Audiobook Choice Engine NYT audio hits + Apple Books US

constant β = 0.1 ρ — — titles self-rank-1 — Store

How the engine ranks

Pick an audiobook you liked. Two algorithms rank every other title against it. CE scores consumer surplus from a quadratic utility; VSM is plain cosine similarity on the same features. ρ is the Spearman rank correlation between the two orderings.

Consumer surplus

Continuous:  CS_d = b_d · q_R · q_C − ½ · b_d · q_C²      (q max-normalised to [0,1])
Discrete:    CS_d = ½ · b_d · 1[cand = ref = value]   (match bonus)
X factor:    CS  += 2·X_R − X_C
Intercept:   α_d  = b_d · (1 + q_R)

Current weights

Features

Continuous: release recency (year − 2000), author stature (log of 1 + the author's NYT audio hits), chart run (log of 1 + months on the NYT audio list). Discrete: genre, same author, series, read by the author, book-club pick, screen tie-in, award mention, co-authored.

Indifference price

P_indiff = P_C − MaxPrice × (CS_R − CS_C)

Shown only when weights come from a fitted price model. With constant weights the score has no dollar meaning, so it reads “—”.

Data

3,132 titles: 1,018 NYT monthly audio bestsellers (Mar 2018 – Sep 2026) and 2,114 comparison titles from Apple Books US. Prices are Apple Books US list prices. Narrator, length and ratings are not yet included.

The Store switch changes links and displayed prices only. In the Australian store, 87% of titles have an Apple Books edition; the rest link to Audible only.

Links to Apple Books and Audible may carry affiliate tags.