DDE / 001 — Diligence Infrastructure
Vol. 01 MMXXVI
Due Diligence Engine

DDengine

Preempt the Deal.
Flash QOE Standalone Cost Value Creation Dis-Synergies
Plate VI — From Trial Balance to Thesis DDE/IMG/001 · 2400 × 1029
01 — How it works § 01.00 / 05
INPUT

Source files in

Financial files from the VDR — ingested as-is. No template wrangling, no analyst pre-formatting, no data-room scraping.

Trial Balance P&L Census QoE + Other VDR Files
9-STAGE PIPELINE

Filter, classify, benchmark, analyze

A deterministic rulebook strips out everything that isn't an operating cost, then AI-assisted classification maps what remains — every decision rule-traced and reviewable.

1–4Exclusion filters — balance sheet, revenue, non-recurring, allocations
5–7Function · subfunction · cost type classification
8Cost behavior — fixed / semi-variable / variable
9Output assembly & management questions
OUTPUT

A structured model out

A fully auditable Excel deliverable: benchmarked cost summaries, standalone cost estimate, value-creation and dis-synergy analysis, a data-integrity scorecard, and targeted management questions.

Re-run in minutes as new data emerges or management provides feedback — the model updates instantly.

DDengine Deliverable.xlsx
02 — What you get § 02.00 / 05
01
Classified Cost Base
Every operating dollar mapped to function & subfunction with a full audit trail.
02
Benchmark Convergence
Institutional-grade industry data meets multi-model AI consensus.
03
Integrity Scorecard
How much of the model you can trust, and where the gaps are.
04
Management Questions
Auto-generated and targeted at the uncertain dimensions — not a generic questionnaire.
Iterative Input and Report Update
05
Entanglement Analysis
Separates above-benchmark non-labor cost into what is structurally tied to the parent versus what is genuinely reducible.
Iterative Input and Report Update
06
Value Creation
Where cost sits above benchmark, and by how much — quantified as an actionable lever.
07
Dis-Synergy & Standalone
Incremental carveout cost and non-labor entanglement, modeled at the subfunction level.

⟲  Management Questions and Entanglement Analysis are iterative inputs — analyst responses and confirmed entanglement levels feed back into the model and update the report, sharpening the cost read with each round.

03 Why DDengine

Sharpen your thesis.

DDengine compresses the analytical work that usually decides whether a deal is worth pursuing into a same-day model.

i.

Speed at the front

From raw VDR files to a benchmarked cost model in hours. Screen more targets, kill the bad ones early, and reserve senior time for the deals that survive.

Same-day turnaround
No data-room prep
Screen the whole funnel
ii.

Convergent benchmarking

Every cost function benchmarked two ways simultaneously — against commercially procured industry databases and a consensus of three independent AI models. Agreement strengthens the signal; divergence surfaces the question.

Institutional-grade benchmarks
Multi-model AI consensus
Cost overhang, quantified
iii.

Not a static report

Every classification is rule-traced and analyst-confirmable. As management responds, new files arrive, and the team confirms classifications, the model updates — structured files shuttle to management and back. The model sharpens with the deal.

Rule-traced taxonomy
Round-trip override system
Data-integrity scorecard
04 The engine

Inside the pipeline.

DDengine is one governed system, not a folder of spreadsheets. A deterministic rulebook handles what can be decided by rule; AI handles only what genuinely needs judgment — and even then, three independent models must converge before it counts. Nothing is a black box. Every decision has a traceable source and documented rationale.

01

Exclusion Filters

Stages 1–4 strip out balance-sheet items, revenue lines, non-recurring charges, and allocations — isolating the true operating cost base against a governed rulebook.

Balance SheetRevenueNon-Recurring
02

Cost Classification

Stages 5–8 map every operating line to function, subfunction, cost type, and fixed-vs-variable behavior — deterministic first, AI-assisted on low-confidence rows only.

FunctionSubfunctionBehavior
03

Benchmark Convergence

Cost ratios tested against commercially procured industry benchmark databases and a consensus of three independent AI models — surfacing where the target runs hot or lean.

Industry BMAI ConsensusCost %
04

Standalone Cost, Value Creation, Dis-Synergy

Quantifies above-benchmark cost as a value-creation lever, and models incremental standalone cost for carveouts — including non-labor entanglement analysis.

StandaloneDis-SynergyCarveout
05

Entanglement Analysis

Separates above-benchmark non-labor cost into what is structurally entangled with the parent versus what is genuinely reducible on day one.

EntangledReducible
06

Scorecard & Mgmt Questions

A data-integrity scorecard rates model confidence at every level; management questions auto-generate against the dimensions DDengine is least certain about — no generic questionnaires.

IntegrityMgmt Q&ACensus
ArchitectureRule-first, AI-on-consensus
AuditEvery decision source-traced
Feedback loopOverride round-trip
PostureNot a black box
05 The deliverable

One file. Every decision traceable.

DDengine ships as a single structured Excel workbook — DDengine Deliverable.xlsx — built for IC memos, deal-team review, and lender scrutiny alike. Executive snapshot on top. Full classification detail, source linkage, and audit trail underneath. Nothing opaque.

9
Pipeline stages
3
AI models in consensus
∞
Audit trail depth
1
Deterministic core
02Executive SnapshotSummary
03Benchmark SummarySummary
04Cost Walk — FunctionsCost Walk
05Cost Walk — SubfunctionsCost Walk
06APQC Benchmark ScorecardBenchmark
07AI Benchmark ConvergenceBenchmark
08Data Integrity ScorecardQuality
09Client Review & TB OverrideFeedback
10Entanglement AnalysisCarveout
11Management QuestionsDiligence
14Census Detail & Labor AttributionCensus
24MethodologyAudit
06 Access

A founding cohort.

DDengine is opening to a small group of private equity firms and corporate development teams first. Founding partners run free in exchange for candid feedback; everyone else pays per deal, no platform fee.

No time-and-material billing. No analyst team billed by the hour. A single senior-level engagement — walking you through the model, not building it in front of you.

Founding Cohort · 5–10 firms
Free

A handful of slots for PE firms and corporate development teams willing to run live deals and tell us, bluntly, what works and what doesn't.

Unlimited runs during the program
Direct line to the founder
Shape the roadmap and taxonomy
No T&M. No hourly fees. Ever.
Per Run
$2–3k / run

After the founding cohort. Pay per target, no subscription, no minimum. Senior advisory only — walking you through the deliverable, not charging for the hours it took to build it.

Full structured deliverable
Override round-trip included
Model updates as inputs change
No team behind the curtain billed separately

A single pre-LOI cost read from a top-tier advisory firm runs five to six figures, takes weeks, and arrives on a slide deck no one can interrogate. DDengine returns a comparable, fully auditable model the same day — at the price of a long lunch.

* Commercially procured benchmark data, where selected, is a pass-through cost billed at cost.

07 — Engage § 07.00 / 07

Run your next target.

Founding-cohort slots are limited and going to active PE firms and corp dev teams. Send a target and we'll run it.

casey@ddengine.com  ·  703-403-6080

Join the founding cohort → See a sample deliverable →
Client Portal
Home
Run type Select the stage of analysis for this submission
Run 1 — Always Initial Run Full pipeline — all 9 stages from raw VDR files
Run 2 — Optional Subfunction Gap Confirmation Resolve unclassified subfunctions before benchmarking
Run 3 — Optional Management Questions Ingest management responses and update the model
Run 4 — Optional Entanglements Confirm entanglement levels and finalize carveout cost
01 Source Files Upload VDR files
02 Configuration Run settings
03 Review & Submit Confirm and run

Source files.

Step 01 / 03 — Required: Trial Balance
Required TB

Trial Balance

The primary source file. DDengine ingests the TB as-is — no pre-formatting needed. Multi-period tabs are supported; the engine will detect column headers and select the relevant period.

Export directly from your accounting system or as-provided from the VDR. Excel (.xlsx) or CSV preferred.
↑
Drop file or click to browse
.xlsx · .xls · .csv
Optional P&L

P&L / Income Statement

If provided, DDengine uses the P&L to anchor revenue for cross-checks and to enable TB-to-P&L period alignment (TB_PERIOD_ALIGN). Without it, benchmarks run on the TB period.

Monthly or quarterly P&L with a clearly labeled header row. Management accounts format is fine.
↑
Drop file or click to browse
.xlsx · .xls · .csv
Optional Census

Employee Census

Headcount roster used to drive labor attribution (tabs 14–18) and the Census Function Summary. Without it, census tabs are suppressed and standalone labor estimates rely on TB classification alone.

Include columns for name/ID, department, title, location, and compensation. Format flexibility — DDengine normalizes.
↑
Drop file or click to browse
.xlsx · .xls · .csv
Optional QoE

Quality of Earnings

Provides a pre-built non-recurring and normalizing adjustment schedule. When present, DDengine cross-references Stage 3 (non-recurring exclusions) against QoE-identified adjustments.

Sell-side or buy-side QoE in any format. DDengine extracts the adjustment schedule — full workbook is fine.
↑
Drop file or click to browse
.xlsx · .xls · .csv · .pdf
Optional VDR

Other VDR Files

Additional data room materials — management presentations, board decks, contracts, or supplemental schedules. Surfaced during Stage 9 (management questions) and stored for analyst review.

Multiple files accepted. Any format is fine for supplemental documents — they are indexed, not parsed by the pipeline.
↑
Drop files or click to browse
Any format · Multiple files
* Trial Balance is the only required file — all others are optional but improve output quality.

Run configuration.

Step 02 / 03 — Settings that govern the pipeline

Every setting is documented.

Settings map 1:1 to engine/settings.py. Defaults are correct for most initial runs — adjust for carveouts, specific periods, or benchmark requirements.

Deal
Identifying information for this run.
Used to label the deliverable workbook and for run history. Internal codename or target company name.
Operating context
OPERATING_CONTEXT · IS_STANDALONE · IS_CARVEOUT — controls which output tabs and analyses are active.
Standalone: target modeled as a going concern. Dis-synergy, entanglement, and carveout tabs are suppressed. Carveout: activates Tab 10 (Entanglement Analysis), dis-synergy modeling, and standalone cost estimation for costs currently shared with the parent.
Industry & NAICS
INDUSTRY_CONTEXT · NAICS_CODE — drives benchmark peer group and NAICS resolution.
Sets the primary industry bucket for APQC and AI benchmark peer group selection. If the target spans multiple sectors, choose the dominant revenue segment. Used for auto-NAICS resolution when no code is provided.
When provided, this code is used directly for L2 APQC and AI benchmark lookups — bypassing the auto-resolution that derives NAICS from INDUSTRY_CONTEXT. Leave blank to let the engine resolve. For benchmarking accuracy, a 6-digit code is preferred over a higher-level selection.
6 digits · leave blank for auto-resolve
Period alignment
TB_PERIOD_ALIGN — controls which TB quarters are selected to construct the LTM window.
Trailing 12M (default): selects the most recent 4 quarters of TB data on its own reporting cadence. Tab 1 period label reflects the TB's own LTM. Match P&L: forces TB quarters to align with the P&L LTM window — required when the TB trails the P&L by one quarter and you want an apples-to-apples comparison. Requires a P&L file.
Units
UNIT_MULTIPLIER · REVENUE_UNIT_MULTIPLIER — scale TB and P&L values to actuals.
The denomination of values in the Trial Balance. Actuals means values are in whole dollars; Thousands and Millions scale accordingly before pipeline processing. Benchmark comparisons are always done in actuals regardless of TB denomination.
The denomination of values in the P&L / income statement. Often matches the TB, but set independently when the VDR provides documents with different scale conventions.
Classification
AI_TB_CLASSIFICATION — controls whether AI assists deterministic classification on low-confidence rows.
The deterministic rulebook (Stages 5–7) always runs first. When On, AI assists only on TB rows where rule confidence is below threshold — every AI decision is still traceable and overridable. Recommended On for all production runs; Off only for speed testing or rulebook-only verification.
AI benchmarks
BENCHMARK_TYPE · FORCE_REFRESH_BENCHMARKS — multi-model AI consensus benchmarking (Tab 7).
DDengine queries three independent AI models and requires convergence before accepting a benchmark estimate. Cached: re-uses the last fresh result for this industry — fast, no additional cost. Fresh: forces a new multi-model run. Auto: uses cache if <30 days old, otherwise runs fresh. Off: suppresses Tab 7 entirely — APQC only.
When Yes, discards any cached AI benchmark result for this industry and forces a full fresh multi-model run — regardless of BENCHMARK_TYPE. Use after a NAICS or industry change, or when the cache is suspected stale. Has no effect when BENCHMARK_TYPE is Off.
Commercially procured benchmarks
APQC_BENCHMARKS · APQC_PASS_THROUGH — institutional industry benchmark database (Tab 6). Billed at cost.
When On, DDengine queries commercially procured APQC Process Classification Framework data for cost-as-% benchmarks at the function level. This is the institutional-grade source that feeds Tab 6 (APQC Benchmark Scorecard). Note: APQC data is a pass-through cost billed at cost — confirm availability for the target's NAICS before enabling on cost-sensitive runs.
Marks APQC data cost as a client pass-through in the engagement summary. Set to Client when APQC cost is billed directly to the deal; Absorbed when included in the run fee.
Override source
OVERRIDE_SOURCE — controls whether the pipeline ingests analyst classification overrides before Stages 5–9.
None: pipeline runs clean, no prior overrides applied — standard for initial runs. Client Review Tab: reads confirmed overrides from Tab 09 of a prior deliverable before Stage 5 — use on re-runs after analyst review. Override file: reads a standalone override CSV/Excel — for programmatic or bulk corrections outside the standard review tab.
Notes
Analyst context passed to the engine — surfaced in management questions, the integrity scorecard, and Tab 24 (Methodology).
Deal context useful to the engine — known data quality issues, specific cost concerns, seller representations, TB basis (cash vs. accrual), known census gaps, or areas to probe in management questions.

Review & submit.

Step 03 / 03 — Confirm before running
Source files
Run configuration
What happens next
DDengine will run all 9 pipeline stages against your uploaded files. You'll receive an email when the deliverable is ready — typically within the hour. The structured Excel workbook (DDengine Deliverable.xlsx) will appear in your Runs history for download.
Trial Balance required to submit.
✓
Run submitted

DDengine is running your analysis. You'll receive an email when the deliverable is ready — typically within the hour.

Runs

Deal Submitted Config Status Deliverable