competitive-landscape-deliverable

Competitive Landscape Deliverable

Core outcome

Transform a detailed competitive landscape workbook into a board/executive-ready deliverable while preserving the raw research. Add or replace a second sheet (Executive Deliverable) that:

1. Leads with the verdict (ACQUIRE / MONITOR / PASS) per company
2. Compresses ~100 raw fields into 7–10-word executive phrases
3. Re-sorts companies by deal priority (configurable)
4. Renders with Pattern brand styling (Wix Madefor Display, Pattern blues)

Default workflow:

Source workbook (.xlsx or Google Sheets URL)
   → scripts/pull_from_sheets.py  (if URL)
   → scripts/build_deliverable.py
   → Same workbook with new "Executive Deliverable" sheet added

When to invoke

Trigger this skill when the user asks to:

• "Convert / summarize / clean up the competitive landscape sheet"
• "Make an executive view / board view / deliverable from the competitor research"
• "Add a one-pager / summary tab to the competitive landscape"
• "Make a McKinsey-style deliverable from the M&A target grid"

Source format

The Pattern competitive landscape template produced by the n8n orchestrator (L8mR6NM9u6VoVdrJ) has:

• Sheet name: Competitive Landscape
• Row 1: title Competitive Landscape Mapping
• Row 4: company headers (column C onward, alphabetical or insertion order)
• Column A: section labels (1. Entity identity, 2. Market taxonomy, etc.)
• Column B: field names (Company Name, Legal Entity Name, ...)
• Rows 5–155: research data, organized into 9 sections
• Cell format: "Rating value — McKinsey-style rationale" (em-dash separator)

See references/template-fields.md for the complete field map.

Robust detection rule: never hardcode rows. Scan column B for known field names (e.g., Company Name, Total Moat Score) and anchor relative positions from there. The engine script does this automatically.

Use the engine script

The mechanical work — read source, build sheet, apply formatting — happens in scripts/build_deliverable.py. Do not generate XLSX/XML by hand. Invocation:

py scripts/build_deliverable.py \
  --source "<path-to-source.xlsx>" \
  --output "<path-or-blank-for-in-place>" \
  [--sort-by priority|fit|moat|source]

The script:

1. Loads the source workbook with openpyxl, preserving the original sheet
2. Detects header row, section column, field column dynamically
3. Reads each cell as (rating, rationale) by splitting on em-dash
4. For each output row, runs the synthesis mapping (see below) and writes a 7–10-word phrase derived from the rating + rationale
5. Builds the Executive Deliverable sheet with verdict block + 8 grouped sections
6. Applies Pattern brand styling
7. Saves in place (or to --output)

If the synthesis logic needs new mappings or richer phrasing, edit build_deliverable.py rather than re-implementing the logic per invocation.

Google Sheets ingestion

For a Google Sheets URL, run:

py scripts/pull_from_sheets.py --url <sheets-url> --out <local.xlsx>

This uses the Google Sheets v4 API (set GOOGLE_SHEETS_TOKEN env var to an OAuth access token, or it falls back to gcloud-application-default-credentials). Output is a local .xlsx snapshot ready for build_deliverable.py.

The Pattern Competitive Landscape sheet ID is 159cKsL9YEWmIoo0BTt6NelobAmAm6qan-VyOagUJuJ4.

Executive Deliverable layout

Title block (rows 1–2)

• Row 1: Competitive Landscape — Executive View (24pt, white, on #0F4761 band)
• Row 2: Concise strategic readout from detailed research (10pt, #4280F4)

Verdict block (rows 3–6)

Row	Label (col B)	Cell content
3	Verdict	ACQUIRE / MONITOR / PASS — green/amber/grey conditional fill
4	Recommended posture	7–10-word phrase from posture vocabulary
5	Score triplet	moat / fit / ma (e.g., 16 / 8.2 / 6.8)
6	Acquisition thesis	10–14-word one-liner

Freeze panes at C7 (below verdict block, after both label columns).

Section blocks (row 7 onward)

Eight sections, each preceded by a styled section header row:

1. Market taxonomy (~11 rows)
2. Product capability (~17 rows)
3. Data and compliance (~10 rows)
4. Commercial traction (~16 rows)
5. Competitive position (~16 rows)
6. Pattern fit (~12 rows)
7. M&A lens (~14 rows)
8. Pipeline operations (~12 rows)

Total ~108 rows + 6-row header = ~114-row deliverable.

See references/output-row-spec.md for the exact label list and source-field mapping.

Phrase rules — fidelity over brevity

The deliverable must be as concise as possible without losing or transforming meaning. Compression is achieved by removing filler and transitional language, NOT by truncating substantive content.

Core principle: preserve the rating, compress the rationale

Source cells follow "Rating — McKinsey rationale". Both halves carry meaning:

• The rating IS the answer for many fields (e.g., Yes, Tier B, Strong, Critical, 4). Never drop it.
• The rationale provides the evidence that makes the rating credible. Compress, but don't truncate mid-clause.

Output format by field type

Source pattern	Output format	Example
Both rating and rationale	Rating — compressed rationale	Yes — embedded across OMS, helpdesk, customer notifications
Rating only (no rationale)	Rating (verbatim)	Tier B
Numeric score + rationale	N — qualifier	4 — captures returns data across 1,200+ carriers
Rationale only (no rating)	Compressed rationale	Recently acquired by Blue Yonder August 2025
List values (Notable Customers, etc.)	First 5 items + (+N more)	Sephora, Levi's, Patagonia, Adidas, Lululemon (+12 more)
Empty source	— (em-dash)	—

Compression rules

Apply in this order:

1. Strip citations (<cite index="..."> tags from agent output)
2. Remove transitional filler (for example, notably, in particular, as evidenced by, the company, which is, etc.)
3. Remove banned hype (leading, robust, innovative, comprehensive, cutting-edge, best-in-class) UNLESS the rating itself contains it (then preserve)
4. Collapse whitespace, strip trailing punctuation
5. If still over budget: break at clause boundary (comma, semicolon) — never mid-sentence

Length targets (NOT hard caps)

• Combined cell (rating + rationale): 12–18 words typical, up to 22 acceptable
• Rating-only cells: as long as the rating itself
• Free-text pass-through fields: up to 22 words
• Hard ceiling for QC warning: 26 words (above this, the script flags for review)

These are fidelity ceilings, not aspirational caps. Going under doesn't earn points if it sacrifices meaning.

Good vs bad transformations

✅ Good (rating preserved, evidence compressed):

Source: "Partial — owns return and exchange records but synchronizes core order data from Shopify or OMS upstream"
Output: "Partial — owns return/exchange records, syncs orders from Shopify or OMS"

✅ Good (rating + rich rationale):

Source: "Strong — implementation complexity creates ~12-month migration cost; deeply embedded in helpdesk, OMS, and customer notification workflows"
Output: "Strong — embedded in helpdesk, OMS, customer notifications; ~12mo migration"

❌ Bad (drops the rating, loses verdict):

Source: "Yes — embedded across OMS, helpdesk, and customer notification workflows"
Output: "embedded across OMS, helpdesk, and notifications"        # missing "Yes"

❌ Bad (truncates mid-clause, loses meaning):

Source: "Tier B — captures order, return, and shipment data but excludes payment card data and PII routed to Stripe"
Output: "captures order, return, and shipment data but excludes"   # mid-clause cut

❌ Bad (over-compresses to the point of distortion):

Source: "Pass — Blue Yonder's August 2025 acquisition eliminates buy option while returns management represents non-core capability"
Output: "Off the table"                                            # loses the WHY

Free-text fields — pass through, don't synthesize

For fields where the source is already a list or descriptive text rather than a rated assessment, use lighter cleanup:

• Notable Customers / Key Investors / Direct Competitors / Comparable Transactions: trim to 5 items, add (+N more) if more
• Funding Raised: pass through ($208M Series D)
• Target Contact / Owners: pass through verbatim

The engine script handles this via is_passthrough_field() and trim_list().

Empty / low-confidence handling

• Truly empty source → —
• Source has rating like Unknown or Unclear → keep verbatim, don't substitute
• If the rationale starts with Likely, Reported, Validate, preserve that qualifier — it's the epistemic signal the executive needs

Field synthesis rules

The 100+ output rows are derived from raw fields per the mapping in references/output-row-spec.md. Defaults:

• Combine 2–3 raw fields into one executive label
• If primary source is empty, fall back to listed adjacents
• For numeric scores (1–5 or 1–10), pass the number through; for categorical, mine the rationale

The full mapping lives in the reference file and is implemented in build_deliverable.py:SYNTHESIS_MAP.

Sort and prioritization

By default: re-sort companies by deal priority:

1. ACQUIRE candidates (by descending strategic fit score)
2. MONITOR candidates (by descending moat score)
3. PASS candidates (by source order)

CLI flags:

• --sort-by priority (default)
• --sort-by fit — descending strategic fit score
• --sort-by moat — descending total moat score
• --sort-by source — preserve raw sheet order

Posture vocabulary

When generating the row 4 "Recommended posture" cell:

Pattern in source	Posture phrase
High fit + high priority + low complexity	Pursue active diligence this quarter
High fit + high complexity	Partner first, revisit acquisition next year
High moat + low Pattern fit	Useful benchmark, weak Pattern fit
Recently acquired	Off-market, monitor competitive impact
Low confidence	Validate thesis before leadership outreach
Strong moat + early stage	Track for inflection, no near-term action

Pattern brand styling

All formatting uses the Pattern visual system (matches Word doc standard):

Element	Font	Size	Color	Fill
Title row 1	Wix Madefor Display SemiBold	24pt	#FFFFFF	#0F4761
Subtitle row 2	Wix Madefor Display	10pt	#4280F4	none
Verdict ACQUIRE	Wix Madefor Display SemiBold	11pt	#1A5C1A	#E2EFDA
Verdict MONITOR	Wix Madefor Display SemiBold	11pt	#7A5C00	#FFF3CD
Verdict PASS	Wix Madefor Display SemiBold	11pt	#8B1A1A	#FFE0E0
Section header	Wix Madefor Display SemiBold	11pt	#0F4761	#F5F8FF
Field label (col B)	Wix Madefor Display	9pt	#000000	none
Body cell	Wix Madefor Display	9pt	#000000	alt rows #F5F8FF
Borders	thin	—	#DDDDDD	—

Cell formatting:

• Wrap text everywhere
• Top-aligned
• Column A (section): 18 width, bold
• Column B (label): 26 width, bold
• Company columns: 32 width, wrap
• Row height: 22 default, 36 for verdict block, 14 for section headers

Quality checks

Before returning the workbook, the engine script asserts:

1. Source sheet Competitive Landscape (or detected source) exists and is byte-identical to input
2. Executive Deliverable sheet exists with ≥80 rows and N+2 columns (where N = company count)
3. Company headers in deliverable match source order or sort criterion
4. Random-sample 12 cells: each is ≤12 words and contains no banned filler
5. Verdict cells contain only ACQUIRE, MONITOR, PASS, or —
6. Output saved as .xlsx (not .xlsm or .xltx)

If any check fails, the script raises and prints the specific failure.

Folder layout

competitive-landscape-deliverable/
├── SKILL.md                     (this file)
├── references/
│   ├── template-fields.md       (full Pattern field map)
│   └── output-row-spec.md       (executive deliverable row spec)
├── scripts/
│   ├── build_deliverable.py     (engine — openpyxl)
│   └── pull_from_sheets.py      (Google Sheets ingestion)
└── assets/
    └── (sample output.xlsx if available)