Portfolio
Engagement case studies, free tools, open-source libraries, and the internal tooling that powers the practice. The problems differ; what they share is data that hasn’t been cleaned, validated, or interpreted yet.
Engagements
Worked examples on Cinderhaven, a fictional specialty food brand built from realistic test data so the work can be shown in full. Campaign Math runs on Quellwood Provisions, an invented specialty-food DTC brand; Payout Math runs on Galewick Climate Systems, an invented HVAC manufacturer. The methodology and deliverables are the real thing; the dollar figures are what that methodology finds in the synthetic data. Every engagement sits somewhere on one arc: recover the money already leaking, prevent the repeat, preserve the margin.
CPG retail & deductions
Product Data Health Audit
Every product record traced against GS1 standards, every downstream failure tracked to the field that caused it, annualized cost in dollars.
Trade Spend & Deduction Recovery
Where the waste is, which deductions are disputable, what's recoverable, and a prioritized action plan.
Chargeback Prediction & Prevention
Chargebacks harmonized into root causes, attributed to the data state at ship time, and ranked into a prevention roadmap by dollars prevented. Proves which data conditions cause the bill and prices each fix.
Remittance Stub Parsing
Four retailer remittance formats parsed into one reconciled deduction ledger. Arithmetic validation flags only the stubs that do not balance, while the dispute window is still open.
EDI Reconciliation: Where the Documents Disagree
A 997 says a document arrived, not that it agrees with the order. Reconciles what the PO, ship notice, invoice, and remittance say for a national mass retailer and two distributors, and puts a dollar figure on each billing mismatch. Across a January 2023 to February 2026 corpus seeded with planted errors: $6.0M of exposure, $3.4M of it on order lines no ship notice matched. Built on Cinderhaven synthetic data.
Monday Morning Report
A weekly operating rhythm for founders: three numbers, the same three every Monday, tiered to revenue, that catch the cash crunch and the velocity slide while they are still cheap to fix.
Fulfillment & OTIF Diagnostic
Reconciles internal fulfillment metrics against retailer scorecards, quantifies the gap, attributes root causes.
Channel Profitability & Capital Allocation
Which channels earn after all costs, where revenue-to-cash leakage concentrates, whether capital flows to the right shelf. Retailer-level cost-to-serve scoring with renegotiation simulation.
Where the Money Comes From
An interactive walk from gross revenue to contribution across 10 channels, ending in a priced answer: where does the next $1M of growth pay? The revenue leader is rarely the contribution leader.
SKU Portfolio Audit
A scored kill list with quantified savings, a fix-or-kill action plan with one lever per SKU, methodology to re-run quarterly.
Retail Readiness & Launch
Readiness across eight dimensions for a specific retailer launch. Cash-flow modeling, GTIN validation, EDI preflight, gap-closing action plan.
Trade Promotion Leakage Audit
Forensic match of promo plan to deductions to settlements. Finds double-funded promos, phantom promos, and rate discrepancies, then prices and prevents them.
Production Demand Forecast & S&OP
Demand signals, capacity constraints, and seasonality in one model. What the next launch does to existing commitments, before you accept the PO.
Validation Pipeline Build
The audit finds the problems once; the pipeline catches them on every new item after that. Automated validation, item-setup preflight, exception protocols your team owns.
The Blast Radius: Recall Traceability
When a shared-ingredient lot touches multiple production runs, a recall isn’t one batch. It’s every batch that lot ever entered. In Cinderhaven’s Scenario B, a one-day recall expands into 23 affected lots, 14 SKUs, six retailers, and 5,785 cases in channel: $52K to $81K in retrieval before legal or brand cost. The variable that sets the blast radius is not ingredient risk but whether the lot genealogy exists when the call comes in.
Door Math: Distribution Penetration Tracker
Tracks which stores are actually carrying a brand’s products versus which were authorized: the gap where revenue dies. Built on Cinderhaven synthetic data (50 SKUs, 6 retailers, 640 doors). Four views: door count with authorization-to-scan gap analysis, ACV% and TDP trend lines, exception drill-down, and a printable buyer-meeting scorecard.
Spin Rate: Penetration × Velocity Quadrant
Once a brand is on the shelf, is it actually selling, or wide but dead? Plots every item by distribution penetration (x) against velocity/SPPD (y), bubble-sized by dollars, across four quadrants: stars, hidden gems, wide-but-dead, and question marks. Ranked expansion and at-risk lists, plus a quadrant-migration view showing what moved since last period. Built on Cinderhaven synthetic data (50 SKUs, 6 retailers, 640 doors).
Void Finder: Authorized but Not Selling
Finds every store where an item is authorized but not scanning, classifies each void as never-scanned or went-dark, and prices the gap from the median velocity of comparable stores in the same volume tier and region. Output is a ranked, broker-ready work list with store numbers and addresses, including a regional never-scanned cluster traced to a single botched shelf reset. Built on Cinderhaven synthetic data (50 SKUs, 6 retailers, 640 doors).
Slot Math: Shelf Share vs Sales Share
Is shelf presence proportional to sales, per retailer and region, and what is the gap worth? At Cinderhaven, six over-shelved cells put $1.6M of scan revenue behind shelf it does not earn; the buyer’s number is shown first, and the club cells are called club-normal instead of an expansion order.
Open the tool →Lift Math: Trade Promotion Incrementality
Two baseline methods scored against known ground truth. At Cinderhaven, 2,170 of 5,735 estimable promotion events lost money under the most forgiving read, and the estimator’s own error is published, including where it is large.
Open the tool →Decompose: Which Lever Moved Sales
Splits a brand's period-over-period sales change into the only three levers that can move it (buying households × purchase frequency × spend per trip) and reconciles them exactly to the sales delta with a Shapley waterfall. The demo catches growth that's actually erosion: sales up $1,954 while penetration falls 2.2 points and price carries the entire number. Board-ready verdict computed from the data, never scripted. Built on a seed-locked synthetic household panel.
Leaky Bucket: Trial vs Repeat Analyzer
Of the households that tried you, how many came back, and is your penetration growth real adoption or just expensive sampling? Two launches side by side: a big trial spike with ~14% repeat (a promotion that looked great in month one) against modest trial with 51% repeat (the quiet winner). Cohort retention triangle with right-censoring shown, not smoothed; every summary number applies a maturity cutoff. The integrity check on Decompose: #3 counts the buyers, #4 says whether they stuck. Built on a seed-locked synthetic household panel.
Rebate & incentive payouts
Payout Math: Incentive Payouts, Re-derived
Of everything an incentive program approved for payout, how much broke the program’s own rules? One program year re-derived: $17,597 approved that should not have been, $7,336 earned and never paid, every line reproducible by hand from the published rulebook.
Open the tool →Campaign & web analytics
Campaign Math: Attribution, Bracketed
Which campaign worked best depends on a rule nobody chose: 9 of 12 campaigns change rank with the attribution rule, one swings 0.69x to 1.88x, and $5,064 of spend bought clicks the data cannot trace. Four transparent rules, every number reproducible by hand.
Open the tool →Ongoing
Monthly deliverables, not projects. Cancel anytime.
Free Tools
Use these without an engagement. No login, no paywall.
The Question Engine: Fifteen CEO Questions, Answered in 30 Seconds
Pick the question your Monday meeting cannot answer: should I fire my biggest customer, can I afford this launch, which SKUs should die. Documented rules run against the data and return a one-sentence verdict, one chart, and the three numbers behind it. Rules, not models: every threshold documented, every verdict auditable. Built on Cinderhaven synthetic data (50 SKUs, 6 retailers).
Retail Readiness Scorecard
Score your brand across eight readiness dimensions. Red flags route to the relevant engagement page.
GTIN Product Data Validator
Validate product data against GS1 standards with retailer-specific context. Generates branded PDF reports with a prioritized fix roadmap.
EDI Preflight
Check EDI transaction sets for common errors before submitting to a trading partner.
Item Setup Pre-Flight
Retailer/distributor item-setup pre-flight: codified partner schemas and a typed validation engine that flags new-item form rejection risk before submission. Runs client-side via Pyodide.
Source on GitHub →Cost of Saying Yes
Model the true cost of a new retailer launch: month-by-month cash flow, not just the revenue projection.
Data Debt Calculator
Estimate your product data debt from five inputs. Sensitivity analysis, compliance timeline, cost-of-delay curve.
Open Source
Published on PyPI. Install with pip.
Data Hygiene Auditor
CLI that audits Excel files for mixed formats, misused fields, placeholder floods, and phantom duplicates.
datascope
Data profiling and quality scoring for tabular datasets. Summary statistics, anomaly detection, quality reports.
Cinderhaven Data Platform
Modern data platform for CPG data shapes: source-to-mart pipelines, quality testing, orchestration, lineage. Built around Cinderhaven, a synthetic demonstration dataset, not a client.
Dimension & Weight Integrity
Guided narrative analysis of case dimension and weight divergences. Five chapters from raw measurement audit through cost modeling and retailer compliance risk.
Source on GitHub →Product Master Data Model
The entity model a specialty food product master should have: one SKU walked from brand to pallet, a GTIN at every packaging level that carries one, and the attribute sets each retailer requires, mapped to the entity that owns them. Written as working Postgres DDL with dbt contracts, not a whiteboard diagram. The model that closes the gaps behind the $93K/yr in Cinderhaven chargebacks the Product Data Health Audit traced to product-data defects.
Source on GitHub →Consultant Tooling
Internal tools that power the practice. Published for transparency.
Compliance Guides
What each rule requires, and by when.
GS1 Sunrise 2027
By December 31, 2027, retail POS systems are expected to scan 2D barcodes. Why your product data is the part most brands haven't checked, and what to do before the transition compresses your timeline.
FSMA 204
Congress has barred FDA from enforcing the Food Traceability Rule before July 20, 2028, but your retailers' programs are already in effect. What the rule requires and where brands fail.
Data Standards Cheat Sheet
GTIN anatomy, GDSN syndication flow, freight class table, retailer portal comparison, GS1 Sunrise 2027, and FSMA 204: every number and standard behind item setup, on one reference page.
Start with a form, not a call.
Tell me about the data problem. I’ll tell you where the impact is hiding.
Get in Touch