What does Live Locks, Bank Matching and Styling CRM conclude, in numbers, about “Matching Rows to Orders: Likelihoods and the Hungarian Method”?

Answer: Chapter 4 matches rows to orders with likelihood scores and the Hungarian method: with a unique offset per order only 0.0258 of rows need a person, no automatic match among 5090 is wrong, and checking falls from 348.3 to 9.0 minutes a session.

Scope of the numbers: Its customers, delays, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial, banking or data-protection advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.0258, 348.3, 5090, 9.0

methodmanual shareminutes/sessionsaved
all by hand1.0000348.30.0000
exact key0.244685.20.7554
Bayes, offsets 0..990.055719.40.9443
Bayes, unique offsets0.02589.00.9742
0 errors in 5090 automatic matches: error rate below 0.00059
to show a rate below 0.001 with no error seen: 2995 matches
to show a rate below 0.0005 with no error seen: 5990 matches
Show the full program
# Chapter 4. p08: What the matcher saves, and how far to trust it. From
# p07: 5225 bank rows in 30 sessions (174 a session on average); manual
# shares 1.0000 (all by hand), 0.2446 (exact key), 0.0557 (Bayes with
# offsets 0..99) and 0.0258 (Bayes with unique offsets), with 0 wrong
# automatic matches among 5090. A row checked by hand takes 2 minutes.
# We print minutes per session, and the 95% upper bound on the error
# rate when n automatic matches show no error, 1 - 0.05 ** (1 / n)
# (about 3 / n), and the audit size that bounds it below a target.
# No sampling.
import math

ROWS = 5225 / 30

if __name__ == "__main__":
    print("     method                  manual share   minutes/session"
          "   saved")
    for name, m in (("all by hand", 1.0), ("exact key", 0.2446),
                    ("Bayes, offsets 0..99", 0.0557),
                    ("Bayes, unique offsets", 0.0258)):
        print("     {:22s}  {:12.4f}   {:15.1f}   {:5.4f}".format(
            name, m, 2 * ROWS * m, 1 - m))
    n = 5090
    print("   0 errors in {} automatic matches: error rate below {:.5f}"
          .format(n, 1 - 0.05 ** (1 / n)))
    for target in (0.001, 0.0005):
        k = math.ceil(math.log(0.05) / math.log(1 - target))
        print("   to show a rate below {} with no error seen: {} matches"
              .format(target, k))

Single source and citation

Fan, Y. (2026). Live-Shopping Flash Inventory Locks, Bayesian Bank-Feed Reconciliation and Personalized Styling CRM (Version 1.0.0), Chapter 4: Matching Rows to Orders: Likelihoods and the Hungarian Method. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-112 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026112,
  author    = {Fan, Yafu},
  title     = {Live-Shopping Flash Inventory Locks, Bayesian Bank-Feed Reconciliation and Personalized Styling CRM},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book