🛝 The Product Wasn't the Proof

Welcome to Playground Post, a bi-weekly newsletter that keeps education innovators ahead of what's next.

This week's reality check: Varsity Tutors for Schools served as many as 1,000 schools and districts, but only 15% of students received the recommended dosage in preliminary, unfinished research. WestEd found positive academic effects in three of four outcomes-based interventions it could rigorously evaluate. And while 73% of faculty say they have handled an AI-related integrity problem.

💎 Data Gem

New York City received nearly 5,700 applications for just 2,000 seats in its first public preschool program for 2-year-olds. Each seat will save a family an average of $26,000 per year.

The Tutoring Giant Shut Down. Only 15% Got the Dosage

For a district, access to a tutor is not the same as tutoring.

Varsity Tutors for Schools promised to connect students with online help day or night. One company site said the service reached more than 1,000 schools and districts.

Then the company ended school services immediately.

The shutdown came after districts spent billions in federal recovery money trying to reverse pandemic learning loss. The American Rescue Plan alone sent schools $122 billion, and district purchases from Varsity peaked above $28 million in 2023, according to a new report.

But the evidence never caught up with the scale.

Research supports high-impact tutoring when students meet at least three times a week, work one-on-one or in small groups, and keep the same tutor over several weeks.

Accelerate and Mathematica were studying Varsity's work across three districts during the 2025–26 school year. The research is unfinished, and the early finding must be treated as preliminary.

Only 15% of students received the recommended dosage.

Varsity posted a separate case study in which fourth graders gained 9.82 points on i-Ready compared with 4.52 points for a control group. But the summary did not name the researchers or include enough methodological detail for outside experts to judge its rigor.

Bernalillo Public Schools in New Mexico stopped using the service after one year. The superintendent said Varsity could not provide promised support in a timely fashion.

For education innovators, the opening is a tutoring operations layer that verifies attendance, frequency, tutor continuity, and replacement coverage before a district pays for access. Independent efficacy audits can give buyers evidence they can compare before the next provider disappears.

Three of Four Programs Worked

Outcomes-based contracting promises a clean correction to education procurement: stop paying vendors only for delivering a product and tie part of the payment to student results.

The model has spread to more than 20 states.

Michigan now requires it for schools seeking access to a $50 million virtual-tutoring fund.

The first independent evaluation offers real reasons for optimism.

WestEd studied 10 tutoring and education-technology interventions across eight districts in California, Florida, Mississippi, and Texas between August 2024 and March 2026, according to The Hechinger Report.

Only four interventions produced clean enough comparisons for rigorous evaluation.

Three of those four improved academic outcomes.

The fourth was an online tutoring program intended to run most of the school year. It lasted about two months, too little time to expect measurable gains.

The other six interventions exposed the operating problem. Schools filled empty tutoring seats with students outside the intended eligibility group, failed to get eligible students to after-school sessions, or changed access rules until researchers no longer had a valid comparison group.

"We have no idea whether or not they were working for kids, or even if kids got the services," said Brittany Miller of the Center for Outcomes Based Contracting.

Districts and vendors did begin meeting regularly, tracking attendance, and adjusting programs as data arrived.

But every participating district also received roughly $80,000 in coaching and technical assistance to design the contract, define outcomes, monitor progress, and solve implementation problems.

The evaluation tested outcomes-based contracting plus intensive support.

It cannot determine which one produced the change. It also found no evidence that the model saved schools money.

The center is now testing support that costs under $10,000 per district.

For education innovators, the market is contract infrastructure that connects student eligibility, attendance, vendor payments, and outcome measures in one shared record. A lower-cost implementation service can help districts run the model without spending $80,000 to learn whether the contract works.

73% Faced AI Cheating. Some Turned Off the Detector

AI made cheating easier to scale.

It did not make cheating easier to prove.

73% of faculty say they have personally handled an AI-related academic-integrity problem, according to a new report.

But studies have found AI detectors produce inconsistent results, falsely flag human writing, and disproportionately misidentify work by non-native English writers.

Some accused students have sued their universities.

Yale, Vanderbilt, Johns Hopkins, and Indiana now ban or discourage using detector output as the sole evidence of cheating.

At least 12 universities, including Northwestern, Georgetown, and New York University, have disabled Turnitin's AI detector.

The replacement is not another detector.

Indiana University's Kelley School of Business now tells faculty to design assignments around process, reasoning, and authentic engagement.

Other recommendations include blue-book exams, in-person oral assessments, and requiring students to document how they used AI while completing the work.

Kevin Yee of the University of Central Florida estimates that about half of faculty want to make as few changes as possible. That may work in face-to-face courses with fewer than 30 students, where relationships make shortcuts harder.

Larger classes create the real product gap.

Oral exams require proctoring and substantial faculty time.

Assessment redesign across an entire institution requires support that individual instructors cannot supply alone.

"We're still relying on the unsupervised written word as evidence of learning. The real trick is acknowledging that that doesn't work anymore," said Tricia Bertram Gallant, UC San Diego's academic-integrity director.

For education innovators, the opportunity is scalable oral-assessment workflow, process-evidence portfolios that capture drafts and reasoning, and faculty redesign systems that turn existing assignments into verifiable learning tasks. The market is moving from detecting generated text to documenting how learning happened.

⚡️More Quick Hits

This week in education:

• 94% of Division I athletic departments lost money — Programs spent $20.8 billion against $13.1 billion in revenue, while the median undergraduate contribution reached $8,500 over four years.

• Subscription cost is a deal-breaker for 63% of school buyers — In a survey of 2,277 decision-makers, 80% named budgets as their biggest challenge and 37% rejected products that would not integrate with existing systems.

• Canvas attackers claimed 3.5 terabytes of data from more than 8,800 institutions — The May breach disrupted classes across dozens of districts and exposed how dependent schools are on a single learning-management system.

• HBCU coaching produced a 6.9% return rate, 2.5 times the national average — Across 39 HBCUs, more than 1,300 stopped-out students returned and institutions recovered or retained an estimated $26.9 million in tuition revenue.

To stay up-to-date on all things education innovation, visit us at playgroundpost.com.