Here’s a scenario that might feel familiar. You get your state or provincial math assessment results in August. You spend September analyzing them. You use October to design a response. By November, you’re implementing something different — and the students in front of your teachers right now are not the same students those scores were measuring. The moment when the data could have changed instruction has already passed.
This is the fundamental problem with designing a math improvement system around a single type of data: the feedback loop is too slow to actually guide the work.
Most districts aren’t data-poor.
They’re data-imbalanced.
They have one very loud instrument — standardized testing — and they treat it as the primary signal for what’s working and what needs to change.
Meanwhile, they’re surrounded by other data they’re not collecting systematically, not analyzing together, and not using to adjust instruction before it’s too late to matter.
There’s a better model. And once you understand it, it changes how you design data cycles, how you talk to principals about evidence, and how you help teachers connect what they observe on Monday to what you’re measuring in May.
Satellite Data, Map Data, and Street Data: What Each Type Is For
The Street Data framework, developed by Shane Safir and Jamila Dugan, describes a three-level hierarchy for educational data that has real practical implications for math improvement specifically. Understanding all three levels — and what each one is for — is the foundation of a coherent math data system.
Satellite data is what most people mean when they say “data.” It’s standardized assessment results: state tests, national benchmark comparisons, proficiency rates by grade band and demographic group. Satellite data gives you the wide-angle view. It tells you where your system is relative to external benchmarks. It’s essential for accountability, for identifying which schools need the most support, and for tracking long-term trends.
But satellite data has a critical limitation: you get it once a year, months after the instruction it’s measuring has already happened. By the time you can act on it, you’re looking at a different set of students in a different school year. It’s like trying to steer a car by looking in the rearview mirror twelve months after you drove through.
Map data operates at a shorter time horizon. These are your benchmark assessments, common formative tasks, interim checks — measures collected two, three, or four times a year within the school year. Map data lets you see learning trends across a grade level or school, identify where specific standards are creating persistent gaps, and make adjustments mid-year rather than waiting for August results.
Map data is more actionable than satellite data. But it still has a lag. A benchmark assessment administered in February tells you where students were in February. It doesn’t tell you what happened in yesterday’s lesson, or why a particular student is stuck, or what a teacher would need to see to know whether their instructional change is landing.
Street data is what you can see when you’re in the room. It’s observable changes in student behavior — the questions students ask, the strategies they reach for, the ways they talk about mathematics with each other, whether they persist when a problem is hard. Street data shows up in learning walks, in coach observations, in the quality of student discourse during a number talk. It’s qualitative, contextual, and immediate.
Street data is what tells you whether the map is actually being followed. It’s the evidence that instruction is shifting, that students are developing new habits of mind, that the professional learning your teachers participated in last Thursday is showing up in how they’re facilitating discourse on Tuesday.
Why Standardized Test Data Alone Can’t Drive Math Improvement
There’s a reason most districts over-rely on satellite data. It’s comparatively easy to collect and analyze. It comes pre-packaged in reports. It gives administrators a clear number to report to boards and communities. It’s familiar.
But the cost of that over-reliance is significant.
When satellite data is the primary — or only — data source informing decisions, you end up making system-level changes in response to a 12-month-old signal. You adopt new resources or adjust pacing guides based on what was happening in your classrooms last year, not what’s happening now. And because the feedback loop is so long, you can’t know whether the changes you made are working until the next test cycle — another year later.
Safir and Dugan (2021) argue that this over-reliance on satellite data is one of the primary equity problems in school systems: it causes leaders to focus on outcomes rather than experiences, on gaps rather than strengths, and on what students don’t know rather than what they’re actively building. For math specifically, this produces improvement efforts driven by proficiency rates rather than by the instructional conditions that produce proficiency.
Meanwhile, the data that could actually change teaching in real time — what students are doing in math class right now — goes largely uncollected, unanalyzed, and unused.
How to Structure a Three-Level Math Data System in Your District
Building a genuinely three-data math improvement system doesn’t require new tools. It requires deliberate structures for collecting and using each type of data at the right time horizon.
For satellite data, the question to anchor your work is: what does this year’s testing data tell us about where we need to concentrate our three-year improvement focus? These numbers are strategic, not operational. They inform priority-setting — which grade bands, which standards clusters, which student populations need the most intensive support — but they shouldn’t drive week-to-week or even month to month instructional decisions.
For map data, the question is: what can we see about student learning in this unit, this quarter, this semester that tells us whether our instructional approach is working? This is where PLCs should be spending most of their analytical time — looking at common formative assessments, and planning targeted instructional responses. Research on effective mathematics PLCs shows that centering collaborative time on actual student work, effective instruction and formative assessment results — rather than logistics or pacing — produces the strongest shifts in teacher practice (Cobb et al., 2018).
For street data, the question is: what are we seeing in classrooms right now that tells us whether the work is taking root? This requires scheduled learning walks with shared look-fors, coaching observation protocols tied to the same instructional priorities you’re measuring in benchmarks, and PLC time that includes teacher reflection on what they observed during instruction — not just what they scored on an assessment.
The critical design principle is alignment: the look-fors in your learning walk should connect directly to the standards you’re benchmarking in your map data, which should connect directly to the priorities in your satellite data trends. When all three data types are measuring the same instructional priorities at different time horizons, you get a coherent picture of what’s working and where to focus.
How a Balanced Math Data Framework Changes Leadership Conversations
One of the most practical benefits of building fluency with the three-data framework across your leadership team is that it changes how you talk about progress.
When satellite data is the only language in the room, every conversation about student learning eventually collapses into “our scores.” That framing creates anxiety, defensiveness, and a tendency to focus on what didn’t work rather than on what needs to be built. It also makes it nearly impossible to have productive conversations about what teachers should be doing differently on Tuesday, because state test data doesn’t answer that question.
When your leadership team is fluent in all three types of data, the conversations become layered and generative. What does our map data tell us about where to focus in Q3? What did our learning walks last month surface about how well students are engaging with the open-ended tasks in our curriculum? What do both of those things together tell us about whether our coaching model is producing the instructional shifts we’re aiming for?
That’s a completely different conversation. And it’s a conversation that can actually change what happens for students in the next thirty days, not the next twelve months.
Your Next Step
If your data system is currently built primarily around satellite data, starting with map data is the highest-leverage next move. MMM’s free Math Coherence Compass Training includes a section specifically on your data architecture — which types of data you’re currently using, how aligned they are to your instructional priorities, and where the gaps are. It’s a practical starting point for building a more complete picture.
Access that coherence compass training here.
References:
- Safir, S., & Dugan, J. (2021). Street Data: A Next-Generation Model for Equity, Pedagogy, and School Transformation. Corwin.
- Cobb, P., Jackson, K., Henrick, E., Smith, T.M., & the MIST Team. (2018). Systems for instructional improvement: Creating coherence from the classroom to the district office. Harvard Education Press.
