Once co-founder Matt Pasternack on helping children become fluent readers—and opening up everything that comes next.
INTERVIEW | by Victor Rivero
What does helping schools connect their software have in common with securing elections—or teaching a child to read? For Matt Pasternack, each experience has offered lessons in what makes a solution work, and what it takes to bring it to more people.
A former middle school English teacher, Clever’s first employee, and co-founder of VotingWorks, Pasternack returned to education to co-found the Once Early-Reading Program. Once combines software, training, and coaching to help parents, caregivers, and school-based instructors deliver individual reading instruction to young children. In this conversation with EdTech Digest, he reflects on his path, the challenge of translating reading research into everyday practice, and a future in which more children arrive at kindergarten already reading.
You started your career in the classroom as a Teach For America corps member teaching middle school English, and later moved into technology and operations, including scaling Clever and co-founding VotingWorks. How did that trajectory—from the classroom to election security and back to early childhood education—shape your philosophy on scaling structural solutions for kids?
Great question. I’ve become a student of product-market fit over the years, and have gotten to experience what it looks like—and doesn’t—in different contexts.
‘I’ve become a student of product-market fit over the years, and have gotten to experience what it looks like—and doesn’t—in different contexts.’
Looking back at my classroom days teaching middle school English in West Harlem in the mid-2000s, what stands out in retrospect is how much variability there was between classes purportedly teaching the same subject to similar kids.
To be fair, those were the Balanced Literacy days, but my memory is that in reading we were expected to have a warm-up, a mini-lesson on metacognition, and independent reading time.
Beyond that, I could choose whatever I wanted to teach and use any books I wanted. We were also supposed to do writers’ workshop. I neglected a lot of that because I really wanted to get reading right.
I mean, I was working 14, 15 hours a day during the week, and a solid chunk of the weekends, but the idea that a 22-year-old was literally writing his own reading curriculum is preposterous to me in retrospect. In the 20 years since the classroom, I’ve hired people for a lot of jobs, and I’ve never given an entry-level hire a job of that complexity, with that much scope, and so few metrics on whether what they were doing was working. I know that sounds like a divergence from the theme of product-market fit, but I’ll show how it relates when we get to my current job.
After teaching, I was the first employee at Clever—SSO and rostering for K–12—and I got to scale a business where we hit product-market fit out of the gate. It was unreal. The common wisdom was that you couldn’t scale quickly in K–12, but every district IT department was suffering from hundreds of hours of data-transfer problems with their vendors and no budget to fix it, and every vendor’s implementation in K–12 was held up for weeks or months by data-transfer problems. So we made data transfer free to districts and sold our solution to the districts’ vendors and grew like wildfire because each district introduced us to all of their vendors and each vendor introduced us to all of their districts. I can’t think of a better illustration of virality in the public sector.
That being said, one thing I really had to figure out was that even though our product mechanic led to virality, to sustain and nurture the relationships we created, we needed to field sizable account management teams to manage both the schools and districts and the applications and vendors. Since then, I’ve never deviated from the belief that scaling structural solutions for kids requires understanding how to scale relationships well. That’s not a pure science. It requires having the right people on the team.
‘I’ve never deviated from the belief that scaling structural solutions for kids requires understanding how to scale relationships well.’
When I left Clever, Clever’s former head of engineering and I started VotingWorks. I had no idea that we were about to walk into two entirely different public-sector product-market fit scenarios.
We initially set out to exclusively build voting machines. I had no idea how complex a sale that would be. I thought K–12 annual, seasonal sales timelines were hard. That doesn’t begin to compare to the seven- to 20-year sales cycles for voting machines. And all sales for all voting-machine buyers in a state happen at the same time, so staffing a sales operation is really hard. And that doesn’t even account for the years required to get initially certified to sell in every state except for one, Mississippi, which is where we started.
At the same time, VotingWorks received Department of Homeland Security funding to rebuild and scale a particular type of statistically validating vote-auditing software. The need was massive, no one was meeting it, and in short order we were serving a third of U.S. states at the state level! We got national press overnight when we had a team on the ground in Georgia supporting the audit of the Georgia 2020 election. The real deal.
There we were, a single small team, with near-overnight product-market fit on one product and a long mountain to climb on the other. We were the constant. The market was the variable. Some young founders now think that if you work 12 hours a day, six days a week, you can solve any problem. That wasn’t my experience.
So back to education, I left VotingWorks to start the Once Early-Reading Program, where we now provide software, training, and coaching to upskill any adult to teach any three- to seven-year-old child to read fluently and independently at a mid-second-grade level. Of course I wanted to hit product-market fit out of the gate. But I knew from my teaching days and from watching all of the vendors we supported at Clever that even though elementary school literacy is a widespread and deeply embedded problem—Sold a Story came out a year after we started Once—schools are often resistant to products or anything resembling a productized experience.
To make matters worse, most product developers over-invest in product development before nailing an implementation that works, making a lot of what Jeff Bezos called irreversible Type 1 decisions without even realizing it. So we were determined to build for the first couple of years with as little technology as possible in order to validate a solution that we could then productize and scale quickly to serve a massive need. We’re turning the corner on that now, but this has been five years in the making.
‘…we were determined to build for the first couple of years with as little technology as possible in order to validate a solution that we could then productize and scale quickly to serve a massive need.’
Once was born out of a family moment during the early pandemic, when you and your co-founder watched your spouses successfully teach your own kindergarteners to read at home in just 15 minutes a day. Now that Once is operating across public, charter, and private schools nationwide, what has been the hardest part of translating that “at-home parental magic” into structured, institutional environments where you are upskilling school support staff and volunteers?
To be clear, we continue to serve both audiences today. Thousands of children today receive Once at home, taught by parents or caregivers, and in school, taught by paraprofessionals, support staff, and high school CTE students. But I understand the thrust of your question is what happens when the magic relationship of a parent-child tutoring session is transformed into an in-school relationship.
I’ll say a few things. First, parent-child tutoring sessions aren’t always magical; sometimes children really do prefer to learn from an adult who is not their parent. But you’re absolutely right that typically no one is more motivated than a child’s own parent or caregiver. So to close that gap we do a few things.
First, we recognize that we are designing for extreme environments and users. Sometimes a school staff member will be hired to perform general responsibilities and then suddenly find themselves assigned to deliver Once. We need to get those folks as excited about this instruction as quickly as possible. That’s a challenge. But there are advantages to designing for these challenges, for extreme users. Breakthroughs we have on product design, training, coaching, etc., flow right through to our larger user base, helping us build a better solution that serves everyone.
So let’s talk about training and coaching. School staff members delivering Once and high school CTE students who can also deliver Once receive a day to a day and a half of training before they begin instruction. We’re big believers that a person can only hold so many new things in their head at one time, so rather than dozens of hours of upfront training, we teach tutors—we call them instructors—the most critical pieces up front and then rely on video observation and weekly, live coaching from our national network of expert Once coaches. All in-school Once instruction is recorded, since the instructor and the child are sitting side by side in front of an open laptop computer.
As important as instructional delivery is, equally important is ensuring that in-school Once instructors hold high expectations for every student and work as explicitly as possible to maximize every moment of instruction. The Once Program team shares a lot of responsibility for that work with the school sites where Once instruction is occurring. The Once Program team checks daily instructional logs, evaluates student progress, collaborates with schools to launch motivation and incentive programs, and basically keeps everything running.
‘The Once Program team checks daily instructional logs, evaluates student progress, collaborates with schools to launch motivation and incentive programs, and basically keeps everything running.’
Once in school is a team sport, and the learning we get from that environment we roll back into the product so that it is easier and easier and more and more effective to use at home as well.
You’ve pointed out that despite massive advances in the science of reading over the last 50 years, national reading proficiency has remained frustratingly stagnant. Why does the system struggle to close the gap between proven cognitive science and everyday classroom execution, and how does Once intentionally bypass that bottleneck?
Gosh, I mean, just listen to the Steubenville / Success for All story on Sold a Story’s first season, or the tragedy that Zig Engelmann endured in the largest education experiment ever conducted, Project Follow Through, which Emily Hanford is documenting in Sold a Story’s second season.
I think fundamentally, what bedevils education is that there is so little consensus on the desired outcomes. Are we trying to get children to read fluently, or comprehend, or love reading, or go to college, or make a lot of money, or be happy?
‘I think fundamentally, what bedevils education is that there is so little consensus on the desired outcomes. Are we trying to get children to read fluently, or comprehend, or love reading, or go to college, or make a lot of money, or be happy?’
And for each of those, does a teacher measure the result, or does a standardized test, or does a self-reported survey, or longitudinal data? Healthcare has grown to a bloated 18% of the U.S. economy, but there’s consensus around what outcomes matter. We want people to live longer and in less distress, until they die. That helps to align incentives.
Fortunately for us, due to the work of Emily Hanford and many, many others, there is much closer alignment on the progress we can expect from kindergarteners, first graders, and second graders on reading. And those expectations are aligned to evidence in cognitive science, with a heavy focus on decoding and comprehension. The core knowledge / schema piece really begins kicking in around third grade. By middle school, the experts are quite divided on what to do, and the outcomes are much worse.
So at Once, we’ve avoided the bottleneck by focusing on the earliest grades, where the science is more settled, and when students’ brains are most easily able to form the neural connections that allow them to decode text.
You are documenting an ambitious journey running randomized controlled trials across several of the nation’s largest school districts. In an edtech market sometimes criticized for making vague promises without clear verification, why are you betting so heavily on rigorous scientific validation, and what do you hope these results will prove to school boards and philanthropists alike?
Life is short. Too short to build an edtech company that gets to scale but doesn’t actually improve student achievement. The reason we go to work every day is because we believe that we’re doing something that matters, and that belief is rooted in evidence.
‘Life is short. Too short to build an edtech company that gets to scale but doesn’t actually improve student achievement. The reason we go to work every day is because we believe that we’re doing something that matters, and that belief is rooted in evidence.’
Now, I wasn’t anticipating this when I co-founded Once, but education evidence is very much a moving target. Not only do stakeholders not always agree on which assessments matter—see discussion above—but even once you have agreement on assessments, there is broad confusion over the difference between statistical significance and effect size, the difference between quasi-experimental studies and randomized controlled trials, the difference between intent to treat and treatment on the treated, how effect sizes translate to metrics that are more readily understandable, when we can extrapolate from small samples, who is responsible for validating research studies, what happens when a provider misstates their results, and the list goes on and on.
I’ve enjoyed digging into these disputes on LinkedIn, and I’m hopeful that as a field we dispel the confusion and get to a point where we can more easily compare interventions.
When we started this work at Once, a lot of people told us that districts don’t care about research data other than to check a box. We said, we don’t care, we’ll work with third parties to produce the research anyway. Now we’re seeing a lot more demands across the sector for evidence. From experience, we know that takes a while to build, so evidence could end up becoming quite a durable advantage for us.
Personally, I am not impressed by companies or organizations that produce a single glowing research report and then don’t commission another study for a decade. At Once, our product is improving constantly, so a research report based on a full year of school data documents student experience on features that might have been pushed two years earlier. We have to continue to run research and collect evidence so that we can see the impact of what we continue to build.
Additionally, longitudinal studies on interventions in education are scarce. It’s great that some intervention improved early elementary decoding, but what happened to those students later on? Could they comprehend? Did they continue to grow disproportionately to their peers? These are the types of questions we’ll be particularly interested in in the future.
With the strong monthly recurring revenue growth of your Once at Home program, you’ve hinted at a future where kids might regularly enter kindergarten already reading fluently. Looking five to ten years down the road, how do you see the relationship shifting between software, parents, and schools? Do you envision a world where structural literacy failure is entirely obsolete before a child even sits at a school desk?
Wow. That would be game, set, match. I imagine the day when a parent walks their three-year-old into the pediatrician’s office, and walks out with a Once subscription covered by health insurance, and two years later that kid enters kindergarten decoding, or at least well on their way.
Schools get to focus their interventionists’ time on the children who didn’t get that support before kindergarten, or who just arrived in the U.S., or who have specific needs for more specialized intervention. Kindergarten, first, and second grade are spent reading books, discussing them, making meaning, writing, creating. The entire Tier 1 curriculum would need to get fundamentally revised.
The very first place we piloted Once was in an elementary school in West Oakland. We were working in a TK [Transitional Kindergarten, a foundational bridge program between preschool and kindergarten for young learners] class. A month into our work, the principal invited us to the all-staff meeting he held each Friday. At the end of the meeting, he announced that the TK students were getting daily, one-on-one, in-person reading instruction so they would learn to read fluently and independently well above grade level.
Suddenly, a group of adults started clapping and cheering. We looked around. It wasn’t the TK or kindergarten teachers. It was the seventh- and eighth-grade teachers. They said, “If all those kids come into middle school reading fluently, we’ll be able to teach the grade-level curriculum we want to teach!”
‘Suddenly, a group of adults started clapping and cheering. We looked around. It wasn’t the TK or kindergarten teachers. It was the seventh- and eighth-grade teachers.’
That’s the dream. Right there.
Now we need to win adoption, school by school and district by district. Fortunately, that’s getting a lot easier. When we first started, we only upskilled paraprofessionals to provide Once instruction. That doesn’t offer districts a lot of options. Now, in addition to upskilling paraprofessionals, we can upskill high school CTE students, volunteers, or parents at home—schools can buy them Once at Home licenses in bulk.
Additionally, some districts have budgets for kindergarten or first grade. Others have budgets for pre-K. We can serve them both. Finally, the Education Freedom Tax Credit could bring billions of dollars to public school tutoring. I’m encouraged to see Scholarship Granting Organizations newly created to facilitate this, with a commitment to transparently reporting on the amount of tutoring delivered and the results achieved.
I’m an optimist, and all of this gives me a lot of cause for optimism.
—
Victor Rivero is Editor-in-Chief of EdTech Digest, Program Director of The EdTech Awards, and Executive Producer of Future Focus Forums (F3), which brings top U.S. school district technology leaders together with edtech companies to gain candid, actionable feedback on products and strategy while establishing deeper, more productive relationships. Write to: victor@edtechdigest.com.
0 Comments