Direct Answer
The Swoopr Research Loop is a repeatable eight-step stock research process: define the question, understand the business, measure its financial trajectory, compare it with peers, value it with scenario ranges, stress-test the thesis, document the evidence, and review when new information arrives. Each step has a tool or guide on this site, and every calculation those tools perform is defined on this page.
Key Takeaways
- The Swoopr Research Loop is an eight-step process: define, understand, measure, compare, value, stress-test, document, review.
- Every metric the Workbench tools accept has one documented formula on this page, so a number always means the same thing everywhere.
- Missing data is reported as missing. No tool on this site substitutes zero for an absent value.
- Valuation is done in scenario ranges with visible assumptions, never as a single point estimate.
- Nothing in the Loop produces a recommendation or a score; the output is documented, stress-tested research.
What Is the Swoopr Research Loop?
The Swoopr Research Loop is a repeatable eight-step stock research process: define the question, understand the business, measure its financial trajectory, compare it with peers, value it with scenario ranges, stress-test the thesis, document the evidence, and review when new information arrives. Each step has a tool or guide on this site, and every calculation those tools perform is defined on this page.
The Loop exists because unstructured research fails in predictable ways: the thesis is never written down, contrary evidence is never collected, valuation happens first and bends everything after it, and six months later nobody can say why the position exists. A documented process does not make the conclusion right, but it makes the reasoning checkable, which is the part you control.
The Eight Steps in Detail
- Define. One research objective stated as an answerable question, the time horizon it applies to, and what evidence would change your mind. A question that cannot be settled by evidence is not a research objective.
- Understand. The business model in plain language: what is sold, to whom, how it is priced, and which two or three variables actually move revenue. Then the competitive context: market structure, key competitors, and where pricing power sits.
- Measure. The financial trajectory from primary sources: multi-year revenue and EPS direction, margin structure, cash generation, balance sheet strength, and share-count change. Every figure gets the period it covers and the filing it came from.
- Compare. The same metrics against a defensible peer group and against the company's own multi-year range. A 20% operating margin means little until you know whether peers earn 10% or 35%, and whether the company itself earned 25% two years ago.
- Value. Scenario ranges from transparent methods: multiples against peers and history, discounted cash flow, or owner-earnings approaches. Each scenario states the assumption that drives it, and the sensitivity of the result to that assumption is reported alongside the result.
- Stress-Test. The risk register: concrete downside cases, the strongest current evidence against the thesis, catalysts with dates where known, and explicit thesis-break conditions: observable events that would prove the thesis wrong.
- Document. The thesis in two or three sentences, its key assumptions each stated so it can be checked, the questions still open, and every source with an as-of date. Facts, assumptions, and estimates are labeled as what they are.
- Review. The triggers that reopen the document: the next 10-Q or 10-K, a named catalyst resolving, or any thesis-break condition firing. Sources older than 180 days are flagged stale and re-verified rather than trusted.
How Is Each Metric Calculated?
These are the definitions behind every metric in the watchlist stock screener and the Measure step of the Research Workbench. The worked examples use Example Manufacturing Co., a fictional company invented for this page, with FY2025 figures: revenue $530M, gross profit $180.2M, operating income $74.2M, net income $47.7M, operating cash flow $85M, capital expenditures $26.7M, EBITDA $95M, total debt $120M, cash $80M, interest expense $9M, average shareholders' equity $310M, diluted shares 49.5M (50.0M in FY2024), share price $16.80.
Growth: year-over-year change and CAGR
Year-over-year growth is the latest period divided by the prior period, minus 1. Example Manufacturing's revenue grew from $500M in FY2024 to $530M in FY2025: 530 / 500 - 1 = 6.0%.
Compound annual growth rate (CAGR) smooths growth over several years: (ending value / beginning value)^(1 / years) - 1. Revenue of $400M in FY2021 reaching $530M in FY2025 is (530 / 400)^(1/4) - 1 = 7.3% per year. CAGR hides interim volatility by construction: a business that fell 30% and recovered shows the same CAGR as one that grew steadily. EPS growth is calculated the same way on diluted EPS and is not meaningful when either year's earnings are negative; the tools report it as unavailable in that case. Try the calculation on the CAGR calculator.
Margins: gross, operating, net, and margin change
Each margin is a profit line divided by revenue for the same period. For Example Manufacturing in FY2025: gross margin 180.2 / 530 = 34.0%; operating margin 74.2 / 530 = 14.0%; net margin 47.7 / 530 = 9.0%. Margin change is expressed in percentage points: FY2024 operating margin was 66 / 500 = 13.2%, so FY2025's 14.0% is a change of +0.8 points. Margins are comparable within an industry, much less so across industries; a grocer and a software company have structurally different gross margins.
Cash flow: free cash flow and FCF margin
Free cash flow (FCF) is operating cash flow minus capital expenditures: 85 - 26.7 = $58.3M. FCF margin divides that by revenue: 58.3 / 530 = 11.0%. FCF differs from net income because of non-cash charges, working-capital swings, and the gap between depreciation and actual capital spending; a company can report profits while consuming cash, which is why the Measure step asks for both. See the free cash flow guide for the line items involved.
Balance sheet: net debt, leverage, and interest coverage
Net debt is total debt minus cash and equivalents: 120 - 80 = $40M. Negative net debt means more cash than debt. Net debt to EBITDA scales that by earnings capacity: 40 / 95 = 0.4x. Interest coverage is operating income divided by interest expense: 74.2 / 9 = 8.2x, meaning operating profit covers interest about eight times over. Coverage is not meaningful when interest expense is near zero, and EBITDA-based leverage flatters companies whose depreciation reflects real, recurring capital needs.
Returns on capital: ROE and ROIC
Return on equity is net income divided by average shareholders' equity: 47.7 / 310 = 15.4%. ROE rises mechanically with leverage, so it is read together with the balance sheet. Return on invested capital uses net operating profit after tax (NOPAT) over invested capital. With a 21% illustrative tax rate, NOPAT is 74.2 x 0.79 = $58.6M, and against invested capital of $430M (debt of $120M plus equity of $310M) ROIC is 58.6 / 430 = 13.6%. Definitions of invested capital vary, most often in how much cash is excluded; whichever definition you use, use it consistently across every company you compare.
Dilution: share count change
Share count change is the latest diluted share count divided by the prior year's, minus 1. Example Manufacturing went from 50.0M to 49.5M diluted shares: 49.5 / 50.0 - 1 = -1.0%, a net buyback. Positive values mean dilution: each existing share owns a smaller slice of the same business. Persistent dilution of a few percent per year compounds into a large drag on per-share results, which is why the screener includes a dilution watch preset.
Valuation multiples: P/E and EV/EBITDA
Price to earnings is share price over diluted EPS. Example Manufacturing's EPS is 47.7 / 49.5 = $0.96, so at $16.80 the P/E is 16.80 / 0.96 = 17.4x. P/E is undefined when earnings are negative. EV/EBITDA uses enterprise value: market cap (49.5 x 16.80 = $831.6M) plus net debt ($40M) gives EV of $871.6M, and 871.6 / 95 = 9.2x EBITDA. EV-based multiples let you compare companies with different debt loads on the same footing. A multiple is a comparison device, not a verdict: it needs a peer set and the company's own history to mean anything, which is what the Compare step supplies.
Discounted cash flow and sensitivity ranges
DCF values a business as the sum of its projected free cash flows, each discounted to present value: a cash flow received in year n is divided by (1 + discount rate)^n. $100 received three years from now at a 10% discount rate is worth 100 / 1.331 = $75.13 today. A full DCF projects cash flows over an explicit period, adds a terminal value, and discounts everything back; the arithmetic is mechanical, and the result is only as good as the growth and discount-rate assumptions feeding it.
That is why the Value step requires ranges, not points: run the same model at low, base, and high growth and at more than one discount rate, and report the spread. If the low and high scenarios produce wildly different values, the honest conclusion is that the valuation is highly sensitive to assumptions you cannot verify, and that finding belongs in the research document. The DCF valuation calculator runs these scenarios with every assumption visible.
How Is Missing Data Handled?
A missing number is reported as missing, everywhere, without exception. The screener labels a company that lacks a needed metric as "insufficient data" and lists the missing fields; Workbench exports print unfilled sections as "Not yet researched"; and sources without an as-of date are listed as undated rather than assumed current. Substituting zero for an absent value is the single most common way screeners and models silently lie, because zero is a real, meaningful value for almost every metric on this page.
Some metrics are also undefined in specific, legitimate situations: EPS growth with a negative base year, P/E with negative earnings, interest coverage with near-zero interest expense, ROE with negative equity. The correct output in those cases is "not meaningful" plus a note, and the definitions above say when each applies.
Frequently Asked Questions
Why does the Research Loop put valuation so late in the process?
Because a price target chosen early becomes the goal the rest of the research bends toward. Measuring the business, comparing it with peers, and understanding its trajectory first means the valuation inputs are grounded in evidence rather than reverse-engineered from a number you already wanted. The Loop also requires scenario ranges instead of a single point estimate, which only works when the drivers behind each scenario have already been researched.
What happens when a metric cannot be calculated?
It is reported as missing, never replaced with zero or a guess. In the screener a company missing a needed metric is labeled insufficient data with the missing fields listed, and in Workbench exports an unfilled section is printed as not yet researched. Some formulas are also undefined in specific situations, such as EPS growth when either year's earnings are negative, or P/E when earnings are negative; in those cases the correct output is no value plus a note, not a forced number.
Where should the input numbers come from?
Primary sources first: the company's own 10-K and 10-Q filings on SEC EDGAR, its earnings releases, and its investor-relations materials. Data aggregators and broker fundamentals pages are convenient but occasionally define metrics differently or lag a filing, so any number that materially affects a conclusion should be traced back to the filing it came from, and recorded with an as-of date.
Does completing the Research Loop tell me whether to buy a stock?
No. The Loop is a documentation and quality-control process for your own research, not a decision engine. It has no scoring, no grades, and no buy, sell, or hold output. What it produces is a written thesis with its supporting evidence, the strongest arguments against it, and the specific conditions that would prove it wrong, so that whatever decision you make is based on work you can re-examine later.
What happens when two primary sources report the same metric differently?
Both figures get recorded with their source and as-of date, and the difference becomes part of the note rather than being resolved by picking one. Differences usually trace to a definition rather than an error: a fiscal period boundary, an adjusted versus reported basis, or a restatement that one source has picked up and the other has not. Documenting which definition each figure uses is what allows a later comparison to be like for like.
Why does the loop end with review rather than with a decision?
Because the process is designed to produce evidence and a stated thesis, not a verdict. A decision depends on the reader's own objectives, holdings and constraints, none of which the research process observes. Ending with review also makes the note a living document: it names what new information would change the conclusion, so the question of when to look again is answered before the note is filed rather than left to memory.
How is a peer group chosen for the compare step?
By economics rather than by index membership or sector label. Companies compared usefully share the same demand drivers, the same cost structure and roughly the same business model, which is often a narrower set than a sector classification produces. The group should be written down before the metrics are pulled, since choosing peers after seeing the numbers lets the comparison be shaped to a conclusion. Where no clean peer exists, comparing the company against its own multi-year history is the alternative.
How does the stress-test step differ from the valuation scenarios?
Valuation scenarios vary the inputs of a model to produce a range of values, so they answer how sensitive the result is to each assumption. The stress test asks a different question: what would have to be true for the thesis to be wrong, and what evidence exists against it today. One is arithmetic on the model, the other is an argument against the conclusion. A note with scenario ranges but no contrary evidence has done only half of the work.
How often should a completed note be revisited?
On the schedule the note itself sets, which is why the review step asks for a date and a trigger rather than a general intention. Periodic filings and results announcements are natural checkpoints. Beyond the calendar, the conditions listed as thesis-breaking are the real triggers: any one of them occurring is a reason to reopen the note regardless of how recently it was reviewed.