Barney introduction from a Strategy Experiment Canvas

You are turning a field-experiment idea into the introduction of a theory/empirical paper, following the exact mechanical recipe from Jay Barney's Academy of Management Review Editor's Comments (2018, Vol. 43, No. 3, pp. 345–348), "Positioning a Theory Paper for Publication." Do not freelance the structure — follow it precisely. This is a recipe, not inspiration.

Step 1 — Get the canvas input

The input is a field-experiment idea filled out in the Strategy Experiment Canvas (Sharique Hasan, Hyunjin Kim, and Rembrand Koning). It has two pages.

Page 1 — the experiment story:

1. Setup/motivation for strategy audience — how the project is framed for a strategy research audience specifically.
2. Biz Challenge / Friction — the applied problem or friction the intervention addresses.
3. Your insight — the core theoretical insight driving the intervention.
4. Solution (your intervention) — the treatment being tested.
5. Null (what's the right comparison?) — the counterfactual/control condition.
6. Impact on the business — the outcome(s), and how they'll be measured.
7. Why & When will it work? — the mechanism and the boundary/moderating conditions under which the effect should (and shouldn't) hold.
8. Setting & Subjects — the empirical context and sample.

Page 2 — the model design:

• Main "Y" / Secondary "Y" / Long-term "Y" — the outcome variables at different horizons.
• B1·(X1) — the intervention/treatment effect.
• B2·(X2) — pre-treatment heterogeneity (a moderator measured before treatment).
• B3·(X1×X2) — the heterogeneous treatment effect: does the treatment effect differ by the pre-treatment moderator?
• Sample, sample size / power, estimation technique, fixed effects / standard errors / attrition — the design mechanics.
• Regression form: Y = b0 + b1(treatment) + b2(pre-treatment x) + b3(T × pre).

If the user pastes canvas content directly, use it as given. If they point to a file, Read it. If fields are thin or missing, ask the user rather than inventing experimental details — canvas content is empirical ground truth for the paper; fabricating it would misrepresent the study. In particular, do not invent a Y variable, a moderator, or a mechanism that isn't in the canvas.

Step 2 — Identify the theoretical conversation

Before writing anything, decide what literature/conversation this experiment's findings would actually contribute to. The canvas fields point to this conversation in a specific way — use it deliberately rather than reflexively picking the nearest strategy buzzword:

• Your Insight is usually the single best signal of the theoretical conversation: it states the causal/theoretical claim in the researcher's own words. Start here.
• Why & When will it work? tells you the mechanism and the boundary condition — this is the richest source for identifying which specific theoretical debate (mechanism-level or boundary-condition-level) the paper engages.
• B2/B3 (pre-treatment heterogeneity / heterogeneous treatment effect) on the design page is a formal boundary-condition test. If the canvas specifies a real X2 moderator, that moderator usually is the theoretical issue the paper resolves — treat it as strong evidence for what Paragraph 2's unresolved issue should be. A canvas with no real X2 (i.e., testing only an average treatment effect, B1 alone) points to a paper that is establishing whether an effect exists at all, which is a weaker but still valid basis for Paragraph 2 (see the acceptable-reasons list below).
• Biz Challenge/Friction and Setting & Subjects describe the phenomenon and context — useful for grounding the paper but are not, by themselves, the theoretical conversation. Per Barney, a theory paper's introduction is framed around the theory, not the phenomenon, even when the phenomenon motivated the work.

State your inferred conversation explicitly to the user before or alongside the draft (one sentence), so they can correct it if you've misread the canvas. Getting the first sentence wrong wastes everything downstream — per Barney, the first sentence is the single highest-leverage sentence in the paper.

Step 3 — Write the introduction: exactly three paragraphs

Paragraph 1 — Join the conversation

• First sentence: state, without hedging, what theoretical conversation the paper is joining. This is not a description of the Biz Challenge/Friction — it is the name of the literature/debate that Your Insight and Why & When will it work? point to.
• Rest of the paragraph (2–3 sentences): distill that conversation's main research traditions and their most important results. This is a distillation, not a literature review — do not list papers or produce a "several scholars have shown..." citation dump. Show command of the conversation in the fewest possible sentences. Draw on the Setup/motivation for strategy audience field if the user has already done some of this framing work, but tighten it to Barney's standard — most motivation write-ups are looser than a publishable first paragraph requires.

Paragraph 2 — The unresolved issue ("However...")

• First word: "However" (or a functional equivalent that performs the same pivot).
• Rest of the paragraph: identify one unresolved theoretical issue in the conversation named in Paragraph 1, and explain why resolving it matters. Acceptable reasons (per Barney): unarticulated empirical implications of the theory; unexamined boundary conditions or assumptions; unexplored implications for another theory. Unacceptable reason: "no one has studied this before" — that alone never justifies a paper.
• Preferred move: if the canvas specifies a real B2/B3 moderator (pre-treatment heterogeneity), build this paragraph around the boundary condition it tests — e.g., "However, existing theory does not specify whether/when this effect holds for [the X2 moderator's substantive meaning], even though [reason this boundary condition matters theoretically, not just empirically]." This is the single strongest, most Barney-native move this canvas supports, because a heterogeneous treatment effect is a boundary-condition test by construction.
• If there is no meaningful X2 (average-effect-only design), instead ground the unresolved issue in Your Insight versus what the received theory currently implies — i.e., the theory hasn't yet had its core empirical implication tested in a setting where it can be causally identified, and explain why that gap matters theoretically (e.g., prior evidence is observational/confounded, or the theory's predicted mechanism has testable implications that haven't been isolated).

Paragraph 3 — The paper's answer

• First sentence: begins "The purpose of this paper is..." (or "The purpose of this study is..."), followed immediately by a single, simple, declarative research question. One research question only — if the draft implies two, force a choice. Draw this from Solution (your intervention) + Null (what precisely is being compared) + Main "Y".
• Rest of the paragraph: preview the answer to that question in one sentence (use Impact on the business / Main Y and, if central to Paragraph 2, the B3 heterogeneity result), then 2–3 short sentences on the most important implications. Secondary/Long-term Y and Setting & Subjects belong here only if they carry a real implication — otherwise leave them for the paper's discussion section, not the introduction. Do not enumerate a list of separate contributions ("this paper also contributes to X, Y, and Z") — that undercuts the central contribution instead of reinforcing it.

Step 4 — Enforce Barney's constraints on the whole introduction

• Length: the full three-paragraph introduction should land around 1.5 double-spaced manuscript pages (roughly 350–450 words). Draft first, then cut ~20%, then cut ~20% again. Report the approximate word count to the user.
• One research question: if you can't compress the paper's aim into one declarative sentence, that's a signal the theoretical issue in Paragraph 2 wasn't specified narrowly enough — go back and narrow it, don't paper over it in Paragraph 3.
• No exhaustive lit review, and no design mechanics (sample size, power, estimation technique, fixed effects) anywhere in these three paragraphs. Those belong in the paper's methods section, not the introduction — even though they're on the canvas, they are not introduction material.
• No hedging in the first sentence of Paragraph 1 or Paragraph 3. Barney is explicit that these are stated "without subtlety."

Step 5 — Deliver in chat first

Output in the conversation, before producing any file:

1. The one-sentence statement of the inferred theoretical conversation (for the user to confirm or correct), noting which canvas field(s) it came from.
2. The three-paragraph introduction itself, clearly separated.
3. Approximate word count against the ~350–450 word / 1.5-page target.

If canvas fields were thin, missing, or you had to infer the theoretical conversation or the boundary-condition argument with real uncertainty, say so plainly rather than silently smoothing over the gap. Let the user confirm or redirect before compiling a PDF — the conversation stated above is the highest-leverage thing to get right, and a PDF of the wrong framing just wastes a compile.

Step 6 — Compile to PDF (when asked, or once the draft is confirmed)

Build a short standalone manuscript-style PDF, not a full paper:

1. Write a .tex file (12pt, article, 1in margins, times, setspace) with:
   • A short title stating the paper's likely research question (not "Barney intro").
   • A one-line italic note under the title: what this document is and which canvas it came from (e.g., "Draft Introduction --- generated via the barney-intro skill from a filled Strategy Experiment Canvas").
   • The three paragraphs inside a doublespace environment, matching Barney's own manuscript-page convention — this is not a stylistic choice, it's how the 1.5-page target is actually measured.
   • A small footer (\footnotesize, centered, below a \vfill) with: the word count against the ~350–450 word target, the full citation for Barney (2018) Editor's comments: Positioning a theory paper for publication, AMR 43(3): 345–348, the canvas source filename, and a flag that citations in the body are unverified / need to be added before submission (unless the user supplied real citations).
2. Compile with tectonic <file>.tex (a portable, dependency-free LaTeX engine — no system TeX install required). If no tectonic binary is available on PATH or in the session's scratch directory, download the right release for the platform from https://github.com/tectonic-typesetting/tectonic/releases/latest into the scratch/bin directory before compiling, rather than asking the user to install LaTeX themselves.
3. Verify the compiled PDF: check the page count (expect ~2 pages including the title block and footer — the three-paragraph body itself should be close to 1.5 pages). If the body alone is pushing past ~1.5 pages, that's a real signal the draft needs the Step 4 20%-cut treatment, not just tighter LaTeX spacing.
4. Save the .tex and .pdf together, named after the paper's likely topic (not generically "output" or "draft") in the same directory as the source canvas unless the user specifies otherwise.