Investment Thesis Template

The output of complete analytical work. The standard structure for thesis.md inside every thesis bundle (see thesis-bundle-standard). Length: typically 3-8 pages of substantive analysis, plus the standardized frontmatter and sections below.

The thesis file reflects the current state of AlphaSteve's view on the name. When refreshed, the prior state moves to revisions/ per the bundle standard. Layered refresh history does not belong in thesis.md.

Upstream

The full thesis is not the agent's first contact with a name. The pipeline is:

1. Daily scan surfaces a candidate (per the funnel-top screens with novelty discipline).
2. First-read (per first-read-standard) triages the candidate in 30-60 minutes with one of four decisions: continue / pass / shelve-with-trigger / avoid. The first-read identifies the hinge question and the top risk; it does NOT triangulate valuation or produce a verdict.
3. Only continue first-reads reach this full-thesis stage. The thesis-builder picks the oldest continue from the Backlog and writes the full bundle per this template.

The first-read is the upstream filter — it ensures the full thesis is only invested in names that have already passed the four selection criteria (MoS visible, moat nameable, structural setup, circle of competence). Don't repeat the first-read's work; build on it.

Autonomy

Per the 2026-05-25 autonomy doctrine (Rules), the agent autonomously writes the full thesis and assigns the verdict. The user reviews the completed bundle post-hoc and can override the verdict via a revision file (see thesis-bundle-standard refresh protocol). The agent is not waiting for pre-approval at this stage.

Mandatory vs optional sections

Standardized frontmatter

---
tags:
  - thesis
ticker: <TICKER>
name: <Company Name>
sector: <GICS sector>
gics_industry: <GICS industry group>
verdict: <buy / pass-with-trigger / pass / avoid>
verdict_date: <YYYY-MM-DD>
first_published: <YYYY-MM-DD>
last_refresh: <YYYY-MM-DD>
methodology: <pure-klarman / greenwald-modified / buffett-modern / mauboussin-compounder>
central_value: <number>
range_low: <number>
range_high: <number>
re_engagement_trigger: <number>
position_size_pct: <number, 0 if no position>
pipeline_stage: <workbench / watchlist / portfolio / closed>
moat_quality: <exceptional / wide / narrow / narrow-contested / none>
gating_tests_passed: <0 / 1 / 2 / 3>
artifacts:
  - thesis.md
  - shadow-matrix.md
  - consensus-gap.md
  - calibration.md
  - dashboard.html
  - model.xlsx
---

Optional frontmatter fields when applicable: buy_date, exit_date, exit_price.

Optional refresh note

If the thesis has been refreshed since first publication, include this single-line note immediately below the heading:

Refreshed YYYY-MM-DD — [one-line reason]. Prior version: YYYY-MM-DD-short-name. (Obsidian resolves bare filenames; do not include the revisions/ path prefix.)

That is the only mention of refresh history in the thesis body. Everything below reflects current state.

[Company] — [Verdict] — [Last refresh date]

1. Bottom line

One paragraph summary:

• The recommendation
• The central value estimate
• The current price
• The implied margin of safety
• The variant perception (in one sentence)
• The position size implication

2. The business

What they do

One paragraph in plain English. What products / services, to whom, for what.

How they make money

One paragraph. The economic engine — revenue source, gross margin profile, fixed cost structure, capital intensity.

Segments and geographic mix

Brief table or summary. Where the revenue and profit come from.

Recent history

3-5 year trajectory of revenue, margins, returns. What changed and why.

3. Structural assessment

Moat

• Type: [from 7 Powers / Greenwald taxonomy]
• Mechanism: specific reason competitors can't replicate
• Evidence in numbers: ROIC, margin stability, retention, etc.
• Power: ROIC–WACC spread
• Width: fade horizon
• Decay risks

Bottlenecks

• Where the firm sits in the value chain
• Bottlenecks upstream (dependencies)
• Bottlenecks owned (rent capture)
• Channel position downstream

Industry structure

• Five forces snapshot
• Capital cycle position
• Concentration / competitive set

4. Variant perception

What consensus believes

• Specific claims about growth, margins, returns
• Source: sell-side consensus, reverse DCF, etc.

What you believe differently

• Specific points of divergence
• Evidence and reasoning
• Type (forecast / structural / time horizon / quality / catalyst)

Steelman of consensus

• The best case for consensus being right
• Why you reject it

Gap-closer

• What event or time horizon resolves the variant
• Timeline

5. Valuation triangulation

Method 1 — EPV (no growth)

• Normalized NOPAT
• WACC
• EPV
• Per-share value

Method 2 — Asset / liquidation value

• Net asset value / liquidation value
• Per-share floor

Method 3 — DCF (cross-check)

• 10-year forecast inputs (briefly)
• Implied per-share value
• Reverse-DCF: what price implies for growth and margins

Method 4 — Comparables

• Comp set
• Multiples comparison
• Position within / outside the range

Method 5 — PMV / Replacement value (where relevant)

• Strategic buyer view
• Replacement cost anchor

Triangulation

• Central value estimate
• Range (with confidence interval)
• Width of range (assessment confidence)

Margin of safety

• Required given confidence level
• Current price vs. trigger
• Position relative to range

Shadow valuation matrix

Mandatory section. See shadow-valuation-matrix for methodology.

Record what each candidate methodology would say for this name. The chosen methodology's row is bolded; the others are shadows recorded for the methodology-calibration forward-test.

• Spread: $A to $D (ratio $D / $A)
• At-current-price verdict spread: [do the four agree on action, or split?]
• Where the spread comes from: [one sentence on which inputs drive the dispersion — growth-value credit, runway band, MoS requirement]

6. Consensus & gap

Mandatory section. See consensus-benchmarking for the methodology.

Consensus map

Table comparing our number to the street, with sources and methodologies. Pull at thesis publication and again at each calibration checkpoint.

Gap decomposition

Required when |gap to consensus mean| ≥ 25%.

For each input where assumptions differ materially, state our value, the consensus's apparent value, and one sentence on why we diverged. The honest answer is rarely "we did better research" — it is "we made different choices on a small number of assumptions."

Consensus steelman

Required when |gap| ≥ 25%, or when we disagree on direction (bull vs bear).

• The strongest single argument for consensus: [one paragraph]
• What would have to be true for consensus to be right: [bullets — specific testable assumptions]
• Evidence that would force AlphaSteve to revise toward consensus: [bullets]
• Evidence that would force consensus to revise toward AlphaSteve: [bullets]
• Which type of evidence is more likely to emerge first, and by when: [one paragraph]

Classification of the gap

One of:

• Structural — methodology choice. The gap is driven primarily by EPV vs DCF, conservative-vs-growth-credited, deep-value vs growth-at-a-reasonable-price. The gap is methodology working as designed.
• Specific — evidence-driven. The gap reflects AlphaSteve having or interpreting evidence the market hasn't yet priced. This is the rare and high-value case.
• Mixed. A combination of structural and specific. Decompose what portion is each.

The classification feeds methodology-calibration. A run of "structural" gaps where the kit subsequently underperforms is the leading indicator of a methodology bias to investigate.

7. Quality and management

Capital allocation track record

• Scorecard across 5-10 years
• Specific decisions and their outcomes
• Grade

Incentive alignment

• Compensation structure
• Insider ownership
• Board quality

Communication quality

• Language style, specificity, mistakes acknowledged

Governance flags

• Any present and material

8. Macro and cycle context

Rate environment

• Current regime and assumed path
• Sensitivity of this name to rates

Cycle position

• Where the industry is in its capital cycle
• Implications for forward expectations

FX and country

• Exposure
• Hedging policy
• Risks

Regulatory backdrop

• Current state
• Active issues

9. Risk and pre-mortem

Top 3 bear case scenarios

For each:

• The scenario (specific)
• Probability estimate
• Plausible downside

Permanent loss scenarios

• Specific events that could go to zero
• Probability and downside

Kill criteria

• 3-5 specific, observable, falsifying conditions
• What you'll do when each fires

Tail and macro

• Tail-risk scenarios specific to this name
• Macro events that would impact

10. Position sizing

Conviction tier

• Core 1 / Core 2 / Mid / Probe
• Justification

Initial size

• Target percentage
• Build pattern (entry, add points)

Sell discipline

• Trim and exit price triggers
• Time horizon

Portfolio fit

• Correlation with existing positions
• Sector / country / risk-driver exposure check

Hard rule compliance

• No single position > X%
• No single industry > Y%
• No governance red flags above threshold

11. What you don't know

A short section enumerating the open questions, ranked by how much they would change the thesis if resolved.

For each:

• The question
• Why you can't answer it now
• How / when you could
• Implication for confidence

12. The next steps

• Date of next review
• Specific developments to watch for
• Continued research items

How to use this template

For a buy thesis

All sections required. Length 6-10 pages typically.

For a pass / hold conclusion

Sections 1-7 required, sections 8-10 abbreviated. Length 2-4 pages.

For a probe position

Sections 1-5 required, others as relevant. Length 2-3 pages.

For an update

Reference prior thesis; document what's changed. Length 1-2 pages.

What this template forces

• A concrete recommendation (no fence-sitting)
• A specific variant perception (no buying retail)
• Multiple valuation methods (no single-number false precision)
• Specific kill criteria (no continuous goalpost-moving)
• Position size logic (no implementation-without-thought)
• Honest acknowledgment of what you don't know

If the thesis cannot be filled in with substance, the work is not done. Producing a thesis below this standard is one of the most expensive errors in equity research.

Linked

• 05-decision-framework — the gate structure
• margin-of-safety-pricing — sizing logic
• variant-perception — section 4
• pre-mortem-and-inversion — section 8
• permanent-capital-loss — section 8
• position-sizing-kelly — section 9