Choice overload on the e-book shelf: when fewer choices sell more, and a "start here" default
Date: 2026-10-31 · Harvest run · Topic was not covered before this run: 'choice overload', 'paradox of choice', 'fewer options', 'scheibehenne', 'iyengar' and 'which book first' had 0 hits across INDEX.md and 668 skill files. The decoy effect is a different topic: see pricing-page-decoy-replication-evidence-nudges.md.
What the evidence actually states
- The famous result. Iyengar & Lepper (2000, JPSP 79(6)) put out a tasting table with either 24 or 6 jams. The 24-jam table stopped more shoppers (about 60% vs 40%), but only about 3% of them bought, against about 30% at the 6-jam table. This is one field study, and it gets quoted far beyond what it supports.
- The average effect is roughly zero. Scheibehenne, Greifeneder & Todd (2010, JCR) pooled 63 conditions from 50 experiments (N = 5,036). The mean effect of a bigger assortment was "virtually zero", but it varied a lot between studies. So "cut the range" is not a law.
- When overload does happen. Chernev, Böckenholt & Goodman (2015, J Consumer Psychology) pooled 99 observations (N = 7,202). Once four factors are allowed for, the effect is reliable. Overload grows with: 1. choice-set complexity: options look alike, there is no obvious best one, and the information is dense; 2. decision-task difficulty: little time, a lot to read, awkward presentation; 3. preference uncertainty: the buyer has no settled preference or expertise in the category; 4. decision goal: the buyer is browsing to buy one thing, rather than just looking.
- Practitioner test libraries. GoodUI has patterns for this: #36 "Fewer Or More Results" (9 tests, most recently on a tours site, 27–28 Oct 2026, about 150k visitors) and #113 "More Or Fewer Plans" (7 tests). Their effect sizes are behind a paywall, so we have not seen them and do not quote them. All we take from this is that serious testers treat assortment size as something to test, with mixed results.
Rule we take from this: cut or structure the choices when the options look alike, the buyer is new to the subject, and they are there to buy. That describes a cold visitor to our shop, who does not know Tŷ Mawr. A returning member who knows the case is the opposite case, and the full list serves them better.
How to do it (smallest steps first)
- Count the decisions a cold visitor faces before the first click. That means titles × version selectors × bundles × membership × library. Write the number down.
- Give one default. Add a "Start here" block above the shelf: one recommended first book, one line on why, one buy button, and one free-read link. A default is the cheapest fix and needs no title removed.
- Group, don't list. Put the titles under 2–3 plain headings ("The place and its history", "The case", "The records"), with at most 3–4 cards per group visible before a "See all" tap. Grouping cuts complexity without shrinking the range, which is the moderator Chernev found.
- Make options different, or merge them. If two cards share a price, a cover style and a blurb pattern, the visitor can't tell them apart. Fix that with one sentence each on who the book is for.
- Hide expert controls from cold visitors. A version dropdown on every card is a decision most first-time buyers can't make. Show the current paid edition, and put "Earlier free edition" behind a small link.
- Measure it. Do a before/after with the purchase rate from /ebooks/ visits, ignoring the first 14 days after the change. Better: a sticky 50/50 test of "Start here + grouped" against the current shelf, read with tools/ab_stats.py or tools/bandit.py. Sizes from tools/ab_stats.py: 1.0% to 2.0% needs about 2,319 visitors per arm, and 1.0% to 1.5% needs about 7,750 per arm. At our traffic, only a large change will show up, so test bold layouts, not tweaks.
How we apply it to Cedwyn (live check 31 Oct 2026 06:1xZ)
The live /ebooks/ page has high complexity and is aimed at buyers with uncertain preferences, which is exactly the overload case: - 9 title cards. 8 of them carry a version selector (v0.0.2 £4.99 / v0.0.1 free), plus a "Buy" and a "free version" link. Every paid title is the same price and uses the same card pattern. - More choices on top: an "All four books · £19.96" bundle, the Academy at £49.99 a month, and a library link to 121 more free e-books. - About 30 visible choice points before a stranger has learned what Tŷ Mawr is. - Bundle wording to check (not checked against the build): there are 8 paid titles, but the bundle reads "All four books" at £19.96, which is exactly 4 × £4.99. The visitor can't tell which four, and gets no saving for buying the bundle. See ebook-pricing-bundles-direct.md (bundle anchor about 2–2.5×).
Proposals (markup or theme changes, so they go through the ONPAGE flywheel or the next planned main change; nothing published by this run): 1. A "Start here" block at the top of /ebooks/. Lead with Tŷ Mawr — the history and heritage of Great House Farm, the broadest entry book. Free read first, then buy v0.0.2. 2. Group the 9 cards into 3 headings, with "See all" on mobile (390 px). 3. Replace the per-card version dropdown with the current edition plus a small "earlier free edition" link. 4. Fix the bundle wording: name the four books, or make it "All 8" with a real saving. The price is the project lead's decision, so this is a proposal only. 5. Read the result with tools/ab_stats.py. Use the 14-day ignore window, and note the change date in the growth-report notes.
Sources
- Iyengar & Lepper 2000, "When choice is demotivating", JPSP 79(6):995–1006
- scheibehenne.com (JCR 37(3):409)
- sciencedirect.com (Chernev et al. 2015)
- kellogg.northwestern.edu
- goodui.org ; goodui.org (test counts only; results paywalled, not seen)
- Live page: cedwyn.bucklerfamily.estate (fetched 31 Oct 2026)
Raw Markdown: choice-overload-ebook-shelf-start-here.md · version and changes: changelog · machine index: agents.json