The House in the Cerulean Sea
It tells a story. Hardcover files this book as fiction, and its readers there call it hopeful, funny and emotional. Facts, not factors: neither plays any part in our call.
Read it
It is a close call (the audio case sits between the bottom 31% and the top 42% of our shelf of books for how far the recording outruns the page), and this one is good enough to be worth protecting: it sits in the top 14% of our shelf of books for how well the print crowds think of a book. The first encounter only happens once, so we give it to the page.
Length does not decide this
12h 11m to listen, at 1×9h 45m to listen, at 1.25×8h 07m to listen, at 1.5×6h 58m to listen, at 1.75×6h 06m to listen, at 2×about 13 hours to readabout 10 hours to readabout 8 hours to readabout 6 hours to readabout 5 hours to read
The runtime is Audible’s own, shown at the speed beside it. The reading time is our estimate: 393 pages at 30 pages an hour, the pace you told us. Ours is 50, about 300 words a page, read at 250 a minute.The runtime is Audible’s own, shown at the speed beside it. The reading time is our estimate: 393 pages at 40 pages an hour, the pace you told us. Ours is 50, about 300 words a page, read at 250 a minute.The runtime is Audible’s own, shown at the speed beside it. The reading time is our estimate: 393 pages at 50 pages an hour, which is about 300 words a page, read at 250 a minute.The runtime is Audible’s own, shown at the speed beside it. The reading time is our estimate: 393 pages at 65 pages an hour, the pace you told us. Ours is 50, about 300 words a page, read at 250 a minute.The runtime is Audible’s own, shown at the speed beside it. The reading time is our estimate: 393 pages at 80 pages an hour, the pace you told us. Ours is 50, about 300 words a page, read at 250 a minute.
If you must listen → commute-safe
About as demanding in audio as Slaughterhouse-Five.
About as much work in audio as most of the books we hold.
About as much work in audio as most of the fiction we hold.
About as much work in audio as most of the long books we hold.
Reading load against listening load
Where that puts this one among the books we hold: in the middle 40% of our shelf of books on both counts (effort on the page and effort in your ears). The middle of the scale is not the middle of the shelf (most books sit below the average on both), so a book at 0.00 is already in the top 43% for effort on the page. Each of the three is measured against the whole shelf separately, so the bottom figure is not the top two subtracted. It is this book’s own distance from the middle.
What the audio disadvantage is made of. Two things and no others: how fast readers say this one moves, and how demanding the reviews we read found it. There is no third: the structural one has no data source, so it contributes nothing rather than being guessed at. Both directions are declared priors we have never fitted against a real verdict, which is the weakest-evidenced part of the whole model.
How the call was made
Where a bar carries ticks, the grey one is the middle of our shelf and a coloured one is a line we publish, cool where we start backing the page, warm where we start backing the recording. The evidence squares at the foot are a different thing: how far each of our four sources went, hatched where there was nothing to measure at all.
- The audio case
- 0.09It sits between the bottom 31% and the top 42% of our shelf of books for how far the recording outruns the page. Inside the band where neither side has made its case on this number alone, so the call is settled by what else is on the page.
- Narration payoff
- −0.21Daniel Henning’s reading is less well received than the book it came from: in the bottom 25% of our shelf of books, on 1,698 ratings of the narration. This is a difference, not a score: it sets the reading against its own book, so a much-loved book with a merely good reading sits low on this shelf even where nobody disliked either. That is measured on standardised figures rather than on the stars in the charts below, and here the two disagree: listeners mark a performance higher than a story as a rule, 4.62 against 4.54 across our shelf, so this recording’s 4.59 for the performance beside 4.53 for the story is not the level pair it looks. Against their own crowds they are −0.16 and −0.02, which is the same sum the toggle over those charts does. Asked the other way (how the reading stands against what the book’s own readers made of it, which is a different crowd rating a different thing), it comes out −0.91. We report that one and never score it: two crowds that do not overlap are held together by nothing but the standardising, which is a cost the figure above does not pay.
- What is at stake
- 1.07It sits in the top 14% of our shelf of books for how well the print crowds think of a book: well enough regarded that a wasted first encounter costs more, so we lean towards protecting the page. Blended across 1,104,226 Goodreads ratings and 2,204 on Hardcover, each scored inside its own population first, and never added together.
For the page
- The voice lets the book down
- The print crowds think highly of it
For the recording
- Audio is neither helping nor fighting this one
- 1,698 listeners behind it
So the case for the page won it, but only just.
A bigger dot counted for more: the voice and the ear are sized by their share of the sum above, the other two by how much stands behind them. A hatched square is a reason that did not apply here, which is not the same as one that counted for little.
- Evidence
- 1.00How much of the picture we actually have, out of 1.00: four questions, each answered as far as the source could answer it, less anything taken off for reading fewer reviews than we aim to.The reading1,698 people have rated this reading: as good as it getsThe second sourceHardcover knows this book, with 2,204 readers there: as good as it getsThe matchHow sure we are that the recording we scored is this book: as good as it getsThe reviewsWe read 300 reviews to judge how demanding it is: as good as it gets
What each crowd complains about
In the reviews we read, how often readers brought it up and how often listeners did, each against what is usual for that same crowd across our whole shelf, which is the grey tick. The track stops at 4×. Listeners are only ever compared with listeners.
Neither crowd brings any of the three up more often than its own crowd usually does. Whatever this book asks of you, its listeners are not saying the format is what makes it hard.
What the two crowds said
Showing the stars as each shop prints them. The two crowds are not measured from the same place: on our shelf Audible’s listeners average 0.35 of a star above Goodreads’ readers, so a recording that is ordinary for Audible can still out-star a book that is genuinely well liked for Goodreads. Switch the toggle to take that difference off.Showing each figure as its distance from its own crowd’s middle on our shelf, in standard deviations, so 0 is a typical book for that crowd. Listeners start 0.35 of a star up, partly because anyone who finishes a twenty-five-hour recording chose it deliberately, and this takes that off. It is not a claim that audio is better, the story figure is still Audible’s own (people rating a story they heard), and none of it moves the verdict above.
overallAudible
overall, pooled by the shop, so we hold no ruler for it
1,890 ratings · performance and story, as the shop pools them
the performanceAudible
the performance, against Audible’s own crowd
How often this goes badly4.2% of the people who rated the performance gave it one or two stars. More than most recordings here. The middle of our shelf is 2.0%, and half of it falls between 1.0% and 3.1%.
71 of 1,698 listeners gave the performance one or two stars
The shape of the room
- United
- Warm, this one
- Split
- Lukewarm
A warm room: most listeners are happy with the performance, and a small group are not.
1,698 ratings
the story, as heardAudible
the story, as heard, against Audible’s own crowd
1,699 ratings
the book itselfGoodreads
the book itself, against Goodreads’ own crowd
1,104,226 ratings
The series
Cerulean Chronicles
No more announced yet · One voice throughout
2 books numbered on Goodreads’ list. Everything Goodreads lists is published, the last of it in 2024, and nothing new is announced, which is not the same thing as finished. We start saying finished after five quiet years.
Daniel Henning reads all of the 2 volumes Audible names in our records.
From Goodreads’ own series index, checked on 11 September 2026.
Every reading we hold
One complete reading of this book, and it is the reading every audio figure above is measured on. If there is another, we have not found it. The shelf grows in reviewed batches, never because somebody searched for a book.
- Daniel Henningour pick12h 11m9h 45m8h 07m6h 58m6h 06m · One voice · Audible UK and US
- The same voice all through
Performance 4.59 · Story 4.53 · 1,698 rated the reading · 4.2% of the people who rated the performance gave it one or two stars
Hear it · Audible’s clip, its length read from them on load
The recording we mean
- Runtime
- 12h 11m at 1×
- Read by
- Daniel Henning
- Kind
- One voice
- ASIN
- 1529087961 · Audible UK
- Rated in
- Audible UK and US · 1,610 + 88 ratings of the performance
- Judged on
- 1,698 ratings of the narration
- Published
- Tor · 2021
Audible sorts a search by newest, so ours may not be the first one you are offered. The runtime is the check: if the one in front of you says 12h 11m, that is the reading these numbers describe. “Judged on” is the crowd behind every judgement we make about this reading: the smaller of the performance and story counts, which is why it is a little under the two shop counts added. We do not link out to a shop, and we take nothing if you buy there.