What does Drift and Noise on Long Pages conclude, in numbers, about “Closing the Loop: Scroll Depth, Controlled Deletions and a Pre-Registered Test”?

Answer: Chapter 8 sets out a loop of recording, estimating, changing one thing and reviewing, the distilled "Long-Page Belief-Drift Estimator" tool, and a pre-registered deletion test with 17943 visitors per arm together with conditions for overturning the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses.

Scope of the numbers: Chapter 4 pushes a page forward section by section: deleting the feature table of an illustrative page gains 0.0081, and the best and worst orders of sections differ by 0.1764. Marked boxes separate published evidence, runnable Python calculations (20 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: 17943

z_alpha/21.9600, z_beta = 0.8416
p A p B relative lift n per arm days at 1,500/arm/day
0.0400.044 +0.100 39475 27
0.0400.046 +0.150 17943 12
0.0400.048 +0.200 10317 7
dwell difference (screens) n decided per arm
0.11570
0.2393
0.3175
Show the full program
# Chapter 8. p82: sample size for the section-deletion A/B test.
# 主要指标:每位访客的付款转化率(含未决定离开者),双侧 alpha = 0.05,
# 检验力 0.8,两比例正态近似:
#   n = (z_a sqrt(2 pbar qbar) + z_b sqrt(p1 q1 + p2 q2))^2 / (p1 - p2)^2。
# 次要指标:已决定访客停留时长的均值差(标准差取 p81 的量级 1.0 屏),
# n = 2 (z_a + z_b)^2 sd^2 / diff^2。基线 0.040 与日流量 1,500 为示意。
import math
from statistics import NormalDist

ZA = NormalDist().inv_cdf(0.975)
ZB = NormalDist().inv_cdf(0.80)


def n_prop(p1, p2):
    pb = (p1 + p2) / 2
    top = ZA * math.sqrt(2 * pb * (1 - pb)) + ZB * math.sqrt(
        p1 * (1 - p1) + p2 * (1 - p2))
    return math.ceil(top ** 2 / (p1 - p2) ** 2)


def n_mean(diff, sd):
    return math.ceil(2 * (ZA + ZB) ** 2 * sd * sd / diff ** 2)


print("z_alpha/2 = %.4f, z_beta = %.4f" % (ZA, ZB))
print("p A    p B    relative lift  n per arm  days at 1,500/arm/day")
for pb in (0.044, 0.046, 0.048):
    n = n_prop(0.040, pb)
    print("%.3f  %.3f  %+.3f         %-9d  %d"
          % (0.040, pb, pb / 0.040 - 1, n, math.ceil(n / 1500)))
print("dwell difference (screens)  n decided per arm")
for diff in (0.1, 0.2, 0.3):
    print("%.1f                         %d" % (diff, n_mean(diff, 1.0)))

Single source and citation

Fan, Y. (2026). Fokker-Planck Drift-Diffusion on Long Sales Pages: Belief-Drift Rates and First-Passage Times to the Payment Barrier (Version 1.0.0), Chapter 8: Closing the Loop: Scroll Depth, Controlled Deletions and a Pre-Registered Test. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-240 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026240,
  author    = {Fan, Yafu},
  title     = {Fokker-Planck Drift-Diffusion on Long Sales Pages: Belief-Drift Rates and First-Passage Times to the Payment Barrier},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book