Coca-Cola Consolidated, Inc. (COKE) — Investment Tree v1

### ⚠️ READ BOTH OF THESE BEFORE ANY NUMBER BELOW 1 — COKE is the BOTTLER, not KO. COKE = Coca-Cola Consolidated, Inc. (CIK 0000317540, NASDAQ), the largest US Coca-Cola bottler, headquartered in Charlotte, North Carolina. It is NOT The Coca-Cola Company (KO), the brand owner and concentrate maker. Their economics are not similar: KO's gross margin is ~1.6× and its operating margin ~2.4× COKE's. Data aggregators routinely conflate the two — this build's consensus-price-target lookup failed for exactly that reason. Any figure here that would suit a 62%-gross-margin concentrate business is a KO figure and does not belong. 2 — This ticker FAILED the long-term durability gate: 13/25. The threshold for a 5-10 year hold is 17/25. COKE scores 13/25, in the "12-16 — Low durability: short-term opportunistic only" band, and it is not boundary-sensitive — reaching 17 would require upgrading three separate questions, not one defensible judgement call. What that does and does not say: zero fatal flags fired. COKE is investment-grade (BBB+), cash-generative (~5% FCF yield) and its territory rights renew indefinitely at its own option. This is a statement that the business is capped, not that it is fragile. The essay was composed anyway, at the owner's explicit direction, with the finding carried in the open rather than buried. See §X.

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-09-06 · Anchor price: $189.33 (2026-09-04 close, Tier B) · Shares outstanding 66,564,294 (56,517,334 Common + 10,046,960 Class B, Tier A) · Market cap $12.60B · EV ~$15.05B · Reporting currency USD · US-domiciled, so no sovereign or FX wedge applies Archetype: franchised bottler with a contractually permissioned margin — a durable franchise whose profitability its own franchisor is free to compress

SOURCE QUALITY — read this before any number below. The spine of this tree is Tier A: the FY2025 10-K, the Q2 2026 10-Q (quarter ended 2026-07-03) and the 2025-11-07 8-K. evidence_2026-09-06.jsonl carries 18 rows — 11 Tier A, 5 Tier B, 2 Tier C. Four things that matter are not Tier A:

1. Every multiple, COKE's and every peer's — Tier B aggregator data, not recomputed from each issuer's own filings on a consistent basis. §VII's scenario multiples inherit that.
2. All KO margin figures — Tier B. The 2.41× operating-margin ratio in §VI rests on them.
3. The S&P rating action (BBB+, outlook Negative, 2025-11-14) — Tier B. Moody's rating: NOT VERIFIED — conflicting stale signals (Baa1 vs Baa2) and the source page returned 403.
4. Public float ~5.18M shares, institutional ownership ~62.9%, analyst coverage ~2, and the December 2022 $3.00 special dividend — all Tier C and unreconciled against primary filings.

NOT VERIFIED anywhere in this build: consensus price target · COKE's forward P/E · average daily volume and typical bid-ask spread · Q2 2025 operating margin · FY2024 gross and operating margin comparators. Anything marked † is Tier C arithmetic derived from Tier A inputs, with its assumption named in assumptions.md §B.

0. Company Fundamentals — what COKE is and how it earns

Figures FY2025 (year ended 2025-12-31) unless noted.

What it is. A capital-intensive franchised distributor. COKE buys concentrate from The Coca-Cola Company, manufactures finished beverages, and sells them into retail, grocery, foodservice and vending accounts within exclusive territories granted by KO. It does not own the brands. It does not set concentrate prices. It cannot expand outside its granted territory without KO.

The contractual architecture is the whole thesis:

• Comprehensive Beverage Agreements (CBAs), signed 2013-2017 during KO's US refranchising, grant the exclusive territory rights. Each carries a 10-year term, renewable indefinitely at COKE's option — genuinely durable, and COKE books the franchise rights as indefinite-lived intangibles.
• COKE pays KO's bottling arm quarterly sub-bottling payments for the grant.
• Sparkling concentrate runs through an Incidence-Based Pricing Agreement. KO has no right to establish the prices COKE charges its customers — but the concentrate price itself moves with an incidence formula, package mix and channel.
• For finished goods bought from KO or other bottlers, KO sets those prices unilaterally, and the 10-K states plainly that "there are no limits on the prices [KO] can charge for concentrate."

So COKE earns the spread between an input price its franchisor controls and an output price it sets itself. The territory is durable; the margin is not contractually protected.

How the money actually works.

Two things in that table are worth stopping on. First: net sales rose in both periods and net income fell in both. Second: COKE's capex intensity is essentially identical to KO's. The commonly asserted "capital-intensive bottler" penalty is not visible in capex at all. The entire gap between a bottler and a brand owner shows up in margin and in who controls it.

I. One-sentence verdict

COKE is a durable franchise with a permissioned margin, a contractually fixed growth ceiling and no governance recourse for an outside holder — a business that is capped rather than fragile, at a price that is cheap on EV/EBITDA, expensive on earnings, and offers a −4.5%† expected twelve-month return on this tree's own probabilities.

II. Company snapshot

III. The five facts that drive everything

1. The 10-K says the input price is uncapped. "There are no limits on the prices [KO] can charge for concentrate." Finished-goods prices are set by KO unilaterally. This is quoted, not inferred.
2. On 2025-11-07 COKE retired KO's entire 18,835,460-share stake at $127.00 (~$2.4B) and KO gave up its board seat — with the commercial contracts untouched. 22.06%† of the share count gone; ~$1,174M† = $17.64†/continuing share of one-time value at today's price; and the counterparty that sets COKE's input cost now owns none of COKE's residual.
3. It was funded with debt, and the cost is visible. H1 2026 interest $62.5M vs $12.8M — +388%. Leverage 2.6× against a 2× threshold. S&P outlook cut to Negative. Buyback authorisation cut $1B → $400M. Capitalised at 10%†, the recurring after-tax drag is $11.20†/share against the $17.64†/share gain.
4. Volume is up and profit is down, for two different reasons. FY2025: sales +4.8%, net income −9.9% — and only 17.4%† of that decline can be post-close interest, because the deal closed with 54 days left in the year. ~83%† of the FY2025 decline is not financing. Q2 2026: sales +10.6%, net income −15.2% — and roughly 66%† of that decline is the interest. Two mechanisms, one direction.
5. The 2026 volume is flattered and the company says so. H1 cases +10.3% reported, +7.1% adjusted for six extra selling days — COKE's own adjustment removes ~31%† of the headline. Management attributes further strength to America250™ and the FIFA World Cup. Neither recurs in 2027.

IV. The H-0 thesis

H-0. COKE's earnings are a spread between an input price its franchisor sets without contractual limit and an output price COKE sets itself, so the margin is permissioned rather than owned; the November 2025 buyout of KO's entire stake was well-priced but removed the last structural alignment between price-setter and residual owner and was debt-funded to 2.6× leverage; against a territory that cannot be expanded without KO's grant and a register 78% voting-controlled by one family, COKE's low headline multiple is better explained as an earned control-and-illiquidity discount than as an available mispricing.

What H-0 concedes, explicitly. The territory rights are genuinely durable and unusual — indefinitely renewable at COKE's option, booked as indefinite-lived intangibles. The buyout was a good trade on price. COKE is not fragile: BBB+, ~5% FCF yield, interest covered ~7.6׆ at the FY2025 operating-income level, zero fatal flags. And capital intensity is not the differentiator — COKE spends the same share of sales on capex as KO does.

What H-0 does not claim. It does not claim the franchise is at risk. It does not claim distress. It does not predict that KO will raise concentrate prices punitively — it claims nothing stops KO from doing so, and that a 25× multiple is paying for an assumption of forbearance with no contractual backing.

Falsifiers of H-0 as a whole (full table with data sources in h0_thesis.md): a disclosed cap on KO's concentrate pricing (FF1, Q1 2027); operating margin expanding ≥100bps for two consecutive quarters while volume normalises (FF2, 2027-12-31); a KO-granted territory expansion (FF3, 2027-12-31); leverage under 2.0× with S&P restored to Stable and the authorisation restored to ≥$1B (FF4, 2027-06-30); FY2027 volume ≥+5% with no event attribution (FF5, Feb 2028); coverage past 5 analysts with the KO discount narrowing below 1.5× absent a margin change (FF6, 2027-12-31).

FF1 and FF3 are the two that would break H-0 structurally rather than gradually. Neither has any indication in this evidence base, and FF3 runs directly against KO's disclosed asset-light refranchising direction.

V. Tree — five branches

L0 — Is COKE's low multiple an EARNED discount or an AVAILABLE mispricing?
│
├── L1A  Permissioned margin        ⚠️ mixed — the contract is clear, the mechanism is not measurable
│    ├── 1.1  Concentrate pricing is contractually uncapped ......... ✅A 支持       (LB, 3)
│    ├── 1.2  FY2025's decline is operational, not financing ........ ⚠️A 部分支持   (LB, 3)
│    └── 1.3  Q2 2026's decline is predominantly financing .......... ⚠️C 部分支持   (sup, 1)
│
├── L1B  The KO separation          ⚠️ good on price, worse on alignment, costly to fund
│    ├── 2.1  The buyout created one-time value for holders ......... ✅A 支持       (LB, 3)
│    ├── 2.2  KO's exit removed the last structural alignment ....... ⚠️A 部分支持   (LB, 3)
│    └── 2.3  The financing cost materially offsets the gain ........ ✅A 支持       (LB, 3)
│
├── L1C  The growth ceiling         ✅ the clearest branch in the tree
│    ├── 3.1  Reinvestment is bounded by contract, not funding ...... ✅A 支持       (LB, 3)
│    ├── 3.2  Material 2026 volume growth is non-recurring .......... ✅A 支持       (LB, 3)
│    └── 3.3  Category mix does not relieve the ceiling ............. ⚠️A 部分支持   (sup, 1)
│
├── L1D  Control and liquidity      ⛔ the conditions are present; the effect is not where it should be
│    ├── 4.1  Family control rose mechanically, not by purchase ..... ✅A 强力支持   (sup, 1)
│    ├── 4.2  The float is too small to exit normally ............... ⊗  证据不足    (LB, 3)
│    └── 4.3  Coverage + control produce a measurable discount ...... ⛔A 证据偏反   (LB, 3)
│
└── L1E  Valuation regime           ✗ the "cheapest in the set" claim does not survive its own table
     ├── 5.1  COKE is cheapest in its peer set on EV/EBITDA ......... ✗B 已证伪     (LB, 3)
     ├── 5.2  The KO multiple gap is explained by margin ............ ✅B 支持       (LB, 3)
     └── 5.3  A verifiable consensus target exists .................. ⊗  证据不足    (sup, 1)

Tally (terminal leaves only): 7 ✅ · 4 ⚠️ · 1 ⛔ · 1 ✗ · 2 ⊗ — 15 leaves. Full evidence, counter-evidence and falsification conditions in leaves.md.

h0 — computed, not felt

h0 = Σ(tier_weight × verdict_value) / Σ(tier_weight) over terminal leaves; weights crux 5 · load-bearing 3 · supporting 1; values 强力支持 0.80 · 支持 0.625 · 部分支持 0.45 · 证据偏反 0.225 · 已证伪 0; 证据不足 excluded from numerator and denominator.

h0 = 16.325 / 33 = 49.47% → 49% — the 35-54% partially supported band, below the 55% colour boundary.

Coverage: excluded weight 4 of 37 declared = 10.8%, well below the one-third threshold — no low-coverage flag. The shape is stated anyway: one of the two excluded leaves is load-bearing at weight 3 (L1D 4.2, the float question), excluded because a primary float figure and any average daily volume both do not exist in this build.

Crux test: no leaf in this tree is a crux, and that is a structural property. With h0 at 49.47% the tree already sits below the 55% boundary, so no single downward flip can cross a band — the worst single flip lands at 43.79%. This thesis is not hostage to any one fact; its weakness is distributed. The one leaf with genuine torque runs upward: resolving L1E 5.1 favourably moves h0 to 53.56% / 55.15% / 56.74% at 部分支持 / 支持 / 强力支持, crossing the 55% boundary at 支持 or better. It is the highest-value single item for the next refresh and it needs only one consistently-computed peer table.

VI. Key findings

1. The margin is permissioned, and that sentence is a quotation. Most "the supplier has power" arguments are inferences from margin trends. This one is in the filing: "there are no limits on the prices [KO] can charge for concentrate." COKE's gross margin is therefore not a moat — it is a tolerance. L1A 1.1, ✅A.

2. The profit decline has two causes and neither is "buyback noise." This is the finding the primer did not compute and the price appears not to reflect. FY2025 net income fell $62.7M† while sales rose 4.8%; the buyout closed with 54 days left in the fiscal year, so at most $10.9M† after tax — 17.4%† — can be financing. The other ~83% happened above the interest line. In Q2 2026 the pattern inverts: roughly 66%† of that quarter's $28.5M† decline is the interest. A market treating the whole thing as transaction mechanics is holding a view the calendar refuses.

3. The buyout was good, and about two-thirds of "good" goes to the lenders. Retiring 22.06%† of the shares at $127 against $189.33 is $17.64†/continuing share — real, executed, Tier A, and the strongest single item in COKE's favour anywhere in this tree. Against it: $1.12†/share/year of after-tax incremental interest, which at a 10%† discount rate capitalises to $11.20†/share. The residual is a good but ordinary outcome. The more informative fact is not in the arithmetic: management cut its own buyback authorisation from $1B to $400M. Companies do not surrender optionality they still have.

4. The control increase cost the family nothing, and the arithmetic proves it. Class B carries 20 votes. Retiring KO's one-vote Common took total votes from 276,291,994† to 257,456,534†, moving the Class B share from 72.73%† to 78.05%† — reproducing the 10-K's disclosed "~72% → ~78%" from independently disclosed inputs. An outside holder cannot force a sale, block a related-party transaction, or replace a director, in any realistic scenario. L1D 4.1, ✅A 强力支持 — the only 强力支持 in the tree, and it is a share-count identity rather than a judgement.

5. "Cheapest in the peer set" is false on its own table. primer.md states COKE carries the lowest EV/EBITDA in the set. PEP is at 12.20× against COKE's 12.35×. COKE is second-cheapest of the four names with data, by 15 basis points. Recorded as ✗B 已证伪 rather than quietly corrected, because that is what this system is for. L1E 5.1.

6. What survives that falsification is the finding that matters. The economically meaningful comparison — COKE 12.35× against KO 23.93× — is a 1.94× multiple gap set against a 2.41× operating-margin gap (31.85% vs 13.2%). The discount is smaller than the economics justify. And the usual escape hatch fails: COKE's capex is 4.32% of sales against KO's ~4.2%. A bottler trading near a brand owner's multiple would be the anomaly worth writing about. L1E 5.2, ✅B.

7. COKE is cheap on EV/EBITDA and expensive on earnings, simultaneously. At 25.31× trailing it sits above KDP's 13.35× forward and essentially level with KO's 25.91× forward. On the normalised post-transaction base of ~$7.45†/share it is 25.4× — so a "depressed denominator" rescue does not work either. The wedge between a 12.35× EV/EBITDA and a 25.31× P/E is arithmetically D&A (~$264.5M†), interest (~$125M† run-rate) and tax. The EV/EBITDA cheapness is largely a description of COKE's capital structure, not a governance discount. This is why L1D 4.3 lands at ⛔A 证据偏反 — the discount-generating conditions are all present, and the effect is not visible where it should be. It also means H-0's phrase "low headline multiple" is only half right, which the tree records rather than smooths.

8. The growth ceiling is contractual, not financial — and that is unusual enough to say twice. COKE cannot buy geography without KO's consent. Capital has exactly three destinations: efficiency inside the existing footprint, deleveraging, buybacks. There is no fourth. This is not a company that can grow its way out of a squeeze, and it is the reason durability Q5 scores 2/5.

9. Two honest gaps, one of them load-bearing. L1D 4.2 (is the float small enough to matter?) is ⊗ 证据不足 because the ~5.18M float figure is Tier C and unreconciled and no average daily volume exists in this build. L1E 5.3 (is there a verifiable consensus?) is ⊗ because the aggregator results were internally inconsistent — most plausibly COKE/KO ticker confusion, the exact error this document opens by warning about. Neither was estimated. The consequence is carried: no implied-upside-versus-consensus arithmetic appears anywhere in this build.

VII. Three valuation scenarios

12-month horizon, to ~2027-09. Full derivations, both routes, in scenarios.md.

The shared base, derived twice. FY2025 net income $570.6M on the post-transaction 66,564,294 shares = $8.57†/share; less annualised after-tax incremental interest ($99.4M† pre-tax → $74.6M† at 25%† = $1.12†/share) = $7.45†. Cross-check from the stated multiple: $189.33 ÷ 25.31× = $7.48†. Two independent routes, three cents apart. Share count held constant in all scenarios — the unused $400M authorisation is ~1% of the count and modelling it would add false precision.

Bull requires deleveraging below 2.0×, S&P restored to Stable, the authorisation used and re-upped, ~100bps of margin expansion, and a multiple re-rating toward KDP. Three independent favourable things, and the re-rating has no identified transmission mechanism in a name with roughly two analysts. The conservative end of the bracket is used for the midpoint for that reason.

Base is a flat stock. Not a failure of imagination — it is what ~25× on a mid-single-digit earnings base with a deleveraging obligation arithmetically produces over twelve months.

Bear requires nothing to break. Volume below +2% once the events lapse, operating margin below 13.0%, and a P/E that stops treating a leveraged franchised distributor as a staples compounder. That is the uncomfortable part.

Implied probability

Solving for the vector that sets expected value equal to spot, holding base at 50%:

245·p + 190·0.50 + 140·(0.50 − p) = 189.33   →   105·p = 24.33   →   p = 0.2317

The market is pricing COKE almost symmetrically, and this tree is the more bearish party. That is an unusual direction for this system and it is the arithmetic expression of a thesis whose conclusion is the discount is earned, not the discount is available.

EV = 245×0.15 + 190×0.50 + 140×0.35 = 36.75 + 95.00 + 49.00 = $180.75

Probability-weighted 12-month value $180.75† = −4.5%† expected return. Asymmetry 1.13:1 (+$55.67 up / −$49.33 down) against this system's 2:1 starter threshold.

Sensitivity, honestly. Three of the six multiples above are chosen rather than observed, because COKE has no populated forward P/E. Move the base multiple to 25.3× (no de-rating at all) and the expected return improves to −1.9%†. No single reasonable input change turns it positive — see implied_prob.md §6.

VIII. Triggers and red flags

Triggers (full table with node links in triggers_redflags.md)

Red flags

Expired-condition sweep: none expired. Earliest deadlines are F-A1.3a (2026-11-30) and F-A1.2a / F-D4.2a (2026-12-31). Next sweep at review_due 2026-12-05.

IX. Control and liquidity — what an outside holding actually is

This section replaces the currency-and-sovereign section that a non-US domicile would carry. COKE is US-domiciled and reports in USD, so there is no translation wedge and no sovereign_exposure band. The structural analogue for COKE is control and float, and it deserves the same first-class treatment.

The vote. Common carries 1 vote, Class B carries 20. Class B is unlisted and converts 1:1. The derivation:

This reproduces the 10-K's disclosed "~72% → ~78%" from independently disclosed inputs. The family bought nothing; retiring KO's low-vote Common shrank the denominator.

What that means concretely for a holder. No realistic path exists to force a sale, block a related-party transaction, replace a director, or influence capital allocation. An equity stake in COKE is an economic interest with a governance interest attached that rounds to zero. That is not a scandal — it is a disclosed, decades-old structure that many holders accept knowingly. It is stated here because a 5-10 year hold makes governance recourse matter more, not less.

The float — and why this section ends in an admission. Public float is reported at ~5.18M shares (Tier C, unreconciled) against ~62.9% institutional ownership (Tier C). Average daily volume: NOT VERIFIED. Typical bid-ask spread: NOT VERIFIED.

The direction is Tier-A-supported: only 56,517,334 Common shares are listed, Class B is unlisted, and 78.05%† voting control implies a large insider block. But the magnitude is the entire content of the question, and this build has no verified basis for it. So L1D 4.2 is ⊗ 证据不足 — a load-bearing leaf at weight 3, dropped out of the h0 calculation rather than estimated. A float-to-turnover test needs a float and a turnover, and this build has neither at a tier that would justify a verdict.

The uncomfortable interaction. If the float genuinely is that thin, then the very illiquidity that would justify a discount also makes the discount unharvestable — you cannot arbitrage a security you cannot exit. That is the shape of the whole problem, and it is why L1D 4.3 records the conditions as present but the measurable discount as absent.

X. Long-term holdability verdict

13/25 — LOW durability. Not suitable as a 5-10 year hold. Zero fatal flags.

Four points below the 17/25 threshold, and — unlike a boundary case — not sensitive to a single defensible judgement call. Moving COKE to 17 would require upgrading three separate questions.

What the score says, precisely. It is not a distress warning. Zero fatal flags fired: capital allocation is not chronic ROIC-below-WACC destruction, disruption risk is nowhere near the 80% threshold, and balance-sheet survivability passes cleanly on an investment-grade rating with interest covered ~7.6׆. It is a statement that the specific claim a 5-10 year "hold" makes — that the business compounds owner value over a decade — is not supported here. Margin is permissioned by a counterparty that just exited the equity; the reinvestment runway is bounded by a territory COKE cannot expand; profit fell in both reported periods while volume rose; and the balance sheet is absorbing a debt-funded buyback under a negative outlook.

Read it as: capped, not fragile. Those are different findings with different consequences. A fragile business should be avoided at any size. A capped business can be owned tactically and should not be owned for a decade.

XI. Position sizing and correlated exposure

Percentage bands only. This system does not record holdings, position values or cost basis anywhere.

Durability 13/25 maps to the K.3 "short-term opportunistic only" tier — which means no long-term core allocation at any size while the score stands.

Correlated exposure. cycle_exposure: uncorrelated — COKE is first-order independent of the AI-capex cycle and does not count toward the K.4 15% correlated-AI bucket. Against a portfolio weighted toward AI-adjacent names it is a genuine diversifier, and that is a real if modest part of its case.

The correlation that does need naming: COKE's economics are a direct function of a single counterparty, KO. Any position in KO alongside COKE is not diversification — it is the same system's profit pool held on both sides of a bilateral contract whose terms favour one side. If both were held, the KO position is partially a hedge on the COKE position and the combined exposure should be measured as one beverage-system sleeve, not two names.

A second, structural note on sizing. With float and average daily volume both NOT VERIFIED, the exit cost of any position here is unknown. Sizing rules assume you can leave. This one has not been shown to satisfy that assumption, which is an independent argument for the 1% cap over and above the durability band.

XII. Investment Scorecard (per MANUAL_en.md Part K.6)

15-question scorecard (analytical-tree Q-list, Format B per K.3.5)

Verdict tally (M1 evidence-tier suffixes per K.3.6): 7 ✅ · 6 ⚠️ · 2 ✗ — Q1✅A, Q2✅A, Q3⚠️B, Q4✅A, Q5⚠️B, Q6✅A, Q7✗A, Q8⚠️B, Q9⚠️A, Q10✅A, Q11✅B, Q12✅B, Q13⚠️A, Q14✗A, Q15⚠️C.

(Note on Q4: a ✅ on the bear-case row means the bear case is well-specified and evidenced, not that it will happen. A steelmanned bear case scores well when it is strong — that is the point of asking.)

K.3.5 Weighted-score derivation

Applying the Format B 4-tier mapping from MANUAL §K.3.5 (verdict values ✅ = 1.0, ⚠️ = 0.5, ✗ = 0.0; ⊗ excluded from numerator and denominator — there are no ⊗ rows here, so the denominator is the full 39):

64% = the 45-64% "wait or skip" band per K.3.5 — one point below the 65% moderate-buy boundary. Unlike the durability score, this one is boundary-sensitive: a single ⚠️→✅ upgrade on any Confirming row would carry it to 65.4%. The boundary sensitivity is disclosed rather than resolved in COKE's favour, because the row most likely to move (Q8, FCF growth) is ⚠️ precisely because the data does not exist.

What the score is made of. The Critical tier is where the damage sits: it contributes only 7.5 of a possible 15.0, because Q14 (moat trajectory) is the tree's second ✗ and Q9 (leverage against the company's own threshold) is a ⚠️. Set against that, the Load-bearing tier scores 10.5 of 12.0 — but read what those rows are: the bear case, the sell triggers and the falsification set. This scorecard is being carried by the quality of its own bear case. That is a strange and telling shape: the analysis is in good order and the business it describes is not compounding.

The h0 / xii_score gap is 15 points (49% vs 64%) — narrow by this corpus's standards, and the narrowness is itself the finding. Where a wide gap says "a better company than this thesis is a thesis", a narrow one says the two measurements agree: the structural quality and the thesis confidence are both middling, and neither is being rescued by the other. There is no catalyst-shaped headroom here to close.

K.3.1 precedence check. The fatal-flag override sits above the weighted score, and here it does not bind: zero fatal flags, balance-sheet survivability passes cleanly. The binding constraint on size is the durability band — 13/25, the 12-16 "short-term opportunistic only" tier — not a flag. That is a stricter constraint than the xii_score would impose on its own, and it is what governs §XI.

Scorecard summary

Final verdict: AVOID as a long-term hold. HOLD-WITH-SIZING as a tactical position only — 0% today, ≤1% ever, and only after two triggers fire together

This is a framework output, not advice to any person. It is what the K.3.1/K.3.5 discipline, the fatal-flag check and the durability band produce when applied to the evidence above.

• 🚫 Long-term sleeve: 0%, hard, independent of price. Durability 13/25 sits in the "short-term opportunistic only" band, four points below threshold and not boundary-sensitive. The claim a decade-long hold makes is the specific claim this evidence does not support.
• 🔍 Today: 0%. Expected twelve-month return −4.5%†; asymmetry 1.13:1 against a 2:1 threshold; xii_score 64% in the wait-or-skip band; and the market's implied vector (23/50/27) is more optimistic than this tree's (15/50/35). There is nothing here to be early to.
• ⬆️ Tactical entry, if at all: ≤1%, and only after T2 and T4 fire together — one clean post-event quarter (~Nov 2026) confirming volume holds, and leverage below 2.0× with S&P outlook restored to Stable (2027-06-30). Two conditions, not one.
• 🚫 Hard cap 1%. Binding. A 2/5 reinvestment runway, zero governance recourse and unverified exit liquidity are three independent arguments against size.
• 🔻 Exit within 30 days if RF1 or RF6 fires — a disclosed concentrate-price step-up, or operating margin below 12.0% for two consecutive quarters.
• ⛔ Suspend the tree entirely if RF5 or RF7 fires — a non-market related-party transaction with the controlling family, or any KO strategic action on the territories. Rebuild from Stage 1.
• 📅 Next review 2026-12-05, after Q3 2026 results.

The 2-minute pitch:

"First, this is the bottler, not Coca-Cola. Coca-Cola Consolidated buys concentrate from KO and delivers finished drinks in territories KO granted it. It earns 13% operating margins where KO earns 32%, and the reason is in the 10-K in one sentence: there are no limits on the prices KO can charge it for concentrate. COKE sets what it charges retailers; KO sets what COKE pays. That is the whole business. Last November it did something dramatic — bought back KO's entire 18.8-million-share stake at $127 when the stock is now $189, retiring 22% of the shares and handing continuing holders about $17.64 a share. Good trade. It also borrowed $1.2 billion to do it, so interest went from $12.8 million to $62.5 million in a half-year, leverage went to 2.6 times against its own 2-times limit, S&P moved the outlook to negative, and management cut its buyback authorisation from a billion to four hundred million. And KO gave up its board seat while keeping every bit of its pricing power. Meanwhile volume is up 7.6% — helped by six extra selling days and by the World Cup and America250, which management named itself and which do not happen again — and net income fell 15%. The stock looks cheap at 12 times EV/EBITDA and is not cheap at 25 times earnings; both numbers are correct, and the difference is depreciation and the new debt. Durability scores 13 out of 25, which is a fail for a ten-year hold — but with zero fatal flags. It is investment grade, it generates cash, and its territory rights renew forever at its own option. It is not fragile. It is capped: it cannot buy new geography without KO's permission, one family holds 78% of the votes, and roughly two analysts follow it. On my probabilities the expected twelve-month return is minus four and a half percent, and the market is already more optimistic than I am. Nothing to do."

Risk types most relevant (per MANUAL_en.md Part K.4):

• Counterparty / contract risk (dominant, and unusual) — the supplier, the franchisor and the price-setter are one entity, and its latitude is contractually unlimited. This is not a supply-chain risk; it is a claim on the margin.
• Leverage and rating risk — 2.6× against a 2.0× threshold under a Negative outlook, with interest expense up 388% and the buyback authorisation cut 60%.
• Governance risk — 78.05%† family voting control, risen mechanically. No minority recourse in any realistic scenario; the specific exposure is a related-party transaction on non-market terms (RF5).
• Liquidity risk (unquantified and therefore treated as binding) — float Tier C and unreconciled, ADV NOT VERIFIED. Sizing rules assume you can leave; that assumption has not been demonstrated here.
• Category / demand risk — long-run US CSD per-capita decline and 2026 GLP-1 findings against COKE's own +7.0% sparkling growth and no GLP-1 commentary at all. Genuinely unresolved, and the still-beverage hedge moves ~0.4pp†/year.
• Valuation risk — a 12.35× EV/EBITDA that a screen reads as cheap sits on a 25.31× P/E. The screen is the trap.
• Correlated-factor risk — uncorrelated to the AI-capex bucket and a genuine diversifier there. But any KO position alongside COKE is the same beverage system held on both sides of a contract that favours one side.

"When NOT to buy" anti-pattern check (per MANUAL_en.md Part K.5):

• ❌ People online talking about it — the opposite problem. ~2 analysts and an unpopulated forward P/E. There is no crowd.
• ❌ The stock went up fast — no evidence of a momentum move in this build; the anchor is a single close.
• ❌ It has "AI" attached — it does not. cycle_exposure: uncorrelated.
• ⚠️ It looks "cheap" — this is the live anti-pattern, and it is the whole reason this tree exists. A screen shows 12.35× EV/EBITDA against KO's 23.93× and calls it a 48% discount. The same security is at 25.31× trailing earnings, and the 1.94× multiple gap to KO is narrower than the 2.41× operating-margin gap. The cheapness is a description of the capital structure, not an opportunity.
• ⚠️ You remember the brand — the second live anti-pattern, and it is severe here specifically. Brand familiarity with Coca-Cola is close to universal and entirely irrelevant to this security, which does not own the brand. Worse, it is the mechanism by which the COKE/KO confusion happens: a familiar name makes a wrong ticker feel checked. This document opens with the disambiguation for that reason.
• ❌ "The next [X]" — no pattern-matching claim is made anywhere in this tree.

Net: 2 anti-pattern flags, and they compound rather than sitting side by side — a famous brand attached to a screen-cheap multiple, on a security that owns neither the brand nor the margin.

XIII. What's NOT in this tree

• L1D 4.2 — whether the float is small enough to matter — is load-bearing at weight 3 and is ⊗. Public float is Tier C and unreconciled; average daily volume does not exist in this build at any tier. This is the largest single gap, and it matters more than usual because §XI's sizing argument leans on it.
• L1E 5.3 — a verifiable consensus target and forward P/E — is ⊗. Aggregator results were internally inconsistent, most plausibly COKE/KO ticker confusion. The consequence is enforced, not just noted: no implied-upside-versus-consensus arithmetic appears anywhere in this build.
• Moody's rating is NOT VERIFIED — conflicting stale signals (Baa1 vs Baa2), source returned 403. It is used nowhere.
• FY2024 gross and operating margin comparators are absent, which is why L1A 1.2 caps at 部分支持: the FY2025 decline's direction is Tier A, its mechanism is not evidenced.
• Q2 2025's operating margin is NOT VERIFIED, and the Q2 decomposition in L1A 1.3 rests on assuming it approximated 13.2%. For a beverage distributor Q2 is seasonally the strongest quarter, so that assumption is conservative in the wrong direction — flagged rather than smoothed.
• No bottler-versus-bottler EV/EBITDA comparison exists. CCEP and KOF are the right comparators and neither carries an EV/EBITDA in this pack. Every EV/EBITDA comparison available here is a bottler against a brand owner. Closing this is trigger T1 and it is the highest-value refresh item in the file.
• The December 2022 $3.00 special dividend is Tier C and unreconciled. It is excluded from all yield and capital-return analysis rather than assumed.
• No DCF. Terminal value would be dominated by an assumption about a concentrate price the counterparty sets without limit — a DCF here is an opinion about KO wearing a spreadsheet.
• Three of the six scenario multiples in §VII are chosen rather than observed, because COKE has no populated forward multiple to observe. The bands are wider in reality than the arithmetic makes them look.