Why "paying kids to revise" is the wrong description of Quizstack, what the research actually shows, and the words to use with a parent, a grandparent or a journalist who calls it bribery.
The objection is right about the wrong thing. Paying a teenager for an A doesn't work; the biggest trial ever run on it found no effect. Paying them for the work that produces the A does, and the effect lasts after the money stops. A Quizstack card pays for the work: a short quiz on this week's own notes, passed today. It never pays for the exam.
When you offer £50 for an A in August, you're asking a fifteen-year-old to plan nine months of revision, judge their own progress, and resist every evening's alternatives, all on the strength of a distant promise. Most adults can't do that with a gym membership. The economists who ran the trials put it plainly: students "have little clue how to turn their excitement about rewards into achievement". So the money sits there and nothing changes.
Offer the reward for a specific, near-term, checkable action instead (read this book and pass a short quiz on it) and behaviour changes, learning improves, and the improvement is still measurable a year after the payments stop. The action has to be small enough to do tonight and clear enough that there's no arguing about whether it happened.
The thing we pay for isn't arbitrary. Testing yourself on material beats re-reading it by a wide margin for anything you need to remember a week later, which is the whole of exam revision. So the sponsor isn't paying for time at a desk, or for a mood; they're paying for the single most effective revision activity there is, done to a pass mark, on the student's own notes.
Roland Fryer's randomised trials paid about 38,000 students in 250+ schools across four US cities. New York paid for test scores and Chicago paid $50 for an A, $35 for a B, $20 for a C: neither raised achievement. Dallas paid second-graders $2 a book, on condition they passed a short quiz on it at 80%: reading comprehension rose, and the treatment group still outperformed a year after the money stopped. The paper's conclusion is the one to quote: rewards for inputs work; incentives tied to output are not effective.
Levitt, List, Neckermann and Sadoff tested rewards in Chicago Heights schools. A reward paid immediately after the test lifted scores by 0.12 to 0.2 standard deviations, about five to six months of extra learning. The identical reward paid a month later did nothing. $20 worked where $10 was patchy. And rewards framed as a loss (here's the money, you keep it if you hit the mark) beat the same reward framed as a bonus. Older students responded only to money; younger ones responded to a $3 trophy.
The standard psychology objection is that paying for something erodes the motivation to do it for its own sake. A meta-analysis of the school incentive trials looked for exactly that and found nothing: no measurable change in intrinsic motivation, whether measured by questionnaire or by performance after the rewards were removed. The overall achievement effect across trials was positive but modest (0.08 SD), which is an argument for pairing the money with a good input, not for skipping it.
Roediger and Karpicke had students either re-read a passage or take a short recall test on it. Five minutes later the re-readers were slightly ahead (81% vs 75%). A week later the testers remembered 56% and the re-readers 42%. In a second experiment, one read plus three self-tests beat four reads at a week by 61% to 40%, even though the four-read group had seen the text more than four times as often. This is the "testing effect", and it's the most replicated finding in the study of learning.
Each finding maps to a design decision. The ones already in the brief are marked; two are new and should go into Rev 6.
The unit of reward is a passed quiz on the student's own notes: small, tonight, checkable. The sponsor can chain cards toward a mock, but every pound is attached to an input.
Each tier pays out as it's passed, so the reward is immediate. Parent #3 asked for exactly this, unprompted: "spread it out".
Dallas paid only on an 80% quiz pass. Ours is a 70% floor with a fresh paper each retake, so the input can't be faked.
We don't teach. Every minute a student spends with us is retrieval practice, the activity with the best evidence behind it.
Loss framing beat bonus framing in every age group. The student screen should say "Nan's £20 is in your stack, yours when you pass by Friday", with a visible countdown, not "you could earn £20".
A voucher a month later did nothing in the trials. The payout rail (Tillo/Runa) must issue the voucher the moment a card is passed, and the enquiry to them should say so.
Same argument, three audiences. Each one leads with the sponsor's own worry and answers it in one breath.
Paying for an A doesn't work; the biggest study ever run on it found no effect. Paying for the work does. Quizstack pays out for one thing: a short quiz on their own notes, passed today. Not the exam. Not the mood. The habit.
Bribery is money for a result you can't check, like "£50 if you get an A in August". We don't do that, and the research says it doesn't work anyway. A Quizstack pot pays for revision as it happens: a card is twelve to eighteen questions on their own notes, with a pass mark and fresh questions if they retry, and the voucher lands the moment they pass. It's the same reason you pay a tutor by the hour and not by the grade.
It's £20 that only reaches them when they've actually done the revision, a bit at a time, and you get told each time they pass. Think of it as paying for the hour, not the report card.
The evidence on this is clearer than the debate suggests. Harvard's Roland Fryer paid tens of thousands of American pupils for grades and test scores and found it made no difference; paying pupils a couple of dollars to read a book and pass a quiz on it raised their reading a year later. The lesson isn't "don't pay". It's pay for the work, pay straight away, and make it something they can do tonight. That's the whole design of Quizstack: a grandparent puts £20 behind a topic, the student earns it card by card on their own notes, and nobody is paid for an exam result.
Don't overclaim. The effects are real but modest: a tenth of a standard deviation across the trials, five or six months of learning in the best-designed one. The honest line is "it works, and the way we've built it is the way that works", not "guaranteed grades".
The trials were American, mostly primary and middle school, mostly low-income districts. Nobody has run one on UK GCSE students with grandparent money. Our ten conversations and the teacher test are the nearest thing we have, so say "the research" for the mechanism and "our early testing" for Quizstack itself.
Older students only responded to money in Chicago Heights, and $10 was patchy where $20 worked. That's consistent with parent #3's "£20 to try, not more" and with keeping the £5 pot as a grandparent-shaped ask rather than the headline.