Actionable insights — The Oil Supply Shock Hiding in Plain Sight

3:54 1. Split reported inventory into working capital and genuinely drawable stock

1. Never net a supply loss against headline inventory. Ask first how much of that inventory is operationally required: oil filling pipelines, oil in transit on vessels (20–30 days each way), tank heels below the suction point. That volume is working capital, not savings — "yes, it's cash. It might sit in the account, but you need it to fulfil payroll."
2. Compute the deficit as (barrels/day lost) × (days lost) and compare it only to the drawable remainder. If the deficit exceeds it, the market must clear the difference through price-driven demand destruction, because there is nothing left to release.
3. Sanity-check the mirror case for what "running out" looks like: a just-in-time market with no buffer prices in both directions violently — tank tops gave −$50/bbl in 2020, so tank bottoms are the symmetric risk.

• Days-of-supply cover that keeps falling while headline stock levels look adequate; Cushing "dragging the bottom"; any published inventory series that doesn't separate pipeline/floating volumes from usable stock.

13:09 2. Lag-account the shut-in — count the months that have actually felt it

1. Map the physical pipeline of a supply loss: onshore tanks drain first, then loaded vessels, then the fields shut in — and only after shipping transit (20–30 days) does the barrel go missing at the destination. Assume 45–60 days before inventories start falling.
2. Subtract that lag from the data window. Five months of published data with a two-month lag is three months of impact — so an apparently calm balance is a partial reading, not a verdict.
3. Apply the lag symmetrically to the recovery: a reopening tomorrow still means the same two months of continued draws afterwards. "We're only just more than halfway through the physical manifestation of this problem."
4. Track cumulative draws against the deficit estimate to see how much of the bill is still outstanding.

• The date the outage began versus the date inventories first turned down — that difference is your lag constant. Preliminary final-month data (the July print) that will be revised.

15:38 3. Separate a one-time stock release from a change in the underlying flow

1. When a blockage briefly clears, expect a "burp": the queued inventory dumps into the market in days, not months. Size it (vessels × cargo, or pre-crisis throughput × days open) and treat it as a one-off, not a supply recovery.
2. Ask whether the underlying flow ever restarted. If wellhead production stayed shut, the price relief is borrowed against the same deficit.
3. Check where the released barrels landed. A burp into a market whose refineries are down does not become fuel — it just relocates the shortage downstream.

• Tanker-tracking spikes in departures without a corresponding rise in field-level output; floating storage falling while onshore stocks don't build.

19:50 4. Model crude and refined products as two markets — and audit the refining layer between them

1. Recognize the convention that hides the problem: "supply" is wellhead crude and NGLs, "demand" is refined-product consumption, and refining in the middle is netted out "like an intercompany transfer."
2. The netting is only valid while refinery runs track end-use demand. If runs fall for a non-demand reason (war damage, policy, embargo drill), the two markets decouple and every single-balance model breaks.
3. Then re-ask the question separately for each side: is crude short, or are products short? A products shortage will eventually pull crude with it, because the tanks have to be refilled by bidding crude back into the refining system.
4. Weight your confidence by data quality: crude inventories are reasonably instrumented; refined-product inventories, especially in emerging markets, are "the blind spot."

• Refinery utilization diverging from consumption proxies; product inventories (gasoline, distillate) 10–15M barrels below normal while crude stocks look fine; regional export bans.

48:11 5. Use the crack spread as a falsification test for "demand destruction"

1. State the prediction each story makes. If the balance is clearing through demand, crude and product prices should both be soft and the spread roughly stable. If it is clearing through a product shortage, cracks blow out while crude lags.
2. Read the actual spread. A 3-2-1 crack at record highs with a diesel crack at ~$102 against a normal $20–30 refining margin falsifies the demand-destruction story outright.
3. Convert the crack into an implied crude price to see what the fuel market is really saying: $85 crude + $70 of excess crack = diesel priced as if oil were $155.
4. Benchmark against the historical price of genuine demand destruction before assuming it has already happened — real rationing episodes (1981, 2008) needed close to $200/bbl in today's dollars.
5. Decide which leg to own. If the shortage began upstream, the spread most likely closes by crude rising, not products falling — so own the crude-linked asset, not the margin.

• The 3-2-1 and diesel cracks versus their normal $20–30 band; the implied-crude price the fuel market is quoting; whether the gap closes from the crude side or the product side.

46:42 6. Find the modelling artifact before believing a demand collapse

1. When a headline reports an implausibly large demand drop, ask how demand is estimated. In oil, refinery runs are a major model input — so a war-driven run cut is reported as a demand cut. "That's the tail wagging the dog."
2. Test the number against its own historical analogues: is the claimed decline larger than 2008 and comparable to a COVID lockdown month? If so, it should be visible everywhere in daily life.
3. Corroborate with at least one independent, high-frequency physical series that does not share the input (see the flight-tracker method below), plus direct observation — "the airports are absolutely packed."
4. If the residual doesn't survive the cross-check, reassign it: the barrels aren't un-consumed, they're coming out of uncounted inventory.

• Balancing items and "missing barrels" in the agency data; a demand series that moves in lockstep with refinery utilization rather than with end-use activity.

20:38 7. Keep one real-time, non-agency demand proxy — commercial flight traffic

1. Pick a physical activity series published daily by someone with no stake in the balance — commercial flights in the air, available free from the consumer flight-tracking sites.
2. Use it directionally, not proportionally: 5–6% traffic growth is not 5% oil-demand growth (that would be 5 mb/d), but "the two series do line up incredibly well."
3. Trust it because it survived the hardest test: the correlation held through COVID and the '08 recession even though air travel fell far more than driving — so it is not a fair-weather proxy.
4. Use it to bound claims rather than to forecast: it can't tell you demand growth to the decimal, but it can tell you with near-certainty when a headline demand number is impossible.

• Year-on-year commercial flight counts rolling over (the genuine demand-destruction signal); freight-carrier flight activity as the industrial read.

29:45 8. Read the gas-oil ratio as a field-ageing signal, not as supply growth

1. Understand the physics: in a Permian-type play, gas is dissolved in the oil in the same wellbore, "just like carbonation in a can of soda." As reservoir pressure falls, gas separates and "whooshes" out preferentially.
2. Therefore rising gas and NGL output alongside falling crude is not a growth story — it is the field getting older. "That's a little bit of a swan song."
3. Track the average well age of the basin, not just the rig count. While drilling accelerates, the average well gets younger; once drilling slows, it ages and the stream gassifies.
4. Discount any liquids-growth forecast that depends entirely on the NGL stream — the split-to-gas dislocation buys time, "but it doesn't last."
5. Hold the framework honestly: G&R got this wrong and says so — "we didn't have that modeled properly a couple years ago" — which is why US gas held up better than they expected.

• Basin gas-oil ratio trending up while oil output flattens; forecast growth that rests on NGLs rather than crude; lease/royalty offerings drifting off the core.

40:19 9. Test whether a producer's capital rule even allows a supply response

1. Work the CEO's arithmetic before assuming high prices bring new barrels. Cash on the balance sheet is credited at ~100 cents on the dollar (it backs out of enterprise value). Spending $100M to create $200M of NPV, in a market capitalizing reserves at half NAV, leaves the stock flat — while the best acreage is consumed.
2. Compare against the alternative use of that dollar: buying back stock at half NAV is "massively accretive." Where reserves trade below half NAV, the rational manager does not drill.
3. Check the price the budget is actually set against — companies hedge off the futures curve, not spot, and the curve "hasn't done nearly as much as the spot curve has." A spot spike with a flat strip produces no capex.
4. Conclude that the supply response is structurally delayed: the constraint is valuation and incentive, not geology, regulation or drilling technology.

• The back end of the futures curve, not spot; buyback authorizations versus capex guidance; a producer's price-to-NAV. The re-rating that finally makes drilling accretive is the turn signal.

42:10 10. Compute the strategic reserve's usable floor, not its headline volume

1. Start from the reported barrels in reserve, then subtract what cannot responsibly be withdrawn: single-cycle early storage caverns (drawing them is irreversible), and the ~10% operational minimum the operator wants to leave in the remaining caverns.
2. Divide the usable remainder by the announced weekly release rate to get weeks-to-tank-bottom.
3. Read the announced release size as a policy ceiling, not a capability: a 400M-barrel OECD headline is a Draghi "whatever it takes" gesture, and the actual release running below the promised pace tells you where the comfort limit really is.
4. Add the refill obligation to the forward balance — a drained reserve has to be rebought, "a very very cheap insurance policy" that becomes future demand.

• Weekly SPR release rate versus the announced schedule; the point at which single-cycle caverns get touched (a genuine desperation signal); the first refill tender.

1:02:08 11. Don't treat the absence of a crisis as evidence of abundance

1. Name the inference being made. "It hasn't broken yet, therefore it was never tight" is the reflexive market conclusion — and it strengthens with every quiet day, because each one is read as fresh confirmation of the original glut.
2. Ask instead what has been consumed to keep it quiet: inventory drawn, buffers spent, one-off releases used up. A calm surface bought with a non-repeatable resource is a worse position, not a better one.
3. Weigh insider revealed preference over public commentary: what did the people with the real balance sheet do, and what did they say the deadline was?
4. Set a falsification date so the view stays testable rather than perpetual.

• Cumulative inventory drawn versus the deficit still outstanding; buffers that cannot be used twice; policymakers' own stated timelines, which price the risk more honestly than the futures curve does.