5D Library › Cart-Abandonment Recovery Kinetics, Sunk-Data Retention Ratchets and High-Ticket B2B Qualification Mechanisms · 中文

Chapter 8. The Full Ledger, an Experiment to Tell a Ratchet from Selection, and a Pre-Registered Holdout

The full ledger of a recovery flow

Calculation

# Chapter 8. p14: The full ledger of a three-message recovery flow per
# 1,000 abandoners (schedule 0.25 h / 24 h / 72 h, model of p01).
# Incremental orders only; recovered orders are refunded at rate 0.06;
# unsubscribes as in p02; a spam complaint (0.001 per message) costs
# RM 200 in lost deliverability. Compared with what attribution claims.
import importlib.util

def load(name):
    spec = importlib.util.spec_from_file_location(name, name + ".py")
    m = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(m)
    return m

p01, p02, p04 = load("p01_return_hazard"), load("p02_how_many"), \
    load("p04_incremental")
RULE, N, ORDER = (0.25, 24, 72), 1000, 98.0
REFUND, SPAM, SPAM_COST = 0.06, 0.001, 200.0

if __name__ == "__main__":
    claimed = p04.after_first(RULE) * ORDER
    inc = N * (p01.returned(RULE) - p01.returned(())) * ORDER
    refunds = inc * REFUND
    unsub = N * (1 - (1 - p02.U) ** 3) * p02.V
    spam = N * 3 * SPAM * SPAM_COST
    net = inc - refunds - unsub - spam
    print("   attribution claims  RM {:8,.0f}".format(claimed))
    print("   incremental revenue RM {:8,.0f}".format(inc))
    print("   refunds            -RM {:8,.0f}".format(refunds))
    print("   unsubscribes       -RM {:8,.0f}".format(unsub))
    print("   spam complaints    -RM {:8,.0f}".format(spam))
    print("   net                 RM {:8,.0f} ({:.0%} of the claim)".format(
        net, net / claimed))
#    attribution claims  RM   35,672
#    incremental revenue RM   15,207
#    refunds            -RM      912
#    unsubscribes       -RM      717
#    spam complaints    -RM      600
#    net                 RM   12,978 (36% of the claim)

With three follow-ups at 0.25 h, 24 h and 72 h, per 1,000 abandoners: attribution claims RM 35,672; the increment actually caused by the follow-ups is RM 15,207; after RM 912 of refunds on recovered orders, RM 717 of future value lost to unsubscribes and RM 600 of deliverability lost to spam complaints, the net is RM 12,978, only 36% of the attributed figure. It is still positive, but little more than a third of what is claimed.

Telling a ratchet from selection

Calculation

# Chapter 8. p15: Telling a ratchet from selection. Cohort curves cannot
# (p06). An experiment can: randomly give half of new subscribers a free
# import of 12 months of history. Under the ratchet of p05 their first
# renewal rises from 0.801 to 0.958; under selection it stays at 0.806.
# Sample size per arm (two-sided 0.05, power 0.80) to detect a smaller
# real effect, if only part of the ratchet is real.
from math import ceil, sqrt
from statistics import NormalDist

N = NormalDist()

def n_per_arm(p1, p2, alpha=0.05, power=0.80):
    za, zb = N.inv_cdf(1 - alpha / 2), N.inv_cdf(power)
    pbar = (p1 + p2) / 2
    num = (za * sqrt(2 * pbar * (1 - pbar))
           + zb * sqrt(p1 * (1 - p1) + p2 * (1 - p2))) ** 2
    return ceil(num / (p1 - p2) ** 2)

if __name__ == "__main__":
    for p2 in (0.958, 0.85, 0.82):
        print("   renewal 0.801 vs {:.3f}: {:,} subscribers per arm".format(
            p2, n_per_arm(0.801, p2)))
#    renewal 0.801 vs 0.958: 67 subscribers per arm
#    renewal 0.801 vs 0.850: 941 subscribers per arm
#    renewal 0.801 vs 0.820: 6,678 subscribers per arm

Chapter 3 showed that cohort curves cannot tell a ratchet from selection, but an experiment can: randomly give half of new subscribers a free import of 12 months of history and compare their first renewal. If the ratchet is wholly real (0.801 against 0.958), 67 subscribers per arm suffice; if only part of it is real (0.801 against 0.850), 941 per arm; if the effect is small (0.801 against 0.820), 6,678 per arm.

Note

A four-step check. First, abandonment follow-ups go only to buyers who agreed to receive marketing messages, with an unsubscribe link in every one (Parliament of Malaysia 2010). Second, coupons go to first-time buyers only, with plain reminders for repeat buyers. Third, renewal strategy follows the randomised experiment of this chapter, not cohort curves. Fourth, recovery, recommendations and advisor results are all measured as increments against a holdout, with the refund rate as a guardrail (Kohavi et al. 2020).

A pre-registered holdout experiment

Calculation

# Chapter 8. p16: A pre-registered holdout for the recovery flow.
# Abandoners are randomised: most get the three messages, a holdout gets
# none. Primary outcome: an order within 7 days. Detect 0.214 vs 0.30
# (about half of the modelled gain of 0.155), two-sided 0.05, power 0.80,
# with unequal arms.
from math import ceil
from statistics import NormalDist

N = NormalDist()

def n_holdout(p1, p2, ratio, alpha=0.05, power=0.80):
    za, zb = N.inv_cdf(1 - alpha / 2), N.inv_cdf(power)
    var = p1 * (1 - p1) / ratio + p2 * (1 - p2)
    return ceil((za + zb) ** 2 * var / (p1 - p2) ** 2)

if __name__ == "__main__":
    print("   outcome: order within 7 days; 0.30 vs 0.214")
    for share in (0.5, 0.2, 0.1):
        ratio = (1 - share) / share
        n2 = n_holdout(0.30, 0.214, ratio)
        print("   holdout {:.0%}: holdout {:,} abandoners, total {:,}".format(
            share, n2, ceil(n2 * (1 + ratio))))
#    outcome: order within 7 days; 0.30 vs 0.214
#    holdout 50%: holdout 402 abandoners, total 804
#    holdout 20%: holdout 235 abandoners, total 1,175
#    holdout 10%: holdout 204 abandoners, total 2,040

Abandoners are randomised into two groups: most receive the three follow-ups, the holdout receives none. The primary outcome is an order within 7 days, testing 0.30 against 0.214. With a half holdout, the holdout needs 402 abandoners and the total 804; with a fifth, 235 and 1,175; with a tenth, 204 and 2,040.

Hypothesis · not established

Pre-registered predictions (not yet tested): the incremental return rate of the three follow-ups is no more than half the attributed return rate; a random import of history does not raise first renewal by more than 5 percentage points. What would show them wrong: an incremental return rate significantly above half the attributed one, or a first renewal in the import group significantly more than 5 points higher.

Summary of the book

The best timing of abandonment follow-ups is flat, and what matters is sending the first message early; the marginal return of follow-ups falls fast. Discount hooks teach repeat buyers to wait, and attributed recovery greatly overstates the effect. A rising renewal curve may come from a data ratchet or entirely from selection, and only a randomised experiment can tell; until then, “lock-in” is no reason to refuse data export. Screening for high-ticket products rests on the price gap of a menu, on questions that are easy for one kind of client and hard for the other, and on ranking limited advisor time, not on generic friction or false urgency.

References

Kohavi, Ron, Diane Tang, and Ya Xu. 2020. Trustworthy Online Controlled Experiments: A Practical Guide to a/b Testing. Cambridge University Press.
Parliament of Malaysia. 2010. Personal Data Protection Act 2010 (Act 709). Laws of Malaysia.

Fan, Y. (2026). Cart-Abandonment Recovery Kinetics, Sunk-Data Retention Ratchets and High-Ticket B2B Qualification Mechanisms: Multi-Pulse Follow-Up Timing, Switching-Cost Accumulation and Screening Equilibria [弃单挽回动力学、沉没数据续费棘轮与高客单顾问筛选机制] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090209 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).