A Pre-Registered Analysis: Our Prime Day Predictions, Filed in Advance

The Premise

Prime Day arrives in roughly two weeks — Amazon has signaled mid-July, per its unbroken pattern — and the retail press will spend the event's aftermath explaining, with great confidence, what was always going to happen. Anyone can predict the past. The Editorial Staff prefers the other direction: we are filing our predictions now, in writing, where they can be graded and, if necessary, held against us.

This is pre-registration, a practice borrowed from clinical research, where hypotheses are filed before the data arrives so that researchers cannot quietly discover they had predicted whatever occurred. The stakes there are higher and the discounts fewer. The honesty transfers.

It is also July 3. Tomorrow is a federal holiday, and the sales attached to it are a preamble; we predict, as a warm-up exercise, that this weekend's "Fourth of July doorbusters" will be last week's prices wearing bunting. We will check on Monday.

Ground Rules

A real deal, here as always, is a price at least 10% below the product's 90-day average, per our Keepa-derived logs. One methodological note, filed in advance: a multi-day event contaminates its own baseline — by day three, the average has begun absorbing the very discount it is meant to measure. We therefore freeze all baselines at 12:01 AM on the event's first day and grade every claim against the frozen number. Predictions are graded pass or fail. There is no partial credit. There never is.

For calibration: our Memorial Day forecast, filed May 19, graded five for six. The miss was the AirPods Pro 2, which we predicted near $169 and which instead rose to a $189 street price as clearance stock ran out. We had not modeled scarcity. We have now.

The Predictions

P1 — Base iPad (11-inch A16, MSRP $349) at $299 or below: 80%. The most reliable event participant we track. It has cleared this bar at every major sale event in our logs since its release, including two this year.

P2 — Apple Watch Series 11 (MSRP $399) at 12% or more below its frozen average: 70%. September's successor announcement is close enough that inventory arithmetic favors the buyer.

P3 — AirPods Pro 3 (MSRP $249) at $199 or below, its first sub-$200 listing: 60%. The Pro 2 crossed the same line at its own first Prime Day. Precedent is the only physics retail respects, though new products are occasionally granted a second summer of exemption.

P4 — AirPods Pro 2 at 10% or more below its frozen average, one final time: 55%. In March we would have said 85%. The clearance pile has been shrinking since April, and scarcity has opinions. If it happens, it will be the last one, and someone should mark the occasion. It will apparently be us.

P5 — MacBook Air M4 13-inch (MSRP $999) listed at $799: 70%. And a sub-prediction we file with the calipers already out: against a frozen average we project near $886, a $799 listing sits 9.8% below — missing our own threshold by approximately $1.60. We predict, in other words, a discount the entire internet will celebrate and our methodology will decline, by the width of a sandwich. We will report the margin to the cent.

P6 — iPad Pro M5 (MSRP $999 and $1,299): no listing more than 5% below its frozen average: 85%. Theater. The Pro line does not attend sales; it makes appearances.

P7 — MacBook Pro M5 14-inch (MSRP $1,999): nothing beyond the standing $150 excursion, failing the threshold again: 80%.

P8 — AirPods Max (MSRP $549) at $449, presented as a Prime Day exclusive: 90%. It was $449 on 61 of the 166 days of the first half, by our count. The banner will be new. The number will not.

P9 — At least one accessory "70% off" doorbuster measured against a list price the product has never actually sold at: 95%. List prices are the fireworks of retail: loud, brief, and pointed away from the ground.

What We Decline to Predict

The exact dates: Amazon announces when it announces. iPhone pricing: carrier promotions are not prices but theater with a 36-month runtime, and we do not grade theater. Television doorbusters by model: event-week televisions materialize bearing derivative model numbers that exist nowhere else and vanish by August, and comparing one to a 90-day average is comparing a ghost to its own absence. We track what persists.

The Grading Protocol

Within a week of the event's close, we will publish the grade sheet: each prediction, pass or fail, misses listed first, margins to the cent, frozen baselines shown. Failure, listed first, is the only honest sort order. Throughout the event, every claimed discount on our deals page will be verified against its 90-day average, exactly as on every other day, because the method does not take holidays either.

The predictions are filed. The event will occur. A forecast that cannot embarrass you is not a forecast; it is a banner, and we have counted enough of those.

— The BuyGetRewards Editorial Staff

Shop Current Deals

We track live prices against each product's 90-day Amazon average, so you can tell a real discount from a banner. Browse today's deals →