Deckmaker
workbook · 1 source(s)

ACCESS Market Analysis Ready

Audit this market-sizing + ROI model before it goes near a leadership deck.from seed (example project) · 2026-06-04
ACCESS Market Analysis.md
STAGE 1
Source prep
done
runs: Truth Setup
STAGE 2
Structure
done
runs: Workbook spec
STAGE 3
Creation
done
runs: Workbook Doctor
STAGE 4
Verification
4 must-fix
runs: Pretty-But-Wrong
00-source-packet.md

Truth Setup — ACCESS Market Analysis

Prompt 1 output. Built before trusting the model, against the real file vault/Kobold Vault/meetings/2026-04-16 Kobold - ACCESS Market Analysis.md.

1. Source inventory

ID Source Type Date Owner Status Can support Should NOT support
S1 ACCESS Market Analysis note meeting note / model 2026-04-16 Kobold (unattributed) unknown a record of the internal sizing exercise any externally-defensible market size or revenue figure on its own

That's the catch on inventory alone: there is only one source, and it cites no upstream sources. Every number in it — market size, prices, well counts — appears without a reference, owner, or as-of date. A model whose inputs have no provenance can't be the authority for a number that travels.

2. Fact / assumption / estimate split

The note labels a block "Assumptions," but the split shows the costume runs deeper than that block.

Stated as fact (but actually unsourced):

  • Total market = 49,709 stages (P&P 27,780 / Ball Drop 6,046 / CT Sleeves 15,883).
  • Total wells 2025 = 5,548; 2026 projected = 5,709 (+2.9%).
  • Completion revenue by method totals $187,750,400 (USD).

Assumptions requiring human ownership:

  • Retail price $5,000/stage, cost $2,000/stage, margin $3,000/stage (60%), ROI 150%.
  • "P&P share held constant" to project 2026 P&P wells.

Estimates / scenarios (not outcomes):

  • ACCESS at 5% / 10% market; fixed 3,000 / 5,000 stages.

Interpretations / judgments:

  • "Plug & Perf dominates — primary conversion target."
  • "CT Sleeves show strong adoption."

Open questions:

  • Where does the 49,709-stage market size come from, and as of when?
  • Is $5,000/stage USD or CAD, and is it Kobold's actual ACCESS price or a placeholder?
  • Who owns the $2,000/stage cost figure?

3. Conflict log

# Conflict Why it matters Resolvable from the note? Human decision
C1 ROI is identical (150%) in every scenario The ROI "output" is just the input ratio restated; it carries no information and can't validate any scenario No Decide whether ROI belongs in the model at all, or compute it against real cost/capital
C2 ACCESS price assumption ($5,000/stage) vs P&P actual revenue-per-stage ($4,969) in the revenue table Two different prices used in one model; which applies to ACCESS? No Confirm the real ACCESS price and reconcile
C3 P&P revenue is exactly $138,000,000 while the others are precise ($21,161,000 / $28,589,400) $138M looks rounded/plugged; the $4,969/stage appears back-derived from it, not independent Partly Confirm whether $138M or the per-stage rate is the real input
C4 Currency unlabeled in the Assumptions block; revenue tables say USD A CAD/USD mix-up silently moves every output No Label every monetary input with its currency

4. Human-review gate

Before this becomes a leadership-facing deck or decision:

  • Attach a real source (and date/owner) to the 49,709-stage market size and the well counts.
  • Confirm the ACCESS price and the cost basis, with currency.
  • Decide what ROI is supposed to mean here — as written it's a tautology, not a result.
  • Decide whether the "dominates / strong adoption" judgments are claims you'll defend or framing.

Do not build the deck until the gate is cleared. The arithmetic is fine; the foundation isn't.

01-structure.md

Structure — workbook spec

A trustworthy workbook keeps truth in separate places. The spec defines where each kind lives before a single formula is written.

Tab Holds Rule
Raw data market size, well counts loaded exactly, each value carries a source ID + date
Assumptions price, cost, margin name, value, unit, source, owner, status
Calculations revenue, scenarios formulas only, no hardcoded outputs
Checks tie-outs, recalc test pass/fail smoke alarm
Summary the ROI / revenue view driven entirely by the tabs above

The test the spec enforces

If I change an assumption, does the relevant output change for the right reason?

A spreadsheet that cannot recalculate is not a model. The current ACCESS analysis fails this test before it's even rebuilt: its ROI is a fixed ratio, so no assumption change moves it. The spec exists to make that impossible in the rebuilt version.

02-workbook-doctor.md

Workbook Doctor — ACCESS Market Analysis

Prompt 2 output. The note is laid out as a model (assumptions, market tables, revenue rollups, an ROI scenario table), so the workbook lens applies even though it's stored as markdown.

2.1 Workbook Map

"Sheet" Purpose Type Status
Assumptions price/cost/margin/ROI inputs assumptions active, unsourced
Market Stage Counts 49,709 stages split by method raw data (claimed) unsourced
Well Counts & Growth 5,548 → 5,709 wells, +2.9% calculations internally consistent
Completion Revenue by Method $187.75M total outputs partly back-derived
ACCESS ROI Scenarios 4 scenarios, all 150% ROI outputs tautological (see scan)
Key Insights "P&P dominates", conversion range interpretation unflagged judgments

Assumption map: none of the foundational inputs carry source / owner / date / unit-currency. The "Assumptions" block lists values but not where they came from.

2.2 Formula Risk Scan

Location Issue Why it matters Evidence Human review?
ROI Scenarios, ROI % column Output doesn't change when assumptions (volume) change The recalc test — the single most important Excel check — fails here ROI = margin/cost = 3000/2000 = 150% by definition; all four scenarios print 150% regardless of stage count Yes
ROI Scenarios, whole table One-variable scaling presented as scenario modeling Every row is just stages × {$5,000, $2,000, $3,000}; there's no independent scenario logic to stress Margin and cost columns are exact multiples of stages in every row Yes
Assumptions / Revenue tables Two prices in one model ($5,000 assumed vs $4,969 actual P&P) Ambiguous which price drives ACCESS economics $5,000 in Assumptions; $4,969/stage in the P&P revenue row Yes
P&P revenue = $138,000,000 Suspiciously round vs precise peers; likely a plug A rounded headline number that the per-stage rate was reverse-engineered from reads as independent but isn't $21,161,000 and $28,589,400 are precise; $138M is exact; 27,780 × $4,969 = $138.04M Yes
All monetary inputs Currency unlabeled in Assumptions A silent CAD/USD mismatch moves every downstream number "$5,000/stage" with no currency; revenue tables tagged USD Yes
49,709 market size; 5,548 wells Numbers with no source or date The whole model rests on these and nothing backs them No citation anywhere in the note Yes

What actually ties out (a good reviewer says both):

  • Revenue rollup: $138,000,000 + $21,161,000 + $28,589,400 = $187,750,400. ✓ matches stated total.
  • Growth: 5,548 + 161 = 5,709, and 161 / 5,548 = 2.9%. ✓
  • Conversion: 5% × 49,709 × $5,000 = $12.43M; 10% = $24.86M. ✓ matches the stated $12.4M–$24.9M.
  • Within each ROI row the margin/cost arithmetic is correct.

The danger isn't broken arithmetic. It's a clean-looking model resting on unsourced inputs with one "output" that's really an input in disguise.

2.3 Repair Plan

Must fix before sharing

  1. Source the market size (49,709 stages) and well counts — citation, owner, as-of date.
  2. Confirm ACCESS price and cost basis, with currency, and reconcile the $5,000 vs $4,969 conflict.
  3. Replace or remove the ROI column. If ROI stays, compute it against real invested cost/capital so it can actually move between scenarios.

Should fix for maintainability 4. Mark the "Assumptions" values as assumption/estimate/placeholder individually. 5. Resolve whether $138M or the per-stage rate is the true input; stop presenting a derived number as independent.

Optional polish 6. Reframe "dominates" / "strong adoption" as sourced claims or label them as framing.

Questions for the owner

  • Is this an internal sizing sketch or intended to back a customer/leadership decision? The bar differs. As written it's the former dressed as the latter.

2.4 Verification Memo

Check Status
Formula consistency Pass (arithmetic is internally consistent)
Source traceability Fail (no inputs cite a source)
Assumptions labeled Fail (values listed, provenance/currency missing)
Checks present Fail (no recalc/tie-out checks; ROI can't move)
Stale data Unknown (no dates on inputs)
Output reliability Fail (ROI is a tautology; market revenue rests on unsourced size)

Not ready for a leadership deck or decision. Internally consistent, externally indefensible.

03-verification.md

Pretty-But-Wrong Review — ACCESS Market Analysis

Prompt 5 output. A second model, fresh context, reads the note as a skeptical reviewer and enumerates. No fixes. Ranked.

Must fix before sharing

  1. ROI is a tautology, not a result. Every scenario reads 150% because ROI is defined as the fixed margin/cost ratio ($3,000/$2,000). Change the stage count and ROI never moves. Presenting it in a scenario table implies the scenarios were tested against profitability; they weren't. Either remove the column or compute ROI against real invested capital so it can vary.

  2. Foundational numbers have no source. 49,709 total stages, the method split, 5,548 wells — the entire model rests on these and none cites where it came from, who owns it, or as of when. Numbers without a date or source, presented as fact.

  3. Two prices, unreconciled. The Assumptions block uses $5,000/stage; the revenue table uses $4,969/stage for Plug & Perf. A reader can't tell which drives ACCESS economics, and the gap isn't acknowledged.

  4. Currency unlabeled on inputs. The Assumptions block gives dollar figures with no currency while the revenue tables are tagged USD. A silent CAD/USD assumption would move every output.

Should fix before important review

  1. Assumptions presented as facts. Price, cost, and "P&P share held constant" are judgment calls shown as settled inputs. Each should be marked assumption / estimate / placeholder with an owner.

  2. $138,000,000 reads as a plug. It's exactly round next to precise peers ($21,161,000 / $28,589,400), and the $4,969/stage rate looks back-derived from it. A derived number is presented as if it were independent.

Polish

  1. Unsourced judgments. "Plug & Perf dominates," "CT Sleeves show strong adoption" — defensible framings, but stated as findings. Tag as interpretation or attach support.

Reviewer's note

What's striking is how much does tie out — the rollup, the growth math, the conversion range are all correct. That's the trap. The arithmetic being right is what makes the unsourced foundation and the tautological ROI easy to miss. Polish read as trust. The file looks done; the truth layer was never built. Clear the human-review gate in 00-source-packet.md before any of this becomes a slide.