Create a Long-Term Compounder Screening Agent for Global Equity Research

Finding long-term compounders requires more than screening for companies with high revenue growth. A business may grow quickly while margins deteriorate, capital requirements rise, or returns on invested capital weaken.

For a global equity team, the challenge is to evaluate these characteristics consistently across companies and over multiple fiscal years. That means combining historical growth, profitability, capital efficiency, and cash-flow measures before deciding which names deserve deeper research.

In this workflow, we use Claude with Financial Modeling Prep's MCP Server to build a repeatable compounder screen. Using completed annual financial data, the screen calculates a defined set of multi-year metrics, applies consistent rules, and turns the results into a compact analyst-review queue.

The objective is not to identify stocks to buy. It is to narrow a research universe systematically, surface companies that warrant further investigation, and flag cases where the underlying data or business economics require analyst judgment.

Key Takeaways

• The screen combines five-year revenue growth with average gross margin, operating margin, ROIC, and free-cash-flow margin.
• Six annual revenue observations are required to calculate a five-year CAGR. The remaining metrics use the latest five completed fiscal years.
• ASML and TSM both qualified as Strong Quality and Strong Compounding in the completed sample run, giving each a High priority for deeper research.
• Missing core history, misaligned fiscal periods, or conflicting fields trigger Review Required instead of a forced classification.

Defining the Long-Term Compounder Methodology

The screening framework combines four characteristics that matter for long-term compounder research: sustained growth, durable profitability, capital efficiency, and cash-flow support. Long-term quality is rarely captured by one ratio, so the screen looks for strength across several measures rather than relying on a single favorable result.

For each company, the screen uses completed annual fiscal periods and calculates:

• Five-year revenue CAGR: Based on six annual revenue observations, from the starting fiscal year to the latest completed fiscal year.
• Average gross margin: The arithmetic mean across the latest five completed fiscal years.
• Average operating margin: The arithmetic mean across the same five-year window.
• Average ROIC: The five-year average of FMP's returnOnInvestedCapital field. It serves as the available capital-efficiency measure and is not presented as cash flow return on investment, or CFROI.
• Average FCF margin: The five-year average of annual free cash flow divided by revenue. This provides a supporting view of cash conversion.

The thresholds are intentionally simple. Strong Quality requires an average gross margin of at least 40%, an average operating margin of at least 20%, and an average ROIC of at least 15%. A five-year revenue CAGR of at least 10% qualifies as Strong Compounding.

Quality and compounding classifications determine research priority through a fixed matrix:

Quality	Strong Compounding	Moderate Compounding	Slow Compounding	Contracting
Strong Quality	High	Medium	Low	Low
Moderate Quality	Medium	Medium	Low	Low
Weak Quality	Low	Low	Low	Low

Strong Quality means all three quality thresholds are met. Moderate Quality means two are met, and Weak Quality means one or none are met. Strong Compounding requires revenue CAGR of at least 10%, Moderate Compounding covers 5% to less than 10%, Slow Compounding covers 0% to less than 5%, and Contracting applies when revenue CAGR is below 0%.

High priority therefore requires Strong Quality plus Strong Compounding. Medium captures companies with enough quality and growth to warrant further investigation but that do not clear both strongest classifications. Weak Quality or a Slow or Contracting profile receives Low priority.

Review priority means priority for deeper analyst research, not investment attractiveness or a buy or sell recommendation.

These thresholds are screening rules, not universal definitions of a compounder. Their usefulness may vary across sectors, business models, countries, and accounting conventions. ASML and TSM are both semiconductor companies, but their operating models differ materially. ASML sells semiconductor-manufacturing equipment and related services, while TSM operates a capital-intensive semiconductor foundry.

The same thresholds therefore provide a consistent first pass, not evidence that the companies' economics are directly comparable. When the screen expands, analysts should benchmark each company against an appropriate peer group before drawing broader conclusions.

Data availability is assessed separately:

Data Confidence	Required Condition
High	All required annual observations are present and fiscal periods align.
Medium	Core CAGR, margin, and ROIC calculations are complete, but a supporting measure such as FCF margin has incomplete history.
Review Required	A core measure lacks enough history, fiscal periods cannot be aligned, or conflicting data prevents a reliable classification.

No missing value is estimated or silently replaced with another metric. Preserving period definitions and unresolved fields is essential when maintaining consistent financial inputs.

FMP Data and Claude MCP Setup

The compounder screen uses three FMP datasets, each serving a specific role:

FMP Dataset	Returned Fields Used	Purpose
Income Statement API	revenue, grossProfit, operatingIncome, fiscalYear, period	Revenue CAGR, gross margin, operating margin, and fiscal-period alignment
Key Metrics API	returnOnInvestedCapital, fiscalYear, period	Five-year average ROIC
Cash Flow Statement API	freeCashFlow, fiscalYear, period	Five-year average FCF margin

To access these datasets through Claude, first connect the FMP MCP Server. An active FMP API key is required. In Claude, open Settings, select Connectors, choose Add custom connector, and enter the remote MCP URL using your API key. The current connection steps are available in the FMP MCP documentation.

FMP MCP Connection URL

https://financialmodelingprep.com/mcp?apikey=YOUR_FMP_API_KEY

Give the connector a name, add it, and start a new Claude conversation. Once connected, Claude can discover and call the available FMP tools without separate API requests or pasted financial data.

For this workflow, request completed annual periods only. Six annual revenue observations are needed to calculate a five-year CAGR because the calculation requires a starting value five fiscal years before the latest completed year. Gross margin, operating margin, ROIC, and FCF margin use the latest five completed fiscal years.

Request only the fields needed for the calculations. Claude can then align the observations by fiscal year, calculate the screening metrics, and apply the quality, compounding, and confidence rules. Fiscal years must remain consistent across the three datasets because misaligned reporting periods can distort comparisons.

If the required annual history is incomplete, the fiscal periods cannot be aligned, or a core metric is unavailable, the workflow returns Review Required instead of estimating or substituting a value.

Building the Compounder Screen

The initial run uses ASML and TSM as a two-company global semiconductor sample. Keeping the demonstration within the same broad industry reduces some sector-level differences, while the companies' different home markets show how the workflow can extend across a global research universe. The companies are evaluated independently rather than ranked against one another.

The five-year measurement window reduces the influence of a single strong or weak fiscal year while remaining recent enough to reflect the company's current operating profile. Six revenue observations are requested because a five-year CAGR requires a starting value five fiscal years before the latest completed year.

Gross margin of 40%, operating margin of 20%, ROIC of 15%, and revenue CAGR of 10% create a demanding quality-and-growth screen. These thresholds demonstrate the classification framework and should be reconsidered when the universe, sector, or research mandate changes.

ROIC is used because it is available through FMP's Key Metrics API and provides a practical measure of capital efficiency. It is not treated as an equivalent to CFROI. The prompt also limits the number of calls and retries so a missing field triggers review instead of repeated requests.

Claude Prompt

Using the connected Financial Modeling Prep (FMP) MCP server, run a compact long-term compounder screen for this predefined sample global equity universe: - ASML Holding N.V. (ASML) - Taiwan Semiconductor Manufacturing Company (TSM) This is a research-screening exercise, not an investment recommendation. Evaluate each company independently; do not rank the companies against each other. Use only completed annual fiscal periods available through the run date. Do not use TTM, quarterly, live-price, or forward-estimate data. Keep the workflow compact. Use only the minimum FMP calls required and no more than one retry total if a request fails. Data required for each company: - latest 6 completed annual revenue observations from the income statement - latest 5 completed annual gross profit and revenue observations - latest 5 completed annual operating income and revenue observations - latest 5 completed annual ROIC observations from key metrics - latest 5 completed annual free cash flow observations Calculate: 5-year revenue CAGR: (latest annual revenue / revenue 5 fiscal years earlier)^(1/5) - 1 Average 5-year gross margin: calculate gross profit / revenue for each of the latest 5 fiscal years, then take the arithmetic mean. Average 5-year operating margin: calculate operating income / revenue for each of the latest 5 fiscal years, then take the arithmetic mean. Average 5-year ROIC: take the arithmetic mean of the latest 5 annual returnOnInvestedCapital observations. ROIC is the available capital-efficiency proxy only. Do not call it CFROI. Average 5-year FCF margin: calculate free cash flow / revenue for each of the latest 5 fiscal years, then take the arithmetic mean. Illustrative screening rules: Quality: Strong Quality: gross margin >= 40%, operating margin >= 20%, and ROIC >= 15% Moderate Quality: two of the three conditions are met Weak Quality: zero or one condition is met Compounding profile: Strong Compounding: 5-year revenue CAGR >= 10% Moderate Compounding: 5% to <10% Slow Compounding: 0% to <5% Contracting: <0% Review priority: High: Strong Quality + Strong Compounding Medium: Strong Quality + Moderate Compounding, Moderate Quality + Strong Compounding, or Moderate Quality + Moderate Compounding Low: any Weak Quality classification, or any Slow Compounding / Contracting profile Review priority means priority for further analyst research, not a buy/sell recommendation. Data confidence: - High: all required annual observations are present and fiscal periods align - Medium: core CAGR, margin, and ROIC calculations are complete, but a supporting metric such as FCF margin has incomplete history - Low confidence / Review Required: a core metric lacks enough history, fiscal periods cannot be aligned, or conflicting data prevents a reliable classification Do not estimate missing values or silently substitute another metric. Return: 1. A compact screening table with: Ticker, Company, 5Y Revenue CAGR, Avg. Gross Margin, Avg. Operating Margin, Avg. ROIC, Avg. FCF Margin, Quality, Compounding Profile, Review Priority, Data Confidence. 2. State the fiscal-year window used. 3. Provide a short research note for each company, maximum 3 bullets: - what primarily drove the classification - one metric or trend that deserves analyst attention - one specific follow-up question for an analyst If data is insufficient, state Review Required rather than forcing a conclusion. Keep the entire response concise and focused on the screening results.

Screening Results

The screening run used FY2020-FY2025 for the five-year revenue CAGR and FY2021-FY2025 for profitability, ROIC, and free cash flow. Both companies had complete annual histories across the required window, resulting in High data confidence.

The first table reports the calculated metrics. The second shows the resulting classifications.

Ticker	Company	5Y Revenue CAGR	Avg. Gross Margin	Avg. Operating Margin	Avg. ROIC	Avg. FCF Margin
ASML	ASML Holding N.V.	18.50%	51.73%	33.26%	30.54%	33.38%
TSM	Taiwan Semiconductor Manufacturing Company	23.50%	56.31%	45.92%	21.14%	22.28%

Ticker	Quality	Compounding Profile	Review Priority	Data Confidence
ASML	Strong Quality	Strong Compounding	High	High
TSM	Strong Quality	Strong Compounding	High	High

Reproducing the Calculations from the Annual Inputs

The required annual inputs were retrieved through FMP MCP on August 29, 2026, using the stated datasets and fiscal-year windows. No value was estimated, interpolated, or replaced.

ASML figures are reported in EUR millions and TSM figures in TWD millions. FY2020 is used only as the starting revenue observation for the five-year CAGR. The profitability, ROIC, and FCF calculations use FY2021-FY2025.

The first table contains the revenue and operating inputs. The second contains ROIC and free cash flow.

Company	Fiscal Year	Revenue	Gross Profit	Operating Income
ASML	FY2020	EUR 13,978.5M	Not used	Not used
ASML	FY2021	EUR 18,611.0M	EUR 9,809.0M	EUR 6,750.1M
ASML	FY2022	EUR 21,173.4M	EUR 10,700.1M	EUR 6,500.7M
ASML	FY2023	EUR 27,558.5M	EUR 14,136.1M	EUR 9,042.3M
ASML	FY2024	EUR 28,262.9M	EUR 14,492.0M	EUR 9,022.6M
ASML	FY2025	EUR 32,667.3M	EUR 17,258.0M	EUR 11,301.4M
TSM	FY2020	TWD 1,339,254.8M	Not used	Not used
TSM	FY2021	TWD 1,587,415.0M	TWD 819,537.3M	TWD 649,980.9M
TSM	FY2022	TWD 2,263,891.3M	TWD 1,348,354.8M	TWD 1,121,278.9M
TSM	FY2023	TWD 2,161,736.0M	TWD 1,175,111.0M	TWD 921,466.0M
TSM	FY2024	TWD 2,894,308.0M	TWD 1,624,354.0M	TWD 1,322,053.0M
TSM	FY2025	TWD 3,848,510.9M	TWD 2,304,925.3M	TWD 1,955,695.8M

Company	Fiscal Year	ROIC	Free Cash Flow
ASML	FY2020	Not used	Not used
ASML	FY2021	31.02%	EUR 10,367.4M
ASML	FY2022	29.25%	EUR 7,036.7M
ASML	FY2023	32.08%	EUR 3,230.0M
ASML	FY2024	24.93%	EUR 9,504.7M
ASML	FY2025	35.40%	EUR 10,647.4M
TSM	FY2020	Not used	Not used
TSM	FY2021	18.83%	TWD 262,724.3M
TSM	FY2022	24.35%	TWD 520,972.7M
TSM	FY2023	17.41%	TWD 286,568.9M
TSM	FY2024	20.00%	TWD 870,170.0M
TSM	FY2025	25.10%	TWD 1,097,584.0M

The five screening metrics for each company reproduce to the reported values when calculated from these annual inputs. The CAGR, profitability, ROIC, and FCF-margin results therefore remain traceable to the underlying observations without estimation or substitution.

ASML: Strong Quality with High Capital Efficiency

ASML clears all three quality thresholds and exceeds the 10% revenue CAGR threshold. Its five-year average ROIC of 30.54% provides particularly strong support for the capital-efficiency side of the screen.

The main follow-up area is cash-flow consistency. Although the five-year average FCF margin is strong, the annual observations vary considerably. That is why comparing margins with cash-flow conversion matters before treating a strong average as evidence of stable annual performance.

Analyst follow-up: How stable has ASML's FCF conversion been across individual fiscal years, and what explains the material variation?

TSM: Strong Growth with High Operating Margins

TSM also qualifies as Strong Quality and Strong Compounding. Its five-year revenue CAGR of 23.50% clears the Strong Compounding threshold, while its average gross margin, operating margin, and ROIC all exceed the Strong Quality thresholds.

TSM's average FCF margin is lower than its average operating margin, making cash conversion an important area for deeper review.

Analyst follow-up: How persistent has the gap between operating margin and FCF margin been across the five-year period, and what has driven the variation?

Both companies receive High review priority. This does not mean they are equally attractive investments or that one should be preferred over the other. It means the historical screen found enough evidence of durable growth, profitability, and capital efficiency to justify deeper fundamental research.

Scaling the Screen with Scheduled Claude Tasks

Once the screening logic has been checked, the same process can become a repeatable financial-data workflow rather than being rebuilt for each review.

Scheduled tasks in Claude Cowork can run automatically or on demand and can use connectors configured in Cowork. Because this screen relies on completed annual financial statements, it does not need to run continuously. A practical approach is to run it after new annual data becomes available or schedule checks during reporting periods.

The same controls used to refresh a financial model after new earnings data apply here: confirm that the annual period is complete, preserve the prior result, recalculate the metrics, and flag any classification or confidence change.

A scaled task could:

1. Retrieve the latest completed annual history for the defined universe.
2. Recalculate the compounder metrics and classifications.
3. Flag companies whose classification or data confidence changed.
4. Produce an updated analyst-review queue.

For a larger global universe, process companies in manageable batches rather than asking Claude to analyze the entire market in one run. Before expanding coverage, match the FMP plan to the required datasets, historical depth, and request volume. The calculation rules and Review Required logic can remain unchanged while the coverage universe grows.

When an Analyst Needs to Step In

The screen can standardize the first pass, but several situations should still trigger human review. An analyst should step in when:

• A company does not have enough completed annual history for the required calculations.
• Fiscal periods do not align cleanly across datasets.
• A core metric such as ROIC is missing or inconsistent.
• An unusually strong or weak average is driven by one exceptional fiscal year.
• Accounting differences across markets make direct interpretation less straightforward.
• The screen identifies a high-priority company, but the business drivers behind the historical results remain unclear.

This last step is especially important. The screen can show what changed in the financial profile, but it should not invent the business reason behind the numbers. Analysts can investigate product mix, capital spending, competitive position, cyclicality, and industry conditions, then compare management's explanation with the reported financials before forming an investment view.

From Screening Metrics to a Repeatable Research Process

A long-term compounder screen becomes more useful when the methodology is explicit, the historical window is consistent, and missing information triggers review instead of being filled automatically.

Using FMP data through Claude MCP, this workflow turns growth, profitability, capital efficiency, and cash-flow measures into a repeatable first-pass research process. The output is not an investment recommendation. It is a structured way to identify which companies deserve deeper analyst attention.

Once the logic has been validated on a small sample, the same framework can be applied to a broader global universe or scheduled as a recurring research task without changing the underlying screening rules.