Sampling Maestro
The Statistical Sampling Engine Built for Examinations That Actually Get Challenged.
Four specialized modes. Complete examination lifecycle. MTC §4.07 compliant. Built on Cochran, Neyman, and Stringer – documented, peer-reviewed, and ready for Daubert.
Transaction Sampling – Full-Lifecycle Variable Sampling
The Transaction Sampling engine implements the complete classical stratified variable sampling workflow for dollar-value populations. It is the mathematically correct approach when you need to estimate a population mean or total – not an error rate.
- Loads populations from CSV, TSV, or Excel (.xls/.xlsx/.xlsm) – hundreds of thousands of items
- Eight sample size calculators including Cochran SRS (Absolute and Relative), Probe/Pilot, Error Rate (MTC §4.07(c)), and Attribute formulas for comparison
- Eleven stratification boundary algorithms: Dalenius-Hodges, Geometric Progression, Equal Width, Equal Frequency, Neyman Optimal, Lavallée-Hidiroglou, Power Method, Natural Breaks (Jenks), Ekman, Kozak, and Sethi
- Three allocation methods side by side: Independent per-stratum (Cochran), Proportional, and Neyman Optimal
- Certainty stratum and de minimis threshold with what-if materiality analysis
- SRS and Systematic randomization with reproducible seed – full audit trail
- Four estimators with Satterthwaite degrees of freedom: MPU, Difference, Ratio, and Regression – optimal estimator highlighted automatically
- Bootstrap confidence intervals: Classical (t-based), Percentile, and BCa (bias-corrected and accelerated)
- Population Total toggle: switch between per-unit mean (μ̂) and population total (τ̂ = N × μ̂) for liability projections
- Diagnostics tab: achieved precision, footing error tests, ratio/regression eligibility (MTC §4.08(b)), MVP analysis
- Downloadable HTML report with cover page, methodology narrative, and all tables
Dollar Sampling (MUS) – Monetary Unit Sampling with Stringer Bound
The Dollar Sampling engine implements the industry-standard Monetary Unit Sampling methodology based on the Poisson probability model. Each dollar in the population is treated as a sampling unit, giving larger-value items a proportionally higher probability of selection.
- Computes sample size from materiality and expected misstatement using Poisson R-factors
- PPS (probability-proportional-to-size) systematic selection with reproducible seed
- Certainty item auto-detection (items with book value ≥ sampling interval)
- Tainting analysis: Tainting = (Book Value − Audited Value) / Book Value
- Stringer bound evaluation: Basic Precision + Most Likely Error + Incremental Allowances = Upper Error Limit (UEL)
- Pass/fail determination: UEL vs. Materiality
- Sensitivity analysis (1–10% of book value)
- Branded HTML report with Stringer bound computation and pass/fail conclusion
Sample Review – Forensic Examination of Third-Party Samples
The Sample Review engine answers a question that comes up constantly in holder defense and dispute resolution: Is the examiner’s sample statistically defensible? Sampling Maestro provides a rigorous, documented answer.
- Loads the population and reconstructs the client’s stratification boundaries
- Attribute formula detection: automatically identifies when the client’s sample was likely sized using an attribute (error-rate) formula rather than the correct continuous-data Cochran formula
- Size Audit: compares client sample sizes against Cochran per-stratum and Neyman Optimal requirements
- Allocation Forensics: CV(Nh×σh) detects equalized Neyman weights; CV(nh) detects equal allocation – both signs of suboptimal or engineered stratification
- Boundary Comparison: client boundaries vs. Dalenius-Hodges optimal, side by side, with Nh×σh Neyman weights
- Precision Analysis: per-stratum and overall precision sliders, breakeven precision
- Improve Sample: extends deficient samples per stratum; Before/After comparison in Diagnostics and Summary
- Sample-Only Review mode: review without a population file by entering N and estimating population statistics from the sample
- View as Pool: analyze the entire population as a single unstratified pool
- Branded report: Approach A (Independent) and Approach B (Neyman) side by side with projected combined precision
Built-In Methodology Library
Accessed via the “?” button – everything a practitioner or expert witness needs to explain and defend the methodology.
- Interactive Statistics 101 module: live canvas visualization of sampling concepts with adjustable confidence level, precision target, and sample size
- Statistics 201 module: advanced confidence interval visualization
- Three user manuals (one per engine) with step-by-step workflow guidance
- Two methodology references with complete mathematical derivations aligned to MTC §4.07–4.08
- 30+ academic citations: Cochran (1977), Neyman (1934), Stringer (1963), Neter & Loebbecke (1975), Dalenius & Hodges (1959), and more
Compliance & Standards
Every engine is built to meet the standards that govern unclaimed property and audit sampling.
Subscription Access – Coming Soon
Sampling Maestro was built by someone who has sat across the table from examiners, reviewed third-party samples under deposition pressure, and knows exactly what the methodology has to survive. Every feature exists because a real engagement needed it. The built-in methodology library, with step-by-step documentation and 30+ peer-reviewed citations, means practitioners can explain and defend every calculation. No comparable tool exists: not one that combines classical stratified variable sampling, MUS with Stringer bound evaluation, and forensic sample review in a single platform, grounded in published statistical science, and documented to withstand a Daubert challenge.
Subscription plans will be announced shortly. If you’d like to be among the first to access the platform or want to schedule a live walkthrough for your team, reach out now.
Request a Live Demo