Walk-to-math, a reboot.

The Old “Walk to Math” — and Why It Earned Its Critiques

In the older version of walk-to-math, students were sorted at the start of the year into stable groups based on a single diagnostic, and those groups often held through the year. Groups moved between classrooms for the entire math period, not a portion of it. The lowest group typically received a below-grade curriculum in the name of “meeting students where they are,” and the teacher assigned to that group was often determined by scheduling convenience rather than math strength. The model looked like differentiation on paper. In practice, it looked a lot like tracking.

The research on what happens next is consistent. A 2022 working paper from Antonovics, Black, Cullen, and Meiselman, using a decade of Texas data, found that exposure to more homogeneous math classrooms modestly helped high-achieving students while neither helping nor harming lower-achieving ones. Read carelessly, that’s an indictment of grouping itself. Read carefully, it’s an indictment of what was happening inside the low group. The group didn’t hurt the students placed in it — but nothing in the design was engineered to accelerate them either. Without a real intervention program with a path back to grade level, homogeneous grouping is a structure without an engine.

A 2021 TNTP study of classrooms using the Zearn platform sharpened the point. Classrooms that approached below-level students through remediation — going back to fill earlier gaps — completed 27 percent fewer grade-level lessons than classrooms that used just-in-time scaffolding to keep students inside grade-level work. In the tracked version of walk-to-math, remediation was often the default, and acceleration was the exception. The result is predictable.

A Different Model, Built on Thirty Years of Evidence

While walk-to-math was accumulating its reputation, reading researchers were quietly figuring out what made cross-classroom regrouping actually work. Programs in the Success for All tradition — developed at Johns Hopkins by Robert Slavin and Nancy Madden, and studied in dozens of experimental trials across thirty years — showed that the grouping structure is a scheduling shell. What drives outcomes is what happens inside that shell: explicit and systematic instruction, diagnostic assessment tied to specific skills, regrouping every eight weeks or so, and the school’s strongest instructors assigned to students with the biggest gaps (Slavin, Madden, Chambers, & Haxby, 2009). The research base on Success for All is one of the largest for any K–5 reading program, and its core design principles have shaped a generation of work on small-group literacy intervention since.

The reading research offers three findings that transfer directly to math. First, grouping is flexible, not fixed: students move in and out of groups as their skills develop, sometimes multiple times a year, so no student is “in the low group for fourth grade.” Second, the curriculum doesn’t lower: students in intervention get scaffolds into grade-level content, not a lower grade’s content. Third, the best teachers teach the hardest content: schools that assign their strongest instructors to students with the deepest gaps consistently outperform schools that assign by convenience.

Targeted Instruction

Within a single classroom, you may not have enough students with the same specific gap to form a genuine skill group — three students wrestling with place value, two with multi-digit subtraction, one ready for fraction work. Across a whole grade, you usually do. The meta-analytic finding that small-group reading programs produce a 20 percent higher effect size when they target a specific skill (Hall & Burns, 2018) has an obvious parallel in math: the more precisely instruction matches the gap, the more it moves the needle.

Three of the six IES Practice Guide recommendations — concrete and semi-concrete representations, number line work, and timed fluency activities — are hard to deliver to a wide-skill-range group. Productive number-line instruction is difficult with students who already have strong number sense sitting alongside students who don’t. In a skill-matched group, every student is working on the same representation at the same time, and the teacher’s moves can be precise.

Flexible and Dynamic

The central problem with older walk-to-math — and the one the NBER paper captured — was that there was no mechanism for students to move up. Walk-to-intervention treats movement as the default. When a student masters a skill, they move to a group working on a more advanced skill. That mechanism is the engine the older model was missing. It also resolves the tension between “meet students where they are” and “hold the line on grade-level expectations”: the group meets the student where they are this month, and the expectation is that they move.

Many elementary teachers report feeling less confident teaching math than teaching reading, and the teachers with the most math anxiety often end up with the students who need the most. Walk-to-intervention distributes math teaching across every adult on the grade-level team — classroom teachers, specialists, and paraprofessionals — which, over a year, builds the kind of collective math capacity that Donohoo, Hattie, and Eells (2018) identified as one of the strongest predictors of student achievement in the research literature. The same structure that helps students helps teachers.

Where the Real Gaps Often Start

The single most consistent thing we hear from elementary leaders across the districts we work with is that kindergarteners are arriving with fewer pre-requisite math skills than in past years. Counting past ten, one-to-one correspondence, subitizing small quantities, understanding cardinality — foundations that first-grade curricula tend to assume are in place are, increasingly, not. And it’s not just a kindergarten phenomenon; those unaddressed foundational gaps compound upward through the grades.

This is a design problem for math intervention specifically. Most commercial intervention programs start at grade-level content with scaffolds, or step back one grade level. They assume the number sense below that floor is already in place. For a student who entered kindergarten without subitizing or one-to-one correspondence, a second-grade intervention that begins with two-digit addition — even a well-scaffolded one — is building on sand.

This is one of the strongest arguments for structuring math intervention around specific skills rather than grade-level bands. In a walk-to-intervention structure, a first-grader who needs more time on cardinality can be in a group pitched at exactly that skill, while other first-graders work on addition strategies or place value. The instruction meets the student at the actual pre-requisite they’re missing, not at the starting point a commercial program assumes. And when that student consolidates the foundational skill, they move. In practice, this often means that teams using walk-to-intervention for math need to make sure their intervention materials cover further back in the skill sequence than their core curriculum does — and that the teachers leading the earliest-skill groups have the early-childhood math knowledge to make that instruction precise.

The Design Decisions That Separate Walk-to-Intervention from Tracking

The difference between walk-to-intervention that works and walk-to-intervention that reverts to tracking comes down to a handful of design decisions made at the front end. A useful side-by-side:

Dimension	Old “Walk to Math”	Walk-to-Intervention for Math
Duration	The entire math period	30 minutes of a 60–90 minute block
Grouping basis	Stable ability tracks	Skill-specific, tied to the current unit
Regrouping cadence	Rarely; often once a year	Every 3–6 weeks or at unit boundaries
Content for lowest group	Below-grade curriculum	On-grade content with scaffolds and built-in acceleration
Core Tier 1 instruction	Replaced by the walk-to period	Preserved for every student before the regrouping window
Assignment of strongest teacher	By scheduling convenience	To the group with the deepest gaps

A district or school considering the model can test its design against those six dimensions. If the answers drift toward the left column, the implementation will drift toward the outcomes the old model is rightly critiqued for. If the answers fall on the right, the model has the same structural logic that a generation of reading research has validated.