comps-analysis
Skill108 starsChanged 3 days agoMITView on GitHub
Builds a comparable company analysis: selects a defensible peer set with stated inclusion criteria, pulls trading multiples and operating metrics, calculates the median and quartiles, and applies the range to a target to produce an implied valuation. Use when the user asks to "build a comps table", "run comparable company analysis", "trading comps for this company", "what multiple do peers trade at", "who are the comps", or "value this against its peers". Do NOT use for a company profile — use company-profile-pack instead; deal accretion — use merger-model instead; or discounted cash flow, LBO, or precedent transaction analysis.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- JerryLiu369/awesome-quant-skillstheir own version, first seen Nov 16, 2026
- vijaysuperpositionist/skillstheir own version, first seen Nov 1, 2026
- The-Utopia-Studio/skillstheir own version, first seen Oct 13, 2026
- anthropics/financial-services-pluginstheir own version, first seen Oct 8, 2026
- davidalmeida90/finance-agent-kittheir own version, first seen Sep 4, 2026
- anthropics/financial-servicestheir own version, first seen Jul 22, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- anthropics/financial-services-plugins★ 41ksame name
- anthropics/financial-services★ 41ksame name
- The-Utopia-Studio/skills★ 6same name
- JerryLiu369/awesome-quant-skills★ 5same name
- davidalmeida90/finance-agent-kit★ 1same name
- vijaysuperpositionist/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository293 other files it gives its agents.
- .github/copilot-instructions.mdCopilot instructions
- accessibility-passSkill
- acroform-writerSkill
- action-items-todoSkill
- agent-evaluation-designerSkill
- agent-harness-explorerSkill
- agent-performance-triageSkill
- agent-rai-impact-assessmentSkill
- agent-red-teamSkill
- ai-demo-assistantSkill
- ai-first-process-redesignSkill
- ai-usecase-assessmentSkill
- anonymised-case-study-writerSkill
- answer-researchSkill
- b2b-outreach-suiteSkill
- blog-content-auditorSkill
- blog-post-structure-passSkill
- booth-loop-videoSkill
- brand-template-enforcerSkill
- brand-voice-passSkill
- breathing-roomSkill
- browser-uat-analystSkill
- business-case-composerSkill
- business-osSkill
- call-for-speakers-digestSkill
- campaign-deck-builderSkill
- cim-teardownSkill
- company-profile-packSkill
- coverage-meeting-prep-debriefSkill
- dd-checklistSkill
- document-diligenceSkill
- earnings-noteSkill
- ic-memoSkill
- merger-modelSkill
- pitch-deck-skeletonSkill
- case-preparation-briefSkill
- case-triageSkill
- chart-builderSkill
- classic-text-adventureSkill
- coding-rationaleSkill
- document-deriveSkill
- handoff-draftSkill
- note-contextSkill
- quality-gap-checkSkill
- clipchamp-videoSkill
- clt-hr-assistantSkill
- commenting-contentSkill
- competitive-battlecard-builderSkill
- competitive-strategySkill
- conditional-chat-reminderSkill
- conference-session-abstract-packSkill
- content-quality-auditorSkill
- copilot-agent-advisorSkill
- copilot-studio-adaptive-card-builderSkill
- copilot-studio-agent-testSkill
- copilot-studio-harness-pickerSkill
- copilot-studio-knowledge-readinessSkill
- copilot-studio-mcp-readinessSkill
- copilot-studio-topic-blueprintSkill
- cowork-consumption-advisorSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 4 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 70
Check input/ and sharepoint_onedrive-SearchDrive for an existing peer set or house screen first — many firms have one, and diverging from …
- Distinctive: line 124
Run these against the finished table.
- Very specific: line 144
Use the xlsx skill: a summary tab with the implied range, a comps table tab, and a sources-and-criteria tab recording the screening rule …
- Distinctive: line 154
Keep inputs visually distinct from calculations — inputs in blue, formulas in black, cross-tab links in green is the convention a banker …
- Distinctive: line 159
Do not build the whole table and then present it.
- Distinctive: line 226
Keep derived cells as live formulas.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Comparable Company Analysis
Builds a trading comps set where the peer selection is defensible, the multiples are calculated consistently, and the implied range carries its assumptions in the open.
When NOT to Use
- Intrinsic valuation from cash flows — outside this skill's scope.
- Leveraged buyout returns analysis — outside this skill's scope.
- Precedent M&A transactions — a different dataset with control premiums
- Post-results company research with no multiples — use earnings-note instead.
- Pre-results research or initiating coverage — outside this package; request a verified host research destination or identify the workspace administrator as dependency owner.
Related Skills
Hand off rather than overlap:
- Cash flow valuation and terminal value — outside this skill's scope.
- Sponsor returns and capital structure — outside this skill's scope.
- Investment committee recommendation — use ic-memo instead.
- Company profile or tear sheet — use company-profile-pack instead.
- Rendering metrics as a table or card — use render-ui instead.
- Demand and supply planning review — outside this skill's scope.
- Contract clause comparison — outside this skill's scope.
- Budget or target consolidation — outside this skill's scope.
Workflow
1. Define the target and the screening criteria
Establish the target company and the basis for the peer set before naming any peers. State the criteria explicitly — industry and sub-sector, size range, geography, growth profile, margin profile, and business model.
Peer selection is the entire analysis. A comps table whose selection rule is unstated is not reviewable, so write the rule down first and let the names follow from it.
State the question the table has to answer, because it decides which columns earn their place:
| Question | Lead with | Leave out |
| Which company is undervalued? | EV/Revenue, EV/EBITDA, P/E, market cap | Operating detail, growth metrics |
| Which is most efficient? | Gross, EBITDA and FCF margin, asset turnover | Size metrics, absolute dollars |
| Which is growing fastest? | Revenue growth, EBITDA CAGR, customer growth | Margin metrics, leverage |
| Which generates the most cash? | FCF, FCF margin, FCF conversion, capex intensity | EBITDA, P/E |
2. Build the peer set
Propose 5-10 peers, each with a one-line justification tied to the stated criteria. Mark any peer that fails part of the criteria as an outlier candidate rather than quietly dropping it.
Check input/ and sharepoint_onedrive-SearchDrive for an existing peer set or house screen
first — many firms have one, and diverging from it needs saying.
3. Choose the metrics deliberately
The 5-10 rule: five operating metrics (revenue, growth, two or three margin and efficiency measures) and five valuation metrics (market cap, enterprise value, three multiples). Ten columns tell the story; past fifteen you are adding noise, not rigor. Edit ruthlessly.
Sector determines which metrics are meaningful and which are actively misleading:
| Sector | Must have | Do not use |
| Software / SaaS | Revenue growth, gross margin, Rule of 40 | Asset turnover, inventory metrics |
| Manufacturing / industrials | EBITDA margin, asset turnover, capex/revenue | Rule of 40, SaaS metrics |
| Financial services | ROE, ROA, efficiency ratio, P/E | Gross margin, EBITDA — not meaningful for banks |
| Retail / e-commerce | Revenue growth, gross margin, inventory turnover | Heavy R&D or capex metrics |
Optional additions where the data supports them — ARR, net dollar retention and CAC payback for SaaS; backlog and inventory turns for industrials; net interest margin for financials; same-store sales for retail.
4. Gather the data
Source priority: user-supplied data, then internal files, then any connected market-data provider,
then host-web_search for public filings and current prices.
Label every figure with its source and date. Where market data comes from a web search rather than a data feed, say so — a comps table sourced from search results and one sourced from a terminal are not the same artifact, and the reader must know which they hold.
Record the source as you capture each figure, never at the end. A deferred source note becomes a missing one. Cite the specific location, not the publication: "FY24 10-K, page 42, line item 'Total Revenue'" or "street consensus as of 14 Sep 2026", not "annual report". Where a figure is an assumption rather than a retrieved fact, record the reasoning in the same place — "EBITDA margin assumed at the peer median; company does not disclose" — so a reader can tell the two apart.
For each company capture: share price, diluted shares, market cap, net debt, enterprise value, revenue, EBITDA, EBIT, net income, and growth and margin metrics for the periods in scope.
5. Calculate multiples consistently
EV/Revenue, EV/EBITDA, EV/EBIT, and P/E, on a stated basis — last twelve months or forward, but the same basis for every company in the table. Mixing LTM for one peer and forward for another is the most common error in a comps set.
Report median and quartiles, not just the mean — one outlier drags a mean badly at this sample size. Show the target's own multiples against the set.
Never average a set of percentages or multiples as if they were levels — margins, growth rates and multiples are ratios, and their median is the honest central measure.
6. Sanity-check before the range goes anywhere
Run these against the finished table. Each one catches a real error class:
- Margin hierarchy diagnostic — compare gross, EBITDA and net margins on consistent definitions. Exceptions can arise from non-operating income, tax benefits or line-item classification; investigate and explain them rather than declaring an error by definition.
- Multiple plausibility — EV/Revenue typically 0.5-20x (varies widely by sector), EV/EBITDA typically 8-25x (fairly consistent across sectors), P/E typically 10-50x depending on growth. A figure outside its band is either a genuine outlier worth naming or a bad input.
- Growth-multiple relationship — higher growth normally carries a higher multiple. A fast-growing peer on the lowest multiple in the set needs an explanation, not a footnote.
- Numerator and denominator agree — enterprise value over an equity metric, or market cap over an enterprise metric, is the error that silently invalidates a whole column.
- Period alignment — the numerator's date and the denominator's period must match; LTM revenue under a current EV is right, quarterly revenue under it is not.
7. Apply the range and build the output
Apply the peer median and quartile multiples to the target's metrics to derive an implied enterprise value, equity value, and per-share range.
Use the xlsx skill: a summary tab with the implied range, a comps table tab, and a
sources-and-criteria tab recording the screening rule and every figure's provenance. Confirm the
file exists before reporting it.
Every derived cell is a live formula, not a pasted result. Margins, multiples, medians and quartiles reference their input cells so the table moves when a price or an estimate is corrected. The only hardcoded values are raw inputs — price, share count, revenue, EBITDA — and each of those carries the source note from step 4. A pasted-in margin is a silent bug waiting for the next update.
Keep inputs visually distinct from calculations — inputs in blue, formulas in black, cross-tab links in green is the convention a banker will expect.
8. Check in as you go, not at the end
Do not build the whole table and then present it. Stop at each of these and confirm before continuing:
- After the peer set — show the names and the criteria each one satisfies, before pulling data
- After the raw inputs — show the input block with sources and periods, before building formulas
- After the operating metrics — show the calculated margins and growth, before adding multiples
- After the multiples — show them against the statistics block, before applying the range
A wrong basis caught at the input block costs a minute; the same error found after the implied range is built means rebuilding everything downstream of it.
Example
Peer set - criteria: enterprise software, $500M-$3B revenue, North America,
15-30% growth, positive EBITDA. 6 peers; 1 excluded as an outlier.
EV/Rev EV/EBITDA P/E Growth EBITDA margin
Peer A 4.2x 18.1x 31.2x 22% 23%
Peer B 3.1x 14.6x 26.8x 17% 21%
Peer C 5.8x 24.3x 44.1x 29% 24%
Peer D 2.9x 13.2x 22.4x 15% 22%
Peer E 4.6x 19.8x 35.0x 24% 23%
---
Median 4.2x 18.1x 31.2x 22% 23%
Q1 / Q3 3.1x-4.6x 14.6x-19.8x
TARGET 3.4x 15.9x 27.1x 19% 21%
(excluded: Peer F at 11.2x EV/Rev - 61% growth, outside the criteria)
IMPLIED (median applied) EV $1.26B Equity $1.09B $27.40/share
IMPLIED (Q1-Q3 range) $20.10 - $29.80/share vs current $22.60
Basis: LTM for all companies. Prices as of 17 Aug 2026 (web-sourced).When Things Fail
- If fewer than 4 credible peers exist — say so and state that the sample is too small for a reliable median rather than padding with poor fits.
- If a metric is unavailable for a peer — leave it blank and note it; never interpolate a multiple from the others.
- If market data cannot be retrieved — build the structure with clearly marked placeholders and state that the table is unpriced.
- If the target's own basis differs from the peers' (different fiscal year end, different reporting standard) — flag it prominently; it can invalidate the comparison.
- If a sanity check fails — fix the input or name the peer as a genuine outlier with the reason. Never leave a broken margin hierarchy or an out-of-band multiple sitting unremarked in a table someone else will read as clean.
- If the data is stale — state the age. An undisclosed old price is worse than a disclosed one.
- If a tool errors — retry once, then continue and name what is missing.
Output Format
A workbook in output/ with the implied range on the first tab; an inline table when the user just
wants the numbers. Always state the multiple basis and the price date.
Guardrails
- Never fabricate a peer's multiple, price, or financial metric. Blank and noted beats invented.
- Never present a peer set without its selection criteria — an unstated rule is not reviewable.
- Never mix multiple bases within one table; state the basis explicitly.
- Never leave a hardcoded input without its source note, and never defer the note to the end — it is written as the figure is captured or it does not get written.
- Never mix an enterprise-value numerator with an equity denominator, or a numerator and denominator covering different periods.
- Date-stamp all market data and label web-sourced figures as such.
- Report median and quartiles alongside any mean; disclose exclusions and why.
- Keep derived cells as live formulas; the only hardcoded values are raw inputs.
- Cap the table at ten columns unless the user asks for more; more metrics is not more insight.
- This is analytical support, not investment advice — state that on the output.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

