Proof Before Purchase

A note on method

This essay is the fourth in a series written from the builder's chair. The first argued that the education market cannot reward what works because the evidence of what works is invisible where money moves.¹ The second took one state's checkbook and showed sixteen years of spending with no record of a result.² The third read the fine print on contract vehicles and found that being buyable is not the same as being bought.³

This one looks at the only place in the system where anybody is legally required to prove anything before a dollar moves. I went through all fifty-one jurisdictions, states plus the District of Columbia, and read what each one requires of an instructional product before public money can buy it.⁴ The question was simple. Where does a state make proof a precondition of purchase, and what counts as proof?

I. Twenty-two lists, four gates

Start with what is common. Twenty-two states publish an approved list of instructional materials. A review panel reads submissions against criteria, and the ones that pass go on the list. This is old practice, older than any technology in the classroom, and in most of those states the list is exactly what it sounds like: a recommendation. A district can consult it, ignore it, and buy something else with local money.

Now the narrower thing. In four of those states, the list has teeth. Appearing on it is a condition of getting the money, or of being allowed to use the product at all.

Four of fifty-one jurisdictions. The other forty-seven either publish a list without attaching consequences to it, or publish nothing at all.

I want to say plainly that these four states did something good and difficult. Every one of these laws was contested. Somebody had to stand up and say that public money should not buy a reading program whose method the research does not support, and then survive the argument that followed. That is real work, and the four states that did it are ahead of the other forty-seven, not behind them.

II. What the gate reads

Here is where it gets interesting, and where I think the conversation has stopped one question too early.

Read those four requirements again and notice what all of them have in common. Colorado asks whether a program is evidence-based. Ohio asks whether materials are high quality and aligned to the science of reading. Texas runs materials through a review against state criteria. Virginia asks whether a program appears on an approved list built the same way. In every case, the thing being examined is the product. Its design. Its alignment. The body of research behind the method it uses.

Not one of them examines the record. No state asks the question a business would ask first: this program has already been sold to districts in this state, so what happened in those districts? Did the children read better? Nobody checks, because nobody is required to, and because in most states the record that would answer it has never been assembled.

The distinction matters more than it sounds. "Evidence-based" in these statutes means the method has research behind it in general. It does not mean there is evidence of what this product did in your state, in districts like yours, over the years it has already been paid for. A program can be evidence-based in the statutory sense and still have a decade of disappointing local results, and nothing in any of these four laws would surface that. The gate is at the front door. It reads the résumé, not the reference.

The state that tried

One state came closer than the four, and what happened to it is worth the detour. In the 2026 session Utah introduced a bill on software in schools that would have done the thing none of the four do. As introduced, it required vendors to obtain independent verification that their software was academically effective before it could be used in Utah classrooms, and directed the state board to verify effectiveness itself.⁵

Those provisions did not survive. The bill was signed in March 2026 in a form that asks the state board to conduct a study of software use in schools and report what it finds. A study is not nothing, and Utah also enacted the strongest grade-band limit on classroom screen time in the country in the same session. But the mandate that would have made proof of effectiveness a condition of use was removed before passage, and as of July 2026 no state has an enacted requirement that an education technology product demonstrate it works before schools may buy it.

I am not telling that story to score a point against Utah. I am telling it because it shows where the resistance actually sits. Requiring proof of design is now politically achievable, and four states have done it. Requiring proof of results is a different fight, and the first serious attempt at it was amended away.

III. What the missing gate costs

The second essay in this series read sixteen years of one state's spending on the companies that supply its curriculum, its reading and math intervention programs, and its tests. Maryland is not one of the four. More than a billion dollars went to those companies, and over the same years, measured on an independent national scale, the state's districts lost ground rather than gained it.⁶

I have been careful throughout this series about what that does and does not prove, and I will be careful again. It does not prove that any product failed. Districts buy many things at once, students arrive with different circumstances, and a correlation across one state is not a verdict on a company. What it proves is narrower and, I think, harder to answer: for sixteen years nobody could have told you whether the money worked, because the two halves of the question lived in different filing cabinets.

A front-door gate would not have changed that. Suppose Maryland had adopted Ohio's law in 2010. Every program purchased would have been on an approved list, reviewed against reasonable criteria, and the outcome record would look exactly the same as it does now, because nothing in an approval process ever revisits the purchase. Sixteen years of renewals would still have run on inertia. The list tells you the product was worth trying. It never tells you it worked.

That is the gap, and it is not a small one. Everything in the market's current architecture faces forward: the pitch, the review, the approval, the purchase order. Nothing faces backward. And the money is spent in the direction nobody is looking.

IV. The gate at the back door

So what would the other gate look like? Less than you might think, because the pieces mostly exist already.

States already publish what they pay. A handful publish it well enough to reconstruct a district's vendor history for a decade and a half, which is how the second essay in this series was written and how The Ledger now gets published every quarter. States already publish how students perform, and independent researchers already put those results on a common scale so that a district in one county can be compared to a district in another and to the country.⁷ The two records exist. They have simply never been required to sit on the same page.

A back-door gate would say something like this: before a district renews a program that public money has already bought, the record of that program's spending and the record of the outcomes it was bought to improve are placed side by side, in public, on a scale nobody in the transaction controls. Not a verdict. Not a ban. Just the record, in front of the person signing, at the moment of signing.

Notice what that does to the incentives. A company whose product works has, for the first time, a public reason to want the record kept, because the record is its best argument. A company whose product does not work loses the thing it currently relies on, which is not deception but simple invisibility. And a superintendent gets the one thing no approved list has ever given her: an answer to the question of whether this worked here.

The four states that built the front gate proved something important, which is that the politics of requiring proof are survivable. Somebody argued for those laws and won. The next argument is smaller than the one they already had, because it does not require judging any company. It requires only writing down what was already spent next to what was already achieved, and letting everyone read the same page at the same time.

Four states ask for proof before the money moves. None asks what the money did. That second question is the cheaper one to answer, and it is the only one that ever tells you the truth.