5D Library › Persuasive Copywriting Algebra, Loss-Framing Asymmetry and Checkout Micro-Architecture · 中文

Chapter 3. The Friction Equations of Checkout Micro-Architecture

The Hick-Hyman law: logarithmic, not exponential

Hick and Hyman found that the time to choose among equally likely options grows linearly with the logarithm of their number: T=a+blog⁡2(n+1)T = a + b\log_2(n+1) (Hick 1952; Hyman 1953). With illustrative parameters a=0.2a = 0.2 seconds and b=0.15b = 0.15 seconds, one option takes 0.350 seconds and 15 options 0.800 seconds. Doubling the options adds only a constant. So the claim that “each extra option makes abandonment rise exponentially” does not come from the Hick-Hyman law; checkout losses come from another mechanism: at every step some people leave.

An independent-hazard model

If each required field loses a share hh of those who reach it and each page change loses a share dd, completion is

C=(1−h)fields×(1−d)pages−1.C = (1-h)^{\text{fields}} \times (1-d)^{\text{pages}-1}.

Calculation

# Chapter 3. p05: Fields, steps and the arithmetic of checkout loss.
# Hick-Hyman law: choice time T = a + b log2(n + 1) grows with the log of
# the number of options n, not exponentially. Checkout loss is modelled
# as independent hazards: each required field loses a share h of those
# who reach it; each page transition loses a share d. Completion is the
# product of survivals: (1 - h)^fields x (1 - d)^(pages - 1).
from math import log2

def hick(n, a=0.2, b=0.15):
    return a + b * log2(n + 1)

def completion(fields, pages, h=0.02, d=0.08):
    return (1 - h) ** fields * (1 - d) ** (pages - 1)

if __name__ == "__main__":
    print("   options n   choice time (s)")
    for n in (1, 3, 7, 15):
        print("   {:9d}   {:.3f}".format(n, hick(n)))
    print("\n   fields  pages  completion   (h = 0.02 per field,"
          " d = 0.08 per page)")
    base = completion(12, 3)
    for f, p in ((12, 3), (12, 1), (8, 3), (8, 1), (5, 1), (3, 1)):
        c = completion(f, p)
        print("   {:6d}  {:5d}  {:10.4f}  {:+.1%}".format(
            f, p, c, c / base - 1))
    print("\n   each field removed multiplies completion by {:.4f};".format(
        1 / 0.98))
    print("   each page removed by {:.4f}.".format(1 / 0.92))
#    options n   choice time (s)
#            1   0.350
#            3   0.500
#            7   0.650
#           15   0.800
#
#    fields  pages  completion   (h = 0.02 per field, d = 0.08 per page)
#        12      3      0.6642  +0.0%
#        12      1      0.7847  +18.1%
#         8      3      0.7201  +8.4%
#         8      1      0.8508  +28.1%
#         5      1      0.9039  +36.1%
#         3      1      0.9412  +41.7%
#
#    each field removed multiplies completion by 1.0204;
#    each page removed by 1.0870.

With illustrative parameters h=0.02h = 0.02 and d=0.08d = 0.08: 12 fields on 3 pages complete at 0.6642; the same 12 fields on 1 page at 0.7847 (+18.1%); 3 fields on 1 page at 0.9412 (+41.7%). Each field removed multiplies completion by 1.0204, each page removed by 1.0870. This is geometric multiplication, worthwhile, but not an “exponential explosion”.

A race against distraction

Another way to write checkout is as a race against distraction: each visitor’s hazard of leaving per second differs (gamma distributed, once every 150 seconds on average), and the time checkout needs depends on the numbers of fields, pages and payment methods.

Calculation

# Chapter 3. p06: Checkout as a race against distraction.
# Visitors differ in patience: the hazard of leaving per second is
# lam ~ Gamma(shape 2, mean 1/150 s). Checkout time T = 6 s per field
# + 7 s per page (load and re-orientation) + 4 s per payment choice
# considered. Completion for one visitor is exp(-lam T); averaged over
# the gamma it is (1 + T / (2 x 150))^-2 (closed form). We check by
# simulation and compare layouts.
import random
from math import exp

def avg_completion(T, shape=2.0, mean=150.0):
    return (1 + T * (1 / mean) / shape) ** (-shape)

def layout_time(fields, pages, pay_options):
    return 6 * fields + 7 * pages + 4 * pay_options

if __name__ == "__main__":
    rng = random.Random(77006)
    layouts = (("3 pages, 12 fields, 6 ways to pay", 12, 3, 6),
               ("1 page, 12 fields, 6 ways to pay", 12, 1, 6),
               ("1 page, 5 fields, 3 ways to pay", 5, 1, 3),
               ("1 modal, 3 fields, 2 ways to pay", 3, 1, 2))
    base = None
    for name, f, p, k in layouts:
        T = layout_time(f, p, k)
        exact = avg_completion(T)
        sim = sum(exp(-rng.gammavariate(2.0, 1 / 300) * T)
                  for _ in range(100000)) / 100000
        base = base or exact
        print("   {:35s} T {:3d} s  {:.4f} (sim {:.4f}) {:+.1%}".format(
            name, T, exact, sim, exact / base - 1))
    print("\n   marginal value of one field at T = 40 s: {:.4f}".format(
        avg_completion(34) - avg_completion(40)))
    print("   marginal value of one field at T = 120 s: {:.4f}".format(
        avg_completion(114) - avg_completion(120)))
#    3 pages, 12 fields, 6 ways to pay   T 117 s  0.5176 (sim 0.5195) +0.0%
#    1 page, 12 fields, 6 ways to pay    T 103 s  0.5542 (sim 0.5536) +7.1%
#    1 page, 5 fields, 3 ways to pay     T  49 s  0.7389 (sim 0.7383) +42.8%
#    1 modal, 3 fields, 2 ways to pay    T  33 s  0.8116 (sim 0.8114) +56.8%
#
#    marginal value of one field at T = 40 s: 0.0282
#    marginal value of one field at T = 120 s: 0.0149

Three pages, 12 fields and 6 ways to pay take 117 seconds, with mean completion 0.5176; the same content on one page takes 103 seconds, 0.5542 (+7.1%); one page with 5 fields and 3 ways to pay, 49 seconds, 0.7389 (+42.8%); one modal with 3 fields and 2 ways to pay, 33 seconds, 0.8116 (+56.8%). Simulation and closed form agree to two decimal places. Returns diminish: near 40 seconds each field removed adds 0.0282, near 120 seconds only 0.0149.

Checkout time and expected completion, with several layouts marked (p06).

Note

Which fields may not be removed. Information needed for payment, consents required by law and disclosure of refund terms may not be dropped or hidden for the sake of conversion; pre-ticked extra charges and hard-to-find cancellation are documented “dark patterns” (Mathur et al. 2019). This chapter’s models deal only with genuinely unnecessary friction.

Summary

The Hick-Hyman law is logarithmic. Checkout loss is a product of survival rates at each step: with illustrative parameters, going from three pages and 12 fields to one modal with 3 fields raises completion by four to six tenths; the shorter the flow, the more each remaining field is worth.

References

Hick, W. E. 1952. “On the Rate of Gain of Information.” Quarterly Journal of Experimental Psychology 4 (1): 11–26.
Hyman, Ray. 1953. “Stimulus Information as a Determinant of Reaction Time.” Journal of Experimental Psychology 45 (3): 188–96.
Mathur, Arunesh, Gunes Acar, Michael J. Friedman, et al. 2019. “Dark Patterns at Scale: Findings from a Crawl of 11K Shopping Websites.” Proceedings of the ACM on Human-Computer Interaction 3 (CSCW): 81:1–32.

Fan, Y. (2026). Persuasive Copywriting Algebra, Loss-Framing Asymmetry and Checkout Micro-Architecture: Prospect-Theory Valuation Ratios, Numerical Specificity Credibility and Cognitive-Fluency Conversion Bounds [说服文案代数、损失框架不对称性与结账微界面工程] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090205 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).