Check Management Cost Claims Against Reported Results With Claude and FMP

Restaurant executives talk about menu pricing and food costs on almost every earnings call, often in the same sentence. A CFO might say a cost ratio improved and credit price increases and efficiencies for it. The first half of that statement is a number that can be checked against the company's own results. The second half is an explanation that reported figures alone may not establish.

This example checks Chipotle Mexican Grill (CMG) management statements from its Q2 2025 and Q4 2025 calls. Claude retrieves four quarterly income statements, eight U.S. CPI observations, and the relevant transcripts through FMP's MCP server, then tests each claim against evidence with the same metric and comparison period.

The two historical claims tested both concern the food, beverage and packaging cost ratio. Both agree with the company's reported ratios. A third statement, a pricing outlook for Q1 2026, is recorded but remains outside the selected results window.

The check does not isolate how much of any cost change came from pricing. Standardized margin and CPI appear as context, and management's explanations remain attribution. The claim determines which evidence can test it.

Key Takeaways

• Q2 2025 food, beverage and packaging costs were 28.9% of revenue against 29.4% a year earlier, a 50-basis-point decline. Q4 was 30.2% against 30.4%, a 20-basis-point decline. Both measured claims are Supported.
• Supported applies to the measured change. It does not allocate the change among pricing, efficiencies, ingredient costs, or other factors.
• For Q2 and Q4 2025, the retrieved FMP costOfRevenue values reconcile to four disclosed restaurant operating cost lines. The calculated margins agree with Chipotle's non-GAAP restaurant-level operating margins at the company's displayed precision.
• The Q1 2026 outlook remains Review Required because matching results are outside this example's scope.
• Quarter-end-month CPI inflation ranges from 2.38% to 3.02% year over year. It is broad context, not a test of Chipotle's prices or costs.

How the Claim Check Is Built

Chipotle is a useful test case because management quantifies cost-ratio changes and its results releases report the corresponding measures. The example covers fiscal 2025, the most recent complete year at the September 2026 analysis date.

Each claim is tested against the measure and comparison period it names. A year-over-year food-cost-ratio claim requires that ratio for the same quarter a year earlier. A sequential movement in a broader margin measures something else and cannot confirm or contradict it.

Prior-year evidence is limited to Q2 2024 and Q4 2024, the comparisons needed for the selected claims.

Context Measures

Measure	Calculation	Role
Standardized margin	FMP grossProfit ÷ revenue × 100	Sequential context; not the claim test
CPI YoY	(2025 monthly index ÷ same month in 2024 − 1) × 100	Broad U.S. inflation context

For the two reconciled quarters, FMP grossProfit equals revenue less Chipotle's four restaurant operating cost lines. Do not call this GAAP gross margin or assume the same mapping holds for every period or company.

CPI does not measure Chipotle's menu prices, ingredient purchases, wages, or tariff exposure. It never confirms or contradicts a company-specific claim in this workflow.

Separate Outcome From Attribution

A statement that a cost ratio fell about 50 basis points because of menu-price increases contains two claims: a measured outcome and an explanation. The outcome can be compared with the reported ratio. The explanation requires evidence decomposing the effects of pricing and other drivers.

Classification	Meaning
Supported	A specific historical claim agrees with evidence for the same metric and period
Mixed	Aligned evidence complicates or conflicts with another aligned observation of the same claim
Unsupported	Correctly aligned evidence contradicts the testable claim
Review Required	Matching evidence is absent from the analysis, the wording is untestable, or the necessary decomposition is unavailable

Missing evidence never makes a claim Unsupported. A metric or period mismatch is Review Required, not Mixed.

Confidence describes evidence completeness and alignment, not management credibility. High applies when the exact historical measure and period are established; Medium reflects remaining interpretive limits for a testable statement; Low reflects an unassessed or untestable statement within this workflow. These are illustrative screening labels.

FMP Evidence and Claude Setup

Follow the FMP MCP connection instructions to connect the server in Claude, confirm access to the required datasets, and test a simple company lookup before running the prompt.

Evidence layer	Recorded MCP tool and operation	Use
Quarterly results	statements / income-statement	Revenue, costOfRevenue, and grossProfit
Cost mapping attempt	statements / as-reported-income-statements	Company-labeled cost fields
Inflation context	economics / economics-indicators	Eight monthly CPI values
Call identification	earningsTranscript / transcripts-dates-by-symbol	Fiscal-quarter and call-date records
Management statements	earningsTranscript / search-transcripts	Selected 2025 and matching 2024 calls

These are labels used in the recorded MCP workflow, rather than literal REST paths. The Income Statement documentation describes the underlying financial dataset.

Retrieval Record

The original run was September 30, 2026; time of day was not captured. On October 16, between 10:06 and 10:08 UTC, the income-statement and CPI requests were repeated to document exact parameters. Their returned values matched the original run.

Operation	Parameters	Returned records
income-statement	symbol=CMG, period=quarter, limit=12	12 quarters; only Q1-Q4 2025 used
as-reported-income-statements	symbol=CMG, period=quarter, limit=8	8 quarters; mapping attempt not used
economics-indicators	name=CPI; from_date=to_date=first day of each selected month	One observation per call
transcripts-dates-by-symbol	symbol=CMG	Call list
search-transcripts	symbol=CMG; year/quarter: 2025/2, 2024/2, 2025/4, 2024/4	One transcript per request

CPI Observations

For each row, from_date and to_date were both the displayed observation date.

Observation date	Returned CPI
2024-03-01	312.345
2024-06-01	313.044
2024-09-01	314.732
2024-12-01	317.604
2025-03-01	319.785
2025-06-01	321.435
2025-09-01	324.245
2025-12-01	326.031

FMP returned name, date, and value without a series identifier, seasonal-adjustment status, units, or index base. The October 16 external check matched all eight observations to FRED's CPIAUCSL series: monthly, seasonally adjusted, all-items CPI for U.S. urban consumers, indexed to 1982-1984 = 100. This identifies the series externally; it is not metadata returned by FMP.

The values represent individual months. They are not quarter averages or real-time historical vintages.

As-Reported Extraction Limitation

The as-reported fields could not be assembled cleanly into the four restaurant cost lines. For Q2 2025, the extraction left approximately $135 million unassigned against standardized costOfRevenue; the Q4 extraction did not reconcile internally.

The extraction was not used to establish the mapping. Chipotle's releases supplied the direct cost-line reconciliation shown below. This identifies a limitation of the retrieved extraction, not a demonstrated discrepancy in the company's reported totals.

Run the Claim Check in Claude

The prompt below retains the complete MCP-only workflow. The issuer and FRED verification described afterward was performed separately.

Objective Use the connected Financial Modeling Prep (FMP) MCP server to test selected management pricing and cost claims for Chipotle Mexican Grill (CMG) against evidence matching each claim's metric and comparison period. Classify each as Supported, Mixed, Unsupported, or Review Required. Keep management attribution separate from demonstrated causation. Scope - Company: CMG. - Financial periods: Q1, Q2, Q3, and Q4 2025. - Claim transcripts: Q2 2025 and Q4 2025. - Prior-year evidence: Q2 2024 and Q4 2024 only, for YoY claims. - CPI: March, June, September, and December in 2024 and 2025. - FMP MCP only. Do not use web search or other sources to fill gaps. 1. Retrieval record Record retrieval date and UTC time; if unavailable, write "time not captured". For every call record the tool, operation, parameters, fields, record count, and status. Do not rename operations or guess parameters. Allow one narrow retry per source and record it. 2. Quarterly financial records Call statements / income-statement: symbol=CMG, period=quarter, limit=12. Keep fiscalYear 2025, Q1 to Q4. Confirm the period basis is quarterly, not FY, TTM, YTD, six-month, or nine-month. Retain revenue, costOfRevenue, and grossProfit at full precision. Check revenue - costOfRevenue = grossProfit. Standardized margin = grossProfit / revenue x 100. For Q2 to Q4: QoQ revenue % = (current revenue / previous revenue - 1) x 100. QoQ margin change = current margin - previous margin, in percentage points. Q1 sequential changes = N/A. Round only for display. 3. Field mapping Call statements / as-reported-income-statements: symbol=CMG, period=quarter, limit=8. Check which company cost lines make up costOfRevenue for Q2 and Q4 2025. If they do not reconcile, write "Not established from retrieved FMP evidence". Do not infer the mapping. Never label standardized margin GAAP gross margin. 4. CPI Call economics / economics-indicators, name=CPI, once per selected month. Set from_date and to_date to that month's first day. Keep returned date and value. CPI YoY % = (2025 value / same month 2024 value - 1) x 100. Report metadata only if returned; otherwise "Not returned by FMP". Label the result quarter-end-month CPI YoY, not quarterly inflation. CPI is context only, never evidence for or against company pricing, food, labor, or tariff claims. 5. Transcripts Call earningsTranscript / transcripts-dates-by-symbol, symbol=CMG. Then retrieve search-transcripts for year/quarter: 2025/2, 2025/4, 2024/2, and 2024/4. Identify calls by fiscal quarter, not calendar year of the call date. Record conflicts between returned dates and transcript content. 6. Claims Select up to five material pricing or cost claims from the two 2025 calls. Keep speaker, call date, short exact excerpt, metric, comparison period, qualifiers, and whether historical or forward-looking. Separate measured outcome from explanation, labeled Management attribution. Test YoY claims only against the same metric in the same prior-year quarter. Never use sequential standardized margin as the test. 7. Classification Supported: specific historical claim, matching metric and period, evidence agrees. Mixed: aligned evidence complicates or conflicts with another aligned observation of the same claim. Unsupported: correctly aligned evidence contradicts the claim. Review Required: missing matching evidence, vague wording, required decomposition unavailable, or statement outside the selected results window. A metric or period mismatch is Review Required, not Mixed. Do not treat a statement as untestable solely because it began as guidance; assess whether matching actuals are within scope. Confidence: High, Medium, or Low for evidence completeness and alignment, not management credibility. Review Required = Yes when further work is needed to judge the statement. Output Table 1: Quarter | Revenue (USD M) | costOfRevenue (USD M) | grossProfit (USD M) | Standardized margin | QoQ revenue | QoQ margin (pp) | Quarter-end-month CPI YoY. Below: period ends, validation result, and the eight CPI observations. Table 2: Claim | Call | Metric and period | Matching evidence | Classification | Confidence | Review Required. Keep cells short. Below add: Statements: speakers and short excerpts. Management attribution. Revenue context: total revenue, comparable sales, discrete revenue items. Follow-up: distinguish the measured claim from untested attribution. Finish with the retrieval record, limitations, and at most four interpretation bullets. Do not give an investment or management-quality rating.

What the Run Returned

All four selected records were quarterly, and revenue minus costOfRevenue equaled grossProfit in each.

Table 1. Standardized Results and Inflation Context

Monetary values in USD millions. Sequential changes use unrounded inputs.

Quarter	Revenue	costOfRevenue	grossProfit	Margin	QoQ revenue	QoQ margin (pp)	CPI YoY
Q1 2025	2,875.3	2,121.6	753.6	26.21%	N/A	N/A	2.38% (Mar)
Q2 2025	3,063.4	2,225.2	838.2	27.36%	+6.54%	+1.15	2.68% (Jun)
Q3 2025	3,003.4	2,267.9	735.6	24.49%	−1.96%	−2.87	3.02% (Sep)
Q4 2025	2,983.5	2,284.7	698.8	23.42%	−0.66%	−1.07	2.65% (Dec)

Period ends are March 31, June 30, September 30, and December 31, 2025. Display rounding can make a subtraction appear different by $0.1 million.

Unrounded revenue inputs in dollars: 2,875,253,000; 3,063,393,000; 3,003,444,000; 2,983,511,000. Corresponding costOfRevenue inputs: 2,121,631,000; 2,225,163,000; 2,267,861,000; 2,284,739,000.

The sequential margin rose in Q2 and fell in Q3 and Q4. That pattern is context, not the test of the year-over-year claims.

Table 2. Claims and Matching Evidence

Claim	Call	Matching evidence	Classification	Confidence	Review
C1: Food, beverage and packaging ratio down about 50 bps YoY	Q2 2025; July 23, 2025	Q2: 28.9% in 2025 vs 29.4% in 2024; −50 bps	Supported	High	No
C2: Same ratio down about 20 bps YoY	Q4 2025; February 3, 2026	Q4: 30.2% in 2025 vs 30.4% in 2024; −20 bps	Supported	High	No
C3: About 70 bps of Q1 2026 pricing impact against expected inflation approaching mid-single digits	Q4 2025; February 3, 2026	Matching Q1 2026 actuals outside this results window	Review Required: outside selected window	Low for assessment in this run	Yes

All three statements came from CFO Adam Rymer. C1 described cost of sales at 28.9%, down approximately 50 basis points from the prior year. C2 described 30.2%, down approximately 20 basis points. For C3, Rymer expected first-quarter pricing impact of about 70 basis points against inflation approaching the mid-single-digit range.

Management attribution: For C1, management credited menu-price carryover and cost-of-sales efficiencies, offset by steak and chicken inflation. For C2, it credited menu pricing, lower dairy costs, and efficiencies, partly offset by beef and chicken inflation and tariffs. The analysis does not allocate the observed change among those drivers.

Revenue context: Q2 total revenue rose 3.0% while comparable restaurant sales fell 4.0%. Q4 revenue rose approximately 4.9% while comparable sales fell 2.5%. New restaurant openings contributed to growth; the Q4 release also reports $27.0 million in gift-card breakage revenue, up $19.1 million year over year, without an effect on comparable sales.

Follow-up: No further work is needed to confirm the two measured cost-ratio changes. Their attribution remains untested. Assessing C3 requires matching Q1 2026 pricing and cost evidence in a separate analysis.

External Verification

The issuer and FRED checks were completed after the MCP run. They are not outputs of the MCP-only prompt.

Reconcile the Standardized Cost Field

Chipotle's Q2 2025 results and Q4 2025 results provide the cost lines used below.

Table 3. Cost-Line Reconciliation

Amounts in USD millions unless stated otherwise.

Cost line or measure	Q2 2025	Q4 2025
Food, beverage and packaging	885.989	900.155
Labor	756.261	760.524
Occupancy	154.250	162.493
Other operating costs	428.663	461.567
Total restaurant operating costs	2,225.163	2,284.739
FMP costOfRevenue	2,225.163	2,284.739
Calculated standardized margin	27.36%	23.42%
Company restaurant-level operating margin, non-GAAP	27.4%	23.4%

For both quarters, the four cost lines sum exactly to the retrieved FMP costOfRevenue value. The calculated margins agree with the company measure at its displayed precision.

General and administrative expenses, depreciation and amortization, pre-opening costs, and impairment, closure, and asset-disposal costs sit outside this mapped cost field. This is why the standardized margin should not be described as GAAP gross margin.

The releases also directly corroborate the 28.9% versus 29.4% and 30.2% versus 30.4% comparisons used in the claim table.

Keep the Transcript-Date Exception Visible

The retrieved FMP Q4 2024 transcript record carries 2024-02-04. Chipotle's Q4 2024 results announcement establishes the call date as February 4, 2025. The source date is retained as an exception; the verified date is used to identify the call. Its content was consistent with fiscal Q4 2024.

The CPI verification identifies the series externally as described above. It does not turn the current retrieved series into an unrevised historical snapshot.

Reading the Claim Results

Both Supported findings concern cost ratios. Chipotle said food, beverage and packaging costs consumed a smaller share of revenue than a year earlier, and the reported ratios agree with the stated declines.

This does not establish pricing power. Management linked the changes to pricing among other factors, while comparable sales declined in both quarters. Total revenue and comparable sales also measure different populations, and Q4 included a discrete gift-card revenue contribution. Isolating price, transactions, mix, and cost effects requires additional disclosures and analysis.

The standardized margin supplies a separate cross-check. Its decline from 27.36% in Q2 to 23.42% in Q4 means the four restaurant operating cost lines consumed a larger share of revenue in Q4 than Q2. That sequential observation neither contradicts nor weakens a year-over-year food-cost-ratio finding: the measures and comparison periods differ.

C3 remains unassessed here because Q1 2026 results are outside the selected window, not because an outlook can never be checked after the period ends. Its Low confidence label describes this run's incomplete assessment, not the credibility of the speaker.

Apply the Same Rule to the Next Claim

A claim needs review when the matching metric or period is absent, its wording is too vague to test, or it needs a decomposition the evidence does not provide. Preserve source conflicts and resolve them before relying on the affected record.

The direct extension is a claim about Chipotle's labor ratio. Q2 2025 labor costs rose from 24.1% to 24.7% of revenue while the food-cost ratio fell. Testing a labor statement requires its own matching ratio and period; the food-cost result cannot stand in for it.

For another company, establish what the standardized cost field includes before interpreting its margin. Keep the claim text, metric, period, evidence, classification, and untested attribution together. The useful output is a reviewable claim record, not a blanket judgment that management is right or wrong.

Frequently Asked Questions

Does Supported Mean Management's Explanation Was Proven?

No. It means the measured historical outcome agrees with matching evidence. Attribution to pricing, efficiencies, or other causes requires separate support.

Can Gross Margin Be Used to Test Any Cost Claim?

No. The metric must match the claim. A broad standardized margin cannot substitute for a specific ingredient-cost ratio, and the underlying cost-field mapping must be established.

Why Include CPI?

It provides broad inflation context. It is not Chipotle-specific pricing or input-cost evidence and does not determine claim classifications.

Why Is the Q1 2026 Statement Still Review Required?

The matching actual results are outside this example's 2025 results window. It needs a separate assessment using the forecast's metric, period, and qualifiers.

Can the Workflow Be Reused?

Yes, with updated company, periods, transcripts, and metric definitions. Preserve the matching rule and report unavailable evidence without substituting a different measure.