Safe pick
The dependable match. Familiar enough to feel instantly right.
A little peek behind the bookshelf
What happens after you send me a reading mood
Dear reader,
From the outside, Next Good Read is meant to feel tiny: you tell me what you feel like reading, and a little book list arrives on your Kindle.
Behind that quiet flow is a finely tuned little recommendation engine. AI is only the first reader: it helps parse pace, tone, tropes, boundaries, and hidden mood. Then I shape the shortlist with shelf memory, freshness checks, repeat-risk scoring, author rotation, anchor health, and the small quality signals that show whether a list is becoming more personal or just more predictable.
The lovely part stays simple on purpose: you share the kind of reading moment you want, choose when it should arrive, and I turn that into five curated picks with different jobs: a safe choice, an emotional match, a hidden gem, a newer doorway, and one brave little wildcard.
Every send teaches the shelf a little more: what has already been offered, which taste directions are working, where the recommendations are getting too familiar, and when the whole thing needs a fresher turn.
Love from Suzy
I wanted it to feel like someone behind the counter actually remembers you: the mood you were in last time, the books that lit up, the ones that did not, and the little clues that make a recommendation land at exactly the right moment.
The five books each have a job
A good recommendation list should not be five versions of the same obvious book. Each issue is shaped like a tiny bookstore table curated for one reader.
Emotional match
The book chosen for the feeling underneath your note.
Hidden gem
Less obvious, but still deeply aligned with the reader’s taste.
Newer release
A fresh doorway so the list does not live only in old defaults.
Wildcard
A stretch pick: adjacent, interesting, and a little bit magical.
The anti-repeat brain
Repetition is the quickest way for recommendations to feel lazy. Next Good Read keeps a memory of what has already been sent and checks each batch for repeated books, repeated authors, repeated taste anchors, freshness, novelty, drift, and overlap with the previous issue.
0-8 weeksrecent books are blocked from resurfacing.
2-6 monthsolder matches are strongly discouraged.
6-12 monthsexcellent matches may rarely return with a reason.
12+ monthsgreat books can become eligible again if the reader’s mood points back there.
Anchor rotation
I also watch for “anchor collapse,” where a reader keeps getting pulled toward the same identity books or authors. Instead of orbiting one obvious title forever, the shelf rotates through nearby taste neighborhoods: the same feeling, new doors.
That is how recommendations stay coherent without becoming stale.
The reader map
This is the human-touch bit. The reader map is how I track what is working, spot when the recommendations are getting too samey, and constantly tweak performance so the shelves stay warm, varied, and useful. It lets the front door stay sweet and simple while the back room keeps learning.
Does the list still match the feeling the reader asked for?
What has been loved, skipped, disliked, or already sent?
Are books, authors, and taste anchors rotating naturally?
Would this feel thoughtful if it came from a favourite bookshop?
This is how I keep an eye on the whole bookshelf: freshness, variety, novelty, drift, anchor health, and whether delivery is behaving itself.
Taste clusters help me see which shelves are growing, which genres need more breathing room, and whether the five recommendation roles are staying balanced.
History views show repeated authors, resurfaced books, timezone spread, and remembered-book signals, which is how I keep nudging the recommendations back toward delight.
Seeded reader data is shown in these dashboard screenshots for privacy, so no real reader details are on display.
What gets measured
Freshness
100%How well recent repeats are avoided.
Diversity
89%Author, genre, role, and anchor spread.
Novelty
98%How much the list avoids stale defaults.
Anchor health
99%Whether taste centers are rotating naturally.
Seeded shelves, real tuning
Before changes go anywhere near readers, I seed development shelves with realistic taste profiles, book memories, send histories, and repeated-recommendation scenarios. It is like setting up a little pretend bookshop full of readers with different moods: cozy fantasy, dark thrillers, systems sci-fi, romance, nonfiction, daily schedules, weekly schedules, and monthly rituals.
The point is not to make things look clever. It is to test whether the recommender behaves like a thoughtful bookseller over time: avoiding obvious repeats, resurfacing only when appropriate, balancing safe and adventurous picks, and staying warm instead of mechanical.
The dream is simple: opening your Kindle should feel a little like stepping into your favourite bookstore, where someone remembers your taste and still manages to surprise you. - Suzy
Built with coffee, book love, and a suspicious number of tiny quality checks.