What does Dense Data Workbenches conclude, in numbers, about “Zero-Training Error-Proof Forms: Constraints, Defaults, Inline Validation, Confirm-or-Undo and Batch Guards”?
Answer: The order of an error-proof form is fixed: what can be derived is not typed; what can be chosen is not keyed; only what can be learned is pre-filled; format belongs to the mask; numbers carry their own check digits; amounts are checked for magnitude; feedback comes when the field is left; undo comes before confirmation; and a batch shows its blast radius first. p61 shows that a default saves hands without necessarily saving errors; p62 that checking the full format on every keystroke manufactures 46,800 false errors a day; p60 that one extra Damm digit catches every adjacent transposition; p63 that a tiered policy cuts dialogs from 4,411 a day to 29; and p64 that a typed-count threshold of 50 rows puts friction on 0.1037 of batches while covering 0.5866 of rows at risk. Chapter 7 builds these practices into four back offices, reconciliation, orders, stock and customer records, and in its last section writes them up as checks the style kit can run.
Scope of the numbers: The book is not accounting, audit or legal advice. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the reporting system of the Tang finance minister Liu Yan, given in paraphrase.
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: 0.1037, 0.5866, 50
| batches 1,080; rows touched 25,133; median 8; mean 23.3; largest 906 | |||
| T | batches typing count | share of batches | share of rows at risk covered |
| 10 | 466 | 0.4315 | 0.9073 |
| 25 | 247 | 0.2287 | 0.7713 |
| 50 | 112 | 0.1037 | 0.5866 |
| 100 | 42 | 0.0389 | 0.3889 |
| 200 | 18 | 0.0167 | 0.2628 |
| largest 1% of batches (10) carry 0.1878 of rows at risk | |||
| cap 300 rows per batch: 9 batches split into 12 extra runs; worst single wrong batch 906 -> 300 |
Show the full program
# Chapter 6. p64: Blast radius of batch actions: where a typed-count confirmation pays for its friction.
# A fictional 30-day log of batch actions at Harbourbridge (seed 6464): 36 batches a day, sizes drawn
# from a log-normal (median 8 rows, sigma 1.5, capped at 1,200 = one day of rows). Assumption, not data.
# If a batch is run on a wrong filter, every row in it is wrong, so rows at risk scale with batch size.
# For each threshold T: share of batches that must type their row count (the friction), and share of
# rows at risk that such batches carry (the protection). Then the effect of a per-batch cap of 300.
import math
import random
DAYS, PER_DAY, CAP_LOG = 30, 36, 1200
if __name__ == "__main__":
rng = random.Random(6464)
sizes = []
for _ in range(DAYS * PER_DAY):
s = int(round(math.exp(math.log(8) + 1.5 * rng.gauss(0.0, 1.0))))
sizes.append(min(CAP_LOG, max(1, s)))
sizes.sort()
n, rows = len(sizes), sum(sizes)
print("batches {:,}; rows touched {:,}; median {}; mean {:.1f}; largest {:,}"
.format(n, rows, sizes[n // 2], rows / n, sizes[-1]))
print(" T batches typing count share of batches share of rows at risk covered")
for t in (10, 25, 50, 100, 200):
big = [s for s in sizes if s >= t]
print(" {:3d} {:20d} {:16.4f} {:29.4f}".format(t, len(big), len(big) / n, sum(big) / rows))
top = sizes[-max(1, n // 100):]
print("largest 1% of batches ({}) carry {:.4f} of rows at risk".format(len(top), sum(top) / rows))
cap = 300
split = [s for s in sizes if s > cap]
extra = sum(math.ceil(s / cap) - 1 for s in split)
worst_before, worst_after = sizes[-1], min(sizes[-1], cap)
print("cap {} rows per batch: {} batches split into {} extra runs; worst single wrong batch {:,} -> {}"
.format(cap, len(split), extra, worst_before, worst_after))Single source and citation
Fan, Y. (2026). Cognitive Unloading of Dense Data Workbenches, Conditional-Format Salience and Zero-Training Error-Proof Form Engineering (Version 0.1.0), Chapter 6: Zero-Training Error-Proof Forms: Constraints, Defaults, Inline Validation, Confirm-or-Undo and Batch Guards. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-138 · v0.1.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026138,
author = {Fan, Yafu},
title = {Cognitive Unloading of Dense Data Workbenches, Conditional-Format Salience and Zero-Training Error-Proof Form Engineering},
year = {2026},
publisher = {5D Library},
edition = {0.1.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · Method: Three Losses, Four Metrics, Evidence Grades and the Liu Yan Reporting System
- Chapter 2 · Scan Paths and the Skeleton of a Table: Column Order, Alignment, Frozen Panes and Density
- Chapter 3 · Preattentive Salience: Let the Anomalies Jump Out, and Only the Anomalies
- Chapter 4 · Alarm Thresholds: Signal Detection, ROC, the Cost Matrix and a Per-Screen Alarm Cap
- Chapter 7 · The Workbench Playbook: Reconciliation, Orders, Stock and Customer Records, and the Distillation Spec
- Chapter 8 · Practice: Rebuilding the Harbourbridge Reconciliation Back Office from Diagnosis to Redesign