Companion to the abstractions PDF · 15 sources · verified August 2026
Abstractions in the Wild
Published articles, earnings calls, and analyst notes whose everyday English quietly runs the reasoning of your three structural models — without ever stating the heuristics they were built to justify.
How to read this. Each chapter opens with the model from your PDF in plain language, then lists five verified sources. For every source: a short account of what the text argues (paraphrased, not quoted), then translation pairs — a statement the text makes, followed in red by what that statement means inside the model. Every source was fetched and checked; links go to free-to-read versions.
Abstraction 1
Price, volume & promises
Reported revenue can move for three very different reasons: prices changed, the company actually delivered more stuff, or the company made a counting choice — like booking orders it hasn't delivered yet, or deciding whether money spent on unsold goods counts as a cost now or sits on the books as an asset. Any claim about revenue or costs is secretly a claim about which of these moved.
The heuristic these texts don't state: revenue rising while costs stay flat suggests the books are being cooked.
Revenue = P × (S + α·A) · Cost = E × (S + γ·I)
P
price per unit
S
units actually delivered
A
business agreed to but not yet delivered (orders, backlog)
α
how much of that promised business is counted as revenue today
E
cost to make one unit (efficiency)
I
additions to inventory
γ
how much inventory spending is counted as a cost today vs. held as an asset
P&G's reported sales slipped 1% even though underlying growth was +5%: price increases of roughly 10% ran against a 6% drop in units sold. The article then dissects the volume drop itself — about half genuine softening demand, the rest from exiting Russia and from Chinese retailers running down their own stockrooms — and reports the CFO's plan to keep raising prices to cover costlier raw materials.
Translation
Sales grew only because of ~10% price hikes while unit volumes fell 6%.
A quarter's growth carried almost entirely by pricing: prices +10% for a second straight quarter, volumes −3% overall and −7% in Europe. Gross margin (the profit left after direct production costs) improved 1.5 points because price rises outran cost inflation. The warning: European shoppers are switching to cheaper store brands as the price gap widens, so more hikes risk accelerating volume losses.
Translation
All the growth came from price; volumes went the other way.
ΔRevenue > 0 carried entirely by the P term while ΔS < 0
Margins widened because price increases outpaced input-cost inflation.
per-unit margin (P − E) grew because ΔP > ΔE — attributing margin to pricing vs. efficiency
Shoppers trade down as the price gap widens, so further hikes would cost more volume.
S depends on P (elasticity): pushing P up drives S down, capping how long price alone can carry P·S
PepsiCo's reported sales fell 0.5% — its first decline in nearly four years — yet underlying revenue still rose 4.5%, purely on higher prices, while volumes fell across segments (snacks −2%, North American drinks −6%, Quaker −8%). The article's causal story: the price hikes themselves, plus stretched household budgets, are what weakened demand. Confidence note: the piece is real and free to read, but its decomposition is coarser than our first pass suggested — CNBC's July 2023 PepsiCo story carries the more vivid per-region price/volume pairs.
Translation
Revenue rose on pricing while every segment sold fewer units.
same P-vs-S split: ΔP > 0, ΔS < 0 segment by segment
The raised prices are themselves hurting demand, and the outlook assumes cautious consumers.
once ΔP → 0, ΔRevenue inherits the sign of ΔS — the article is forecasting the P lever running out
ASML's quarterly orders — signed commitments for chip-making machines not yet delivered, which the industry calls bookings — hit €13.2B against €6.3B expected. The article calls bookings the industry's most-watched number, and pairs the record order intake with the company raising its sales outlook for the following year. Nothing about accounting games: orders are treated as credible future business.
Translation
Orders are the headline metric, tracked separately from this quarter's sales.
the order book is the A term — agreed but undelivered — measured apart from recognized revenue P·S
Record orders now justify a higher revenue forecast for next year.
additions to A convert into future revenue as machines ship — future P·S read off today's A, not today's S
The surge reflects chipmakers expanding capacity for AI chips.
growth in A signals future real volume S — a quantity story, not a price story
An analyst divides ASML's €38.8B backlog (the accumulated pile of signed-but-unfilled orders) by next year's €34–39B revenue guidance and concludes next year is already fully covered by existing orders. Since a machine takes 16–20 months from order to delivery, today's orders reveal what customers expect to need in 2027–28. The investment question therefore flips: demand is settled; the risk is whether ASML can build fast enough.
Translation
Backlog covers more than 100% of next year's revenue guidance.
comparing the stock of A against next period's P·S: future revenue pre-committed inside A
Long lead times make today's orders a two-year-forward demand signal.
the 16–20 month schedule is the timetable on which units migrate from A into delivered units S
The binding risk is execution, not demand.
the constraint on future revenue is the A→S conversion rate, not new inflow into A
Abstraction 2
Fixed costs & the lever
Split a company's costs into fixed (paid no matter what: factories, fleets, leases) and variable (paid per unit sold). Because the fixed part doesn't move with sales, profit swings more than proportionally when sales change — in both directions. And buying more physical capital is a trade: higher fixed costs in exchange for a cheaper cost per unit, with today's reported profit depressed while the capacity is being built.
The heuristic these texts don't state: a higher share of fixed costs means the company does badly when sales fall.
Earnings = S × (P − V(K)) − F(K) · break-even: S × (P − V) = F
S
units sold
P
price per unit
K
physical capital — factories, planes, warehouses
V(K)
cost per unit; falls as K rises (machines make each unit cheaper)
A hotel owner emerging from COVID reports revenue per room rebounding 372% off the trough — and 56 cents of every extra revenue dollar dropping straight through to operating profit. Management credits cost discipline: cutting hard in the crash, then preventing costs from creeping back as guests returned, down to merging a hotel's restaurants to cut staffing. Two consecutive quarters of positive property-level profit mark the turn.
Translation
56% of each incremental revenue dollar became profit.
measured incremental margin (P − V)/P ≈ 0.56: with F already being paid at empty-hotel occupancy, each extra room-night adds ≈ P − V to earnings
The job on the way up is stopping cost creep.
holding F and V flat while S recovers, so ΔEarnings ≈ ΔS·(P − V) isn't eroded
Property profit turned positive as rooms refilled.
S recrossing the break-even condition S·(P − V) = F from below
Delta's cost per seat-mile is running 6–8% above 2019 — not because anything got less efficient, but because the airline is flying a ~20% smaller network while already carrying the cost of a full restart. The CFO's fix is volume, not cuts: restore the flying and the in-place cost base spreads over more seats. In CFO Dan Janki's words on the call: "You bring back the capacity, you get that leverage." Separately, buying newer planes is framed as paying money up front for permanently cheaper operations (~$400M of benefit in 2021, ~$650M in 2022).
Translation
Unit costs are high because the network is 20% smaller, not because costs grew.
unit cost = V + F/S — shrink S and the F/S term inflates with no change in the cost base
Restoring capacity is what brings unit costs down and profit back.
as S recovers with F fixed, Earnings = S·(P − V) − F rises more than proportionally
Fleet renewal is a permanent structural cost improvement.
spending on K to lower per-unit cost V(K) — the capital-for-efficiency trade
The Motley Fool transcript·Jul 2021·earnings callverified
Asked how to model profit margins, TI's CFO tells analysts to assume 70–75% of every incremental revenue dollar becomes profit over the long run — and not to extrapolate any single quarter, since one-offs (a Texas winter storm cost ~$50M that quarter) distort the level without changing the slope. The same call discusses building more company-owned chip factories, framed as a lasting cost advantage.
Translation
Model us as 70–75% of incremental revenue falling through to profit.
The CFO explains a margin increase as factories running fuller — the same fixed factory cost spread over more chips — and forecasts that newly opened overseas factories will drag margins down 2–4 points until they reach scale. The stated philosophy: higher capital spending always precedes higher growth, and the aim is for revenue to grow faster than that spending; the CEO adds that TSMC's core job is building capacity two to three years ahead of demand.
Translation
Margins rose mainly on higher factory utilization.
higher S spreads the fixed F(K) over more wafers — earnings moving more than proportionally with S
New factories dilute margins until their volume ramps.
F(K) jumps the moment K is added, before its S arrives — capacity investment depressing current earnings
Revenue growth should outpace capital-spending growth.
the condition for E = S·(P − V(K)) − F(K) to expand even as F(K) rises
Amazon's near-zero reported profit is a choice, not a symptom: the business throws off billions in operating cash, and the company reinvests essentially all of it in warehouses and cloud infrastructure, deliberately steering consolidated profit toward zero while building capacity for a much larger future. Warehouses placed near customers cut the cost of shipping each package. The essay says nothing about recessions — the conclusion is about capturing future market share.
Translation
Cash generation is strong even though profit prints near zero.
the spread S·(P − V) is large; reported E ≈ 0 only because spending is scaled to absorb it
All the cash goes into fulfillment and cloud capacity.
deliberately rising K, whose F(K) (depreciation, infrastructure) soaks up today's earnings while raising the ceiling on future S
Warehouses near customers make each delivery cheaper.
capital lowering variable cost: V falls as K rises
Abstraction 3
The inventory ledger
Inventory obeys bookkeeping arithmetic: what you end with = what you started with + what you made or bought − what you sold. The catch is that production and purchasing are decided in advance, from a forecast — and holding stock costs money. So any inventory jump decomposes into exactly two stories: the forecast missed, or the pile-up was deliberate (a buffer against tariffs, shortages, or future demand). Texts below run the identity to tell the second story, or to work out true demand along a supply chain.
The heuristic these texts don't state: inventory rising fast while sales stay flat suggests an involuntary pile-up from an unforeseen slowdown.
I_end = I_start + Q − S · Q chosen early = forecast + planned buffer
I
inventory on hand
Q
quantity produced or purchased (committed before demand is known)
S
quantity actually sold
buffer
stock deliberately held for later periods — costly, so it needs a reason
An audit of the assumption that firms stockpiled before the 2025 tariffs. Imports surged — but measured inventories barely rose and the ratio of inventory to sales actually fell, because consumers were buying ahead at the same time: goods flowed out the door as fast as they came in. The genuine excess works out to roughly three weeks of extra supply, meaning only one to two months of shelter before tariff costs bite.
Translation
Imports spiked, yet stock on shelves stayed flat — so where did the goods go?
the identity forced explicitly: Q up but I_end flat ⟹ S must be up too
Consumers were pre-buying, so warehouses stayed lean.
inflow and outflow rose together; the stock barely moved
The buffer buys one to two months before tariffs hit costs.
coverage time = excess stock ÷ the sales run-rate; the build that exists sits in the deliberate branch
UK Office for National Statistics·Feb 2021·statistical analysisverified
The UK's statistics agency shows imports from the EU spiking in the month before each Brexit cliff-edge date (£1.7B, £0.8B, £1.7B), concentrated in machinery and production inputs, and corroborates it with surveys: 18.6% of manufacturers reported stockpiling versus a 6.4% economy-wide average. Once the trade deal landed, reported stockpiling collapsed. One honest caveat: the late-2020 build is hard to split between Brexit preparation and pandemic buffering.
Translation
Purchases spike right before a known deadline while demand sits still.
timing identifies the branch: a Q surge with S unchanged can't be a forecast miss — the motive is observable
Stockpiling reports collapsed once the deal removed the risk.
the planned buffer unwinding: with its reason gone and holding costs still positive, set Q below S to run I back down
Brexit and COVID motives are hard to disentangle.
the attribution problem inside the decomposition — a stock change doesn't label its own cause
Costco's CFO states the plan in the first person: pull inventory purchases forward before threatened tariffs land — a tactic already used against long shipping times and port-strike risk — alongside vendor negotiations and switching what gets sourced where. He also bounds the exposure: only some of the roughly one quarter of sales that are non-food is imported at all.
Translation
We'll buy ahead of the tariffs on purpose.
the ex-ante Q decision itself: set Q above the sales forecast because pre-tariff units are cheaper than replacements — a planned build by construction
We did the same against shipping delays and strike risk.
the buffer rationale: extra stock held against supply disruption, not a demand bet
Only a slice of the business is import-exposed.
bounding what fraction of Q warrants the pre-buy treatment
A semiconductor analyst frames the 2022–23 chip downturn as a supply-chain inventory correction: chipmakers deliberately shipping less to distributors than end customers are consuming, to drain overstuffed channels — the mirror image of the shortage, when every layer over-ordered at once. Tracking days-of-inventory across a dozen chipmakers, he reads the builds decelerating as the classic signal a correction is ending, and calls the bottom for Q1 2023.
Translation
Chipmakers are shipping below what end customers consume, to drain the channel.
the identity at the distributor tier: channel ΔI = sell-in − sell-through < 0, chosen deliberately — so reported sales understate true demand S
During the shortage every layer over-ordered simultaneously.
each tier commits its Q on the tier below's forecast — the over-ordering cascade (the bullwhip)
Inventory builds decelerating is the tell that the correction is nearly over.
the flow gap (sell-in − sell-through) shrinking toward zero: reading the identity's re-balancing as cycle timing
The Motley Fool transcript·Aug 2026·earnings callverified
Goodyear's CEO splits an earlier decline in tire volumes into three roughly equal parts: deliberately dropping low-margin product lines, distributors working off inventory they'd over-accumulated, and competitive pressure. Management then argues the middle piece is over: shipments into distributors are back in line with what distributors sell onward to customers — so Goodyear's reported sales once again track real end demand instead of understating it.
Translation
Shipments in are back in line with sales out.
distributor-tier identity with the ΔI term ≈ 0: sell-in = sell-through ⟹ our reported sales = end demand
Distributors entered the year with heavy stock to work through.
I_start above target, so distributors set purchases below their own sales to burn it down — the deliberate branch, run in reverse
The volume decline splits into product exits, destocking, and competition.
partitioning reported volume moves into a channel inventory-adjustment component vs. real-demand components
Across the fifteen
What the collection shows
The decomposition is often handed over, not inferred. Your PDF's analyst has to guess which variable moved and lean on priors. In the wild, companies frequently disclose the split themselves — organic growth stated as price vs. volume, orders reported separately from revenue, margin walks itemized — so journalists spend their words interpreting a decomposition they were given.
Real texts resolve the either/or with outside evidence. Where your abstractions end with a prior over branches (forecast miss vs. planned buffer; efficiency vs. book-cooking), these texts identify the active branch directly: an exogenous deadline (ONS), a first-person statement of intent (Costco), or a measured flow gap (Goodyear, TKer). That's the main structural difference between the heuristic user and the reporter — not different models, different evidence.
Same machinery, opposite headlines. The fixed-cost structure your heuristic reads as downturn fragility is celebrated in five of these texts as the engine of recovery profits and future dominance. Nothing in the model changed — only the sign of ΔS. That's strong evidence the abstraction, not the heuristic, is the shared object.