Before you set Q4 budget, open igl-brand-kb/audience/profiles/05-carol-garden-witness.md section 8 and replace the line "fading fast (11.9% of last-120d; lane −22.5% in 90 days vs store −12.4%)" with the same-weeks-last-year pair and the ad spend printed beside it: June–August 2025 the bird-feeder lane sold $23,912 on $6,592 of Meta spend, which was 22.9% of the whole ad account; June–August 2026 it sold $1,129 on $68, which was 0.24%. Then, in Meta Ads Manager before Thanksgiving (26 Nov 2026), you un-pause the three existing bird-feeder ads that last delivered on 2026-01-15 rather than building new creative for that lane — PostID_Ad_001_Excuse Me in campaign 40_MANUAL_CBO_Scaling_WW_11.13..2025_Dup, and its two copies in 40_MANUAL_CBO_Scaling_US_10.12.2025 and 10_CBO_Scaling_US_SQUIRRELS_11.13.2025 — because they are PostID ads carrying the shirt's accumulated 1,232 ad comments, and they ran at 2.00x, 1.83x and 2.22x Meta-reported return in the last season they were funded. Finally, add one column to the Tuesday WeScale scorecard you already submit: each lane's share of Meta spend printed next to its revenue, so no lane is ever again recorded as fading in a month nobody bought it impressions.
Spitznagel's rule that a measurement window shorter than the payoff cycle turns a working strategy into an apparent loss led to the year-over-year test the dossiers never ran — and running it showed the opposite of what the dossier concluded: the lane that is 49.5% of all-time revenue was not tested and found wanting, it was defunded and then written up as dying.
The candidate that won tonight's assay proposed the right instrument and predicted the wrong reading. Its claim was that IGL's five WHO dossiers judge each buyer lane on a 90- or 120-day window, that this is shorter than the annual cycle gardeners live on, and that Carol's lane is therefore being recorded as decaying during her off-season — her own dossier says "her season OPENS at Thanksgiving." So: measure the same weeks last year instead. That instrument is correct and the dossiers genuinely lack it.
Run it and the verdict inverts. The same summer window a year apart is not a gentle seasonal dip: the bird-feeder lane went from $23,912 to $1,129, −95.3%, far worse than the −22.5% the dossier reports. Year-over-year would have confirmed the decay verdict, not overturned it. What overturns it is the number that sits in neither the dossier nor any knowledge base: over those same months the lane's share of the Meta ad account fell from 22.9% to 0.24% while total account spend stayed flat (−3%, $28,803 to $27,926). Sales did not fall away from a lane that was being advertised. Advertising fell away from the lane, and the sales followed it out.
The reason this matters in September rather than in the abstract is that the lane was not a poor performer at the moment it was cut. In the same summer months when it was funded it returned 1.54x Meta-reported revenue against 1.35x for everything else in the account; in December 2025, its own season, it took $13,882 of spend and returned 2.01x against 1.53x for the rest, and did $25,537 in revenue, up 32% on December 2024. It was the best-returning thing in the account, and its budget went to $948 in January, $79 in March, $4 in July. Over the same stretch the account's overall return fell from 1.39x to 0.97x. The rest of the store did grow — everything except this lane went from $53,792 to $102,679 January-through-August, +90.9% — which is the real and good news the single store-wide "−9.6%" hides. Both halves are invisible in any number that aggregates the lanes together.
No single shelf produces this. The founders shelf supplies the principle: Spitznagel's Universa bleeds 0.5% a month on options that expire worthless, sometimes for three years, and would read as a failure on every window shorter than its payoff — "being patient now in order to be strategically impatient later." The brand shelf supplies the artifact: five dossiers written on 2026-08-29, all marked CURRENT SoR, all judging on 90 or 120 days, one of them calling the company's largest lane "fading fast" with the caveat "partly seasonal: her season OPENS at Thanksgiving" left unresolved. The mastermind shelf supplies both the proof that the longer lens changes verdicts (Eric Corry, call #15: "Jan 2025 for sloths $981 in sales. Jan 2026 $84K," in a month brands write off) and the named failure mode (call #25's limiting-beliefs round: Chris Hall's "I can't overcome seasonality"). Tim's library supplies the agrarian version — F. H. King, 1911: "a durable system is governed by accumulated stocks and recurring inflows and outflows, not one season's output alone." And the ops database, which is not a knowledge base at all, supplies the fact that overturns the conclusion all four were converging on.
This also corrects a nugget shipped two nights ago. #026 (2026-08-31) measured the same collapse from the units side and diagnosed it as creative failure — "IGL is not dangerously dependent on one shirt, it lost its one winner and has not produced another." The winner was not lost. Its impressions were, from 1,002,719 in a summer to 5,005.
founders-kb/episodes/070-mark-spitznagel-the-dao-of-capital — the seed. Universa pays 0.5%/month on options that expire worthless, waiting up to three years for the payoff; "being patient now in order to be strategically impatient later"; the conifer that is slow first and accelerates later. Establishes that a measurement window shorter than the payoff cycle turns a working strategy into an apparent loss.igl-brand-kb/audience/profiles/05-carol-garden-witness — status: CURRENT SoR, created 2026-08-29 (5 days old). Lines 110-112: "$176k, 49.5% of all-time, one SKU = 36% of company units — and is fading fast (11.9% of last-120d; lane −22.5% in 90 days vs store −12.4%, partly seasonal: her season OPENS at Thanksgiving)." Line 32 confirms the annual cycle. These are quoted figures from a document, not a measurement — and it is the document this nugget corrects.igl-brand-kb/audience/profiles/02-donna-seed-keeper — status: CURRENT SoR, 2026-08-29, lines 108-109: "Flagship fatigued (Quietly Plotting: $21k lifetime → $444/90d)". A tomato-harvest-plotting phrase judged on a summer window, with the prescribed response being to replace it. The same window defect, second lane.wescale-kb/calls/call-15 — 2026-01-29. Eric Corry, verbatim: "Jan 2025 for sloths $981 in sales. Jan 2026 $84K," which the notes frame as landing "in January, which many brands treat as a dead month." The cohort's own proof that a year-over-year lens overturns a verdict on a written-off month.wescale-kb/calls/call-25 — 2026-04-09, the limiting-beliefs round: Chris Hall's "I can't overcome seasonality," Alex's "my snowsports brand can't scale through summer." Seasonality misreads are a named, recurring cohort failure that nobody has connected to a measurement-window fix.mylibrary-kb/books/farmers-of-forty-centuries — F. H. King, 1911, rights-verified public domain, enriched 2026-08-20. Framework 11: "a durable system is governed by accumulated stocks and recurring inflows and outflows, not one season's output alone," plus "One snapshot is not a growing season." The rule already sits on Tim's shelf in agricultural form and has never been pointed at the dossiers.igl-brand-kb/seasonal/six-season-calendar-operational — theme-aggregate. CURRENT SoR, updated 2026-03-03. Supplies each lane's real season window (Advent opens 29 Nov 2026, Thanksgiving 26 Nov; Epiphany 6 Jan – 17 Feb), which is the mechanism the dossier's season column would use.fact_product_sales_daily, fact_meta_daily and bridge_product_ads in G:\iris\igl-ops-sage\data\igl.db (operator queries run 2026-09-03; sales table covers 2024-07-30 to 2026-09-02, ad table 2024-07-22 to 2026-09-02). Every figure in this card that is not a document quotation comes from these three tables.igl-brand-kb/audience/profiles/05-carol-garden-witness.md section 8 and replace the line "fading fast (11.9% of last-120d; lane −22.5% in 90 days vs store −12.4%)" with the same-weeks-last-year pair and the ad spend printed beside it: June–August 2025 the bird-feeder lane sold $23,912 on $6,592 of Meta spend, which was 22.9% of the whole ad account; June–August 2026 it sold $1,129 on $68, which was 0.24%. Then, in Meta Ads Manager before Thanksgiving (26 Nov 2026), you un-pause the three existing bird-feeder ads that last delivered on 2026-01-15 rather than building new creative for that lane — PostID_Ad_001_Excuse Me in campaign 40_MANUAL_CBO_Scaling_WW_11.13..2025_Dup, and its two copies in 40_MANUAL_CBO_Scaling_US_10.12.2025 and 10_CBO_Scaling_US_SQUIRRELS_11.13.2025 — because they are PostID ads carrying the shirt's accumulated 1,232 ad comments, and they ran at 2.00x, 1.83x and 2.22x Meta-reported return in the last season they were funded. Finally, add one column to the Tuesday WeScale scorecard you already submit: each lane's share of Meta spend printed next to its revenue, so no lane is ever again recorded as fading in a month nobody bought it impressions. ## Why (the one-liner) Spitznagel's rule that a measurement window shorter than the payoff cycle turns a working strategy into an apparent loss led to the year-over-year test the dossiers never ran — and running it showed the opposite of what the dossier concluded: the lane that is 49.5% of all-time revenue was not tested and found wanting, it was defunded and then written up as dying. ## The insight The candidate that won tonight's assay proposed the right instrument and predicted the wrong reading. Its claim was that IGL's five WHO dossiers judge each buyer lane on a 90- or 120-day window, that this is shorter than the annual cycle gardeners live on, and that Carol's lane is therefore being recorded as decaying during her off-season — her own dossier says "her season OPENS at Thanksgiving." So: measure the same weeks last year instead. That instrument is correct and the dossiers genuinely lack it. Run it and the verdict inverts. The same summer window a year apart is not a gentle seasonal dip: the bird-feeder lane went from $23,912 to $1,129, −95.3%, far worse than the −22.5% the dossier reports. Year-over-year would have confirmed the decay verdict, not overturned it. What overturns it is the number that sits in neither the dossier nor any knowledge base: over those same months the lane's share of the Meta ad account fell from 22.9% to 0.24% while total account spend stayed flat (−3%, $28,803 to $27,926). Sales did not fall away from a lane that was being advertised. Advertising fell away from the lane, and the sales followed it out. The reason this matters in September rather than in the abstract is that the lane was not a poor performer at the moment it was cut. In the same summer months when it was funded it returned 1.54x Meta-reported revenue against 1.35x for everything else in the account; in December 2025, its own season, it took $13,882 of spend and returned 2.01x against 1.53x for the rest, and did $25,537 in revenue, up 32% on December 2024. It was the best-returning thing in the account, and its budget went to $948 in January, $79 in March, $4 in July. Over the same stretch the account's overall return fell from 1.39x to 0.97x. The rest of the store did grow — everything except this lane went from $53,792 to $102,679 January-through-August, +90.9% — which is the real and good news the single store-wide "−9.6%" hides. Both halves are invisible in any number that aggregates the lanes together. No single shelf produces this. The founders shelf supplies the principle: Spitznagel's Universa bleeds 0.5% a month on options that expire worthless, sometimes for three years, and would read as a failure on every window shorter than its payoff — "being patient now in order to be strategically impatient later." The brand shelf supplies the artifact: five dossiers written on 2026-08-29, all marked CURRENT SoR, all judging on 90 or 120 days, one of them calling the company's largest lane "fading fast" with the caveat "partly seasonal: her season OPENS at Thanksgiving" left unresolved. The mastermind shelf supplies both the proof that the longer lens changes verdicts (Eric Corry, call #15: "Jan 2025 for sloths $981 in sales. Jan 2026 $84K," in a month brands write off) and the named failure mode (call #25's limiting-beliefs round: Chris Hall's "I can't overcome seasonality"). Tim's library supplies the agrarian version — F. H. King, 1911: "a durable system is governed by accumulated stocks and recurring inflows and outflows, not one season's output alone." And the ops database, which is not a knowledge base at all, supplies the fact that overturns the conclusion all four were converging on. This also corrects a nugget shipped two nights ago. #026 (2026-08-31) measured the same collapse from the units side and diagnosed it as creative failure — "IGL is not dangerously dependent on one shirt, it lost its one winner and has not produced another." The winner was not lost. Its impressions were, from 1,002,719 in a summer to 5,005. ## Evidence trail - founders-kb/episodes/070-mark-spitznagel-the-dao-of-capital — the seed. Universa pays 0.5%/month on options that expire worthless, waiting up to three years for the payoff; "being patient now in order to be strategically impatient later"; the conifer that is slow first and accelerates later. Establishes that a measurement window shorter than the payoff cycle turns a working strategy into an apparent loss. - igl-brand-kb/audience/profiles/05-carol-garden-witness — status: CURRENT SoR, created 2026-08-29 (5 days old). Lines 110-112: "$176k, 49.5% of all-time, one SKU = 36% of company units — and is fading fast (11.9% of last-120d; lane −22.5% in 90 days vs store −12.4%, partly seasonal: her season OPENS at Thanksgiving)." Line 32 confirms the annual cycle. These are quoted figures from a document, not a measurement — and it is the document this nugget corrects. - igl-brand-kb/audience/profiles/02-donna-seed-keeper — status: CURRENT SoR, 2026-08-29, lines 108-109: "Flagship fatigued (Quietly Plotting: $21k lifetime → $444/90d)". A tomato-harvest-plotting phrase judged on a summer window, with the prescribed response being to replace it. The same window defect, second lane. - wescale-kb/calls/call-15 — 2026-01-29. Eric Corry, verbatim: "Jan 2025 for sloths $981 in sales. Jan 2026 $84K," which the notes frame as landing "in January, which many brands treat as a dead month." The cohort's own proof that a year-over-year lens overturns a verdict on a written-off month. - wescale-kb/calls/call-25 — 2026-04-09, the limiting-beliefs round: Chris Hall's "I can't overcome seasonality," Alex's "my snowsports brand can't scale through summer." Seasonality misreads are a named, recurring cohort failure that nobody has connected to a measurement-window fix. - mylibrary-kb/books/farmers-of-forty-centuries — F. H. King, 1911, rights-verified public domain, enriched 2026-08-20. Framework 11: "a durable system is governed by accumulated stocks and recurring inflows and outflows, not one season's output alone," plus "One snapshot is not a growing season." The rule already sits on Tim's shelf in agricultural form and has never been pointed at the dossiers. - igl-brand-kb/seasonal/six-season-calendar-operational — theme-aggregate. CURRENT SoR, updated 2026-03-03. Supplies each lane's real season window (Advent opens 29 Nov 2026, Thanksgiving 26 Nov; Epiphany 6 Jan – 17 Feb), which is the mechanism the dossier's season column would use. - Outside the knowledge bases — fact_product_sales_daily, fact_meta_daily and bridge_product_ads in G:\iris\igl-ops-sage\data\igl.db (operator queries run 2026-09-03; sales table covers 2024-07-30 to 2026-09-02, ad table 2024-07-22 to 2026-09-02). Every figure in this card that is not a document quotation comes from these three tables. ## Assay verdicts - novelty — PASS, 3/5. "No nugget has touched the measurement window on the WHO dossiers, and no cohort document offers a window longer than P30 (call-20's audit SOP)." Deducted because "Carol's own dossier already prints the caveat 'partly seasonal', so the hunt is finishing a sentence the source started." Operator note: the correction below lands past where the source's sentence stops — the spend-share collapse appears in no knowledge base, and independent greps for "PostID", "share of spend", "13,882" and "22.9%" across all thirteen bases and all 28 prior nuggets return zero hits. - grounding — PASS, 5/5. Every quote re-opened and verbatim: Carol lines 110-112, Donna 108-109, call-20 line 205 ("across P7, P14, P30 timeframes" and stops there), call-15 line 304, call-25 line 22. - actionability — PASS, 5/5. "Names the specific decision it changes and it is a live, imminent one: do not move Q4 budget away from the lane that is 49.5% of all-time revenue on the strength of a June-through-August window." - Summed strength 13/15 — top of three 3-of-3 survivors in round 1. ## Operator corrections applied before shipping 1. The move's hedge was unnecessary and was deleted. The candidate wrote "if the table does not reach back that far, pull the Shopify order export for Nov 2025-Feb 2026 once and keep it." fact_product_sales_daily covers 2024-07-30 to 2026-09-02 — over two years. The year-over-year comparison is directly runnable, so the nugget runs it instead of asking Tim to. 2. Running the proposed test reverses the candidate's conclusion, and this is the decisive correction. The candidate's stated reason for protecting Carol's Q4 budget was that a summer window misreads a seasonal trough as decay. Measured: same weeks a year apart, the lane is −95.3%, materially worse than the −22.5% the dossier reports. Its instrument was right and its reading was wrong. The finding that actually protects the budget is the 99% spend cut, which the candidate never looked at. 3. A causal claim was scoped down to what the data supports. This card does not assert that restoring spend restores revenue. It asserts the narrower and fully evidenced thing: the lane was not tested at low demand, it was defunded, so "fading" is not a conclusion the data licenses. The named action is a funded retest before Carol's season, not a prediction. 4. All return-on-ad-spend figures are labelled Meta-reported, not Shopify-verified. IGL's attribution is known to be partial, so these are the platform's numbers and are used only to compare the lane against the rest of the same account over the same days — a comparison that survives an attribution gap applied evenly. 5. Nugget #026's diagnosis is corrected rather than quietly contradicted, and the contradiction is stated in the body so the catalog does not carry two opposed readings of the same collapse. 6. Sixth night running, the decisive correction came from igl.db and nowhere else — and this time in its newest form: the candidate did not declare the measurement impossible (#028's failure) or quote a stale document (#007's), it proposed a live, correct query and then reasoned about the answer without running it. Thread: ep-070 Spitznagel (0.5%/month bleed, three-year payoff, "patient now to be strategically impatient later") → the measurement window is the time horizon → Carol dossier 2026-08-29, "fading fast, −22.5% in 90 days… her season OPENS at Thanksgiving" → the other four dossiers, all 90/120-day verdicts → call-20's audit SOP stops at P30 → call-15 Eric Corry $981 → $84K year-over-year January → call-25 "I can't overcome seasonality" as a named limiting belief → King, Farmers of Forty Centuries, "one snapshot is not a growing season" → six-season calendar supplies the real season windows → operator: run it against igl.db, and the lane is −95.3%, not −22.5% → check the spend and it fell from 22.9% of the account to 0.24%