Signals Desk Weekly | Multi-Year CAGR Strength Taking Shape Across Five Names (Aug 24-28)

Five years is long enough for a margin story to stop being a cost programme and start being a business model. This week's screen ran five-year revenue and EBITDA compound growth side by side across 899 index constituents and kept only the names where earnings compounded materially faster than sales: Salesforce, Sterling Infrastructure, Neurocrine Biosciences, CME Group and Exelixis. The spreads run from roughly 5 percentage points to almost 32.

The comparison is built on FMP's Income Statement API, taking revenue and EBITDA from matched fiscal years so every company is measured on the same footing. What follows is the five names, what each spread is actually made of, and how the same framework scales into a repeatable CAGR screen.

Key Takeaways

• Each of these five spreads has a different mechanism behind it. Cost reduction, mix shift, launch reuse, fixed-cost leverage and research reprioritisation are not the same signal, and they do not decay in the same way.
• Base-year quality decided this screen more than the arithmetic did. Around a dozen wider spreads were set aside because the starting year carried an impairment, a pandemic trough or an acquisition that reset the revenue line.
• Two of the five are already showing the spread narrowing in current results, which is arguably more informative than the five-year figure that surfaced them.
• A CAGR spread is a research trigger. EBITDA that never appears in operating cash flow, or that was produced by deferring spend rather than earning it, reads very differently once the cash statement is attached.

Five Spreads and What Each One Is Made Of

Salesforce, Inc. (CRM)

5-Year Revenue CAGR: 14.34%
5-Year EBITDA CAGR: 31.84%

Salesforce is the cleanest cost-programme story in large-cap software, and the numbers behind the spread are unusually legible. EBITDA margin moved from 15.5% to 31.7% between the January 2021 and January 2026 fiscal years, driven by headcount and real-estate reduction, tighter sales-capacity planning, and a deliberate rotation away from acquisition-led growth toward buybacks. The most recent completed year delivered revenue of $41.5 billion, non-GAAP operating margin of 34.1% and free cash flow of $14.4 billion.

The quarter reported on 26 August, inside the week this screen covers, shows the margin holding but the growth composition changing. Revenue rose 11%, current remaining performance obligation reached $33.5 billion, and non-GAAP operating margin stayed at 34.1%. The full-year guidance raise, however, was $200 million made up of roughly $100 million organic, $200 million from the pending Contentful and Fin transactions and $100 million of currency drag. GAAP margin guidance was trimmed while the non-GAAP figure held, which is where acquisition cost tends to sit.

That is the tension worth carrying forward. Three acquisitions inside about twelve months, one of them near $3.6 billion, from a company that spent three years telling the market it had stopped buying growth. Whether the cost discipline survives consolidation of those businesses shows up year by year through FMP's Income Statement Growth API. A second-order question sits underneath: management has described internal support headcount falling from roughly 9,000 to around 5,000 as agents absorbed volume, and the same logic eventually points back at a seat-based pricing model.

Sterling Infrastructure, Inc. (STRL)

5-Year Revenue CAGR: 15.21%
5-Year EBITDA CAGR: 32.18%

Sterling earned its spread by changing what it sells rather than by spending less. The company walked away from low-bid heavy highway work and redeployed capacity into E-Infrastructure, meaning data centre and semiconductor site development, where returns sit well above traditional civil contracting. Adjusted EBITDA margin moved from 16.3% in 2024 to 20.2% in 2025 and reached 22.0% in the June quarter. Transportation revenue is shrinking on purpose, and its margin still rose 500 basis points to 19.5%, because what remains is better work.

The June quarter was the largest in the company's history: revenue up 90%, adjusted EBITDA up 104%, E-Infrastructure at $905.0 million and now 78% of the business, and signed backlog up 116% to $4.33 billion. Guidance was raised. The market reaction was nonetheless negative, because E-Infrastructure segment margin drifted from the high twenties toward the mid twenties.

That reaction points at the most interesting feature of this name, which is mix dilution caused by its own success. Specialty electrical services acquired through CEC run near 12% margins against legacy site development in the high twenties, and the lower-margin business is growing three to four times faster. Every line can improve while the blended figure falls, which is precisely the case a consolidated CAGR obscures and the Revenue Product Segmentation API resolves. Three constraints deserve attention regardless: electrician availability, which management said curtailed bookings, an expected sequential backlog decline in the September quarter on award timing, and federal transportation funding that expires in September 2026.

Neurocrine Biosciences, Inc. (NBIX)

5-Year Revenue CAGR: 22.29%
5-Year EBITDA CAGR: 35.10%

Neurocrine's spread comes from reusing infrastructure rather than building it. INGREZZA scaled toward a run rate above $2.8 billion against a specialty salesforce sized years earlier, so incremental prescriptions carried very high contribution margin. The company then pushed two further products through the same commercial organisation instead of standing up new ones: CRENESSITY into congenital adrenal hyperplasia, and VYKAT XR from May 2026. That is portfolio leverage, and it is a more repeatable mechanism than a cost programme because it can be applied again to the next asset.

The June quarter showed all three contributing at once. Total revenue rose 39% to $959 million, with INGREZZA at $716 million, CRENESSITY at $184 million and VYKAT XR adding $54 million from 18 May. Full-year INGREZZA guidance was raised. Both selling and research expense guidance moved up alongside it, which is the recognisable shape of a company spending its operating leverage on the next launch rather than banking it.

The balance sheet changed with it. The $2.9 billion Soleno acquisition closed on 18 May, funded partly through a $1.0 billion credit facility, so a previously debt-free profile now carries a fixed charge against those margins, and the Balance Sheet Statement API is where that interaction becomes trackable. Two items frame the next eighteen months: Phase 3 readouts for osavampator and direclidine have moved into the second half of 2027, lengthening the interval the current research run rate has to justify, and management expects VYKAT XR discontinuation to settle at 25% to 30%.

CME Group Inc. (CME)

5-Year Revenue CAGR: 5.95%
5-Year EBITDA CAGR: 11.28%

CME produces the most mechanical spread in the group. Revenue grew 39% across the window while operating margin expanded roughly 850 basis points, for the straightforward reason that an additional cleared contract costs the exchange almost nothing to process. Two structural elements reinforced it. Market data, which is recurring and high margin, reached a record $803 million in 2025 across 33 consecutive quarters of year-over-year growth. Cross-margining now saves clients more than $95 billion of margin per day, which operates as a retention mechanism rather than a revenue line.

The June quarter is where that mechanism ran backwards for the first time in this cycle. Revenue was essentially flat, expenses rose to $599.1 million from $562.7 million, and adjusted operating margin fell to 69.5% from 71.0%. Average daily volume slipped 1.2% and rate per contract eased to $0.678 from $0.690, even as market data set another record at $238.1 million, up 20%.

The open question is whether that reflects one quarter of product-launch and cloud-migration spend or a genuine step-up in the run rate. Single-stock futures, round-the-clock gold, Treasury clearing and compute futures all arrived in the same window, while the regulator stayed the round-the-clock crude launch, so cost landed ahead of any revenue. A leadership transition compounds it, with Lynne Fitzpatrick due to become chief executive in March 2027. Holding a five-year margin series and one deteriorating quarter in the same frame is what the Financial Ratios API is for.

Exelixis, Inc. (EXEL)

5-Year Revenue CAGR: 18.63%
5-Year EBITDA CAGR: 50.55%

Exelixis carries the widest spread in the screen, with EBITDA margin moving from 12.1% to 39.7%. Two things produced it. Cabozantinib scaled into a commercial base built years earlier, carrying gross margin near 96%, so almost every additional dollar of sales dropped through. Then in January 2024 the company cut roughly 13% of headcount, terminated a Phase I programme and a discovery partnership, and research spending fell from $910.4 million in 2024 to $825.0 million in 2025 while selling and administrative cost rose only 5.4%. Net income grew 50.1% on 7.0% revenue growth.

The August quarter shows the mechanism still working and shows its boundary at the same time. Revenue rose 10.6% and non-GAAP earnings reached $0.91 per share against $0.75, but full-year revenue guidance came down to $2.50 billion to $2.55 billion on a slower neuroendocrine tumour ramp. Research guidance was cut by $50 million in the same release, so the profitability guide held while the top line fell. It is difficult to find a cleaner illustration of margin defended by spending less.

The concentration underneath is the reason this name needs the caveat rather than the headline. Essentially all product revenue comes from one molecule, and the litigation settlement licenses a United States generic from 1 January 2031. The structure survives that only if zanzalintinib converts, and its first approval decision carries a 3 December 2026 target date. The research reduction that flattered current margins is, on one reading, deferred spending on the asset that has to replace the franchise. FMP's Financial Estimates API is where you can see whether consensus is modelling the cliff, the replacement, or neither.

Where the Margin Actually Came From

Five spreads, five mechanisms, and they are not interchangeable. Sterling is the only one of the group expanding margin by changing what it sells, which is the most durable version of the signal and also the one that dilutes itself as the acquired lower-margin services scale. CME's leverage is the most purely arithmetic and therefore the most exposed to a cost step-up, which is exactly what the June quarter delivered. Salesforce completed its cost programme and is now partly re-spending the proceeds on acquisitions. Neurocrine is deliberately converting leverage into new launches rather than retaining it. Exelixis is holding margin by deferring research against a dated exclusivity cliff.

Ranked that way, a useful distinction emerges. Mechanisms that work by subtraction, meaning headcount, research budgets and discretionary cost, cannot repeat indefinitely because the base eventually runs out. Mechanisms that work by substitution, meaning selling different work into a different market, can. A five-year CAGR spread records a decision already taken; the forward question is only ever whether the mechanism behind it can run again.

Base-year quality mattered more here than the calculation did, because a five-year window is quietly reset by a single impairment or an acquisition that redefines the revenue line, and the statements that expose it sit alongside the growth series on the FMP platform, which is what allowed roughly a dozen wider spreads to be rejected rather than published. From there the tests run in sequence. The Cash Flow Statement API establishes whether reported EBITDA arrived as cash or as working capital. The Owner Earnings API strips out the capital intensity that EBITDA ignores, which matters most for Sterling and Neurocrine. The Price Target Consensus API and the Financial Estimates API then indicate how much of the improvement analysts have already written into their models.

Read that way, the screen does what a screen should. It narrows 899 names to five worth the time, and it moves the question from whether margins improved, which the arithmetic already answered, to whether the improvement was earned in a way that can happen again.

Building a Consistent CAGR Screening Framework

Building a useful CAGR screen is less about the formula itself and more about maintaining discipline in the underlying dataset. The calculation is straightforward; what determines whether the result is meaningful is data consistency. Every company needs to be evaluated using the same reporting periods, identical financial line items, and the same time horizon. Once those inputs are standardized, growth rates become comparable across industries, capital structures, and business models. The workflow below shows how to structure that process using FMP's Income Statement data — starting with a single company and then scaling the exact same logic across a broader universe.

Step 1: Pull Income Statement Data

Begin with a single symbol to establish the baseline. Query the standard Income Statement API to retrieve the full set of historical reporting periods needed for the calculation.

As long as your API key is active, one request gives you the raw time series you'll be working with. For example:

Endpoint:

https://financialmodelingprep.com/stable/income-statement?symbol=AAPL&apikey=YOUR_API_KEY

Step 2: Gather Historical Figures

From the JSON output, select the specific metric you want to analyze — revenue, EBITDA, EPS, or another line item. Arrange the values in proper chronological order before doing any math. This step is easy to overlook, but it's critical: CAGR only makes sense when the starting and ending points are clearly defined and consistently ordered.

Step 3: Calculate CAGR

Once the first and last data points are set, calculate CAGR using the standard formula:

CAGR = (Ending Value / Beginning Value)^(1 / Years) - 1

This reduces several years of performance into a single annualized figure, making it easier to compare growth profiles across companies without getting lost in interim volatility.

Step 4: Scale Screening with Bulk API

After validating the method on one symbol, broaden the workflow using the Income Statement Bulk API:

https://financialmodelingprep.com/stable/income-statement-bulk?year=2025&period=FY&apikey=YOUR_API_KEY

Running the same calculation at scale lets you build filters — for instance, highlighting companies that clear a five-year revenue CAGR threshold — while ensuring every ticker is processed under the same ruleset. Once the bulk pull is in place, updating or rerunning the screen is effectively a single action.

Scaling the Framework Without Changing the Methodology

The strength of this type of screen comes from consistency, not complexity. Once the formula, reporting periods, and financial line items are defined, the objective is to keep the methodology fixed while gradually widening the universe being tested. Expanding coverage should not require changing the framework itself — only the number of companies moving through it.

That's why the workflow is easiest to validate in a smaller environment first. Within the Basic plan, the Income Statement endpoints provide enough historical coverage to align reporting periods properly, normalize the selected metrics, and verify that the CAGR calculations are producing comparable outputs across companies. At this stage, the emphasis is less about scale and more about making sure inconsistencies in filings or missing data are not distorting the screen.

From there, expanding into the Starter plan simply broadens the sample size. The screening logic remains identical, but the larger universe makes sector-level comparisons more useful. Patterns that initially appear company-specific can then be evaluated against peers, industries, or market-cap cohorts to determine whether the operating leverage signal is isolated or part of a broader trend developing within a segment of the market.

The Premium plan extends that same structure further by increasing historical depth and geographic coverage. The underlying process still does not change. What changes is the scope of observation — allowing the same framework to be applied across wider datasets without introducing new assumptions or altering the screening criteria midstream. That continuity is what makes the process repeatable over time rather than dependent on one-off observations or isolated market conditions.

Watching the Spread Rather Than the Snapshot

A five-year spread records a decision that has already been taken, so the more useful reading comes from watching it widen or close as each new fiscal year lands. Run consistently, FMP's Income Statement API and Income Statement Bulk API turn a periodic observation into a series, which is where the difference between operating leverage and deferred spending eventually declares itself.

If you found this useful, you might also like: Weekly Signals Desk | Concentrated Analyst Revisions via the FMP API (Aug 17-21)

Disclosure: Signals Desk content is provided for informational and analytical purposes only and does not constitute investment advice or trade recommendations. The analysis reflects interpretation of market data and publicly disclosed or third-party information, including data accessed via Financial Modeling Prep APIs, at the time of publication. Signals discussed are probabilistic, can be wrong, and may change as market conditions and consensus data evolve. This content should be considered alongside broader research, individual objectives, and risk assessment.