authored Fixing Classroom Observations (TNTP, 2013) and Making Licensure Matter (Colorado Department of Education, 2013).
If you have used the survey dashboard we built for a national charter network, or any of the survey systems Integral Ed has built since, you have seen Insight's DNA: comparison groups the school chooses, hover text for every item, and the comments right there next to the charts.
What we believe about data
We empathize deeply with what our partners are trying to achieve, and we believe data should be used in service of serving communities, schools, teachers, and students better. In practice that means four things.
- Simple signals. We would rather give a leader three findings they can act on this month than thirty they can admire.
- Action over accountability. Leading indicators help organizations steer. Lagging indicators confirm and validate hypotheses and help set the leading indicators. Systems that rely on lagging indicators to guide decisions are stuck steering through the rearview mirror.
- Spotlight the positive. David was powerfully influenced by the theory of positive deviance: in every system, some people facing the same constraints are getting better results. We lead partners to use data to find those outliers succeeding in a similar context with the same constraints.
- Depth behind the numbers. Interviews and case studies of positive outliers dig into what they do differently. Open-ended survey responses belong in the report, not in an appendix. They explain the numbers and help leaders turn a hypothesis into a plan.
Why the data team sits inside a design studio
David founded Integral Ed in 2011. It has grown into a twenty-person learning and media design studio with more than 250 projects for schools, charter networks, nonprofits, and education publishers, including KIPP Foundation, EL Education, the Danielson Group, and Project Lead The Way. He personally leads the studio's survey, data aggregation, automation, and reporting work, from instrument design through analysis and synthesis, and has built dashboards and custom reporting for KIPP Foundation, Charlotte-Mecklenburg Schools, KIPP Forward, the DC Schools Fund, and EdTech marketing teams.
Sitting next to instructional designers, writers, and media producers changes how the data work turns out. A dashboard is a piece of communication, and the people who build training and storytelling for a living are good at asking who will read it, what they will do next, and what will make them stop. It also means the work can keep going past the dashboard when it needs to: into a learning platform, a report, video, or resource hub, or a course.
David leads the data work himself, hands on, backed by Integral Ed's designers, developers, and project leads, several of whom he mentored from their first client project to running their own. Twenty years with these kinds of datasets is what lets him set up the right checks on day one and know which signals are worth surfacing on a dashboard, and which are noise.
Agents and automation, responsibly
Integral Ed uses AI agents in two places, and keeps a hard line around both.
In data workflows, agents generate scripts that check a dataset against the common and likely pitfalls before anyone analyzes it; infer missing values by triangulating other data points and assigning a degree of confidence; and fuzzy-match datasets that share no primary key, where identifiers such as school names come in variants, again with a stated confidence for a human to review.
In learning, agents role-play the socially demanding parts of a job inside a course, such as delivering feedback, running a reading conference, or student-centered career counseling, and give agentic feedback on a draft or a response against a client's rubric or a set of measurable goals. That work lives on the eLearning side of the studio.
The line: client data stays in the client's environment. We help organizations working with sensitive data adopt compliant automation without sharing anything with outside LLMs or moving it beyond the tools they already trust, whether that is Google Drive, Microsoft Teams, or Dropbox.