"How long should this novel be?"
And how much can you tell from "too long" and other rejections for measurable reasons?
One of the most frequently asked questions among writers’ groups is a version of ‘how long should my novel be?’ And alternating at the top of the most frequent answers are a) ‘90,000 words’, and b) ‘as long as it needs to be’.
Are both of these true? How can both of these be true? But to my mind, they are both true…ish. So this post unpicks what I think is going on, and ends with a TLDR for the essentials of wordcount for different forms and genres.
I blogged a while back about what to do when your story is far too long, which entailed thinking about what lengths different forms of fiction should be. Since then, my sense is that the commercial limits have got shorter. A decade ago an agent said the production costs of print rise significantly after around 140,0001 words, but the other day a friend quoted an agent putting the same boundary at 100,000 words, and the other professional writers in the room nodded.2
1) The ‘How Long?’ question can simply be about process, not product
If you remember that writing an 80,000 word novel is essentially a task of making 4.800 billion decisions,3 then it’s obvious why many writers find having some fixed reference points very helpful. For example, Ian McEwen says that one of the first things he knows about a new project is ‘the maths’: that it’s three parts of 20,000 each, or five parts of 15,000, or whatever. It works for him.
For writers like these, working to some kind of wordcount serves a similar function as deciding that they’ll stick to an internal, first person narrator, or they’ll only cover five years in your real-historical character’s eighty-year life: it narrow the options down to something manageable.
2) When it is about product, it may become a reason for agents to say No
Another agent I was talking to described their principal editorial job as ‘removing all the reasons for an editor to say “No”.’ If agents know that editors - often for very good reasons - don’t want to buy books with certain characteristics (including length) then agents will try to make sure the books they’re selling don’t have those characteristics. So, arguably, if an agent is offered a book which is longer than 100,000 words, the wordcount increases the chances of them saying No.
I see why they might, but equally every agent and publisher has stories of one or more books they sold which are full of characteristics that they’d have sworn they didn’t want and wouldn’t take on. What’s more, though it’s true that all the boundaries are fuzzier in literary fiction than in commercial fiction, when both Ducks, Newburyport (1000pp) and Orbital (126pp) can be shortlisted for the Booker, there are clearly very few absolute conclusions you can usefully draw about these things.4
I genuinely don’t think the literary barbarians are at the gate, just that culture evolves along with the society and technology that supports it, and it’s not always a comfortable experience for creatures who were adapted to how things used to be: it’s no good protesting, social-media-outrage-style, that 377,000 words was good enough for Charles Dickens, so are we saying Bleak House is no good??? The 21st century book trade just doesn’t work the same way.
So how seriously is the writer to take these reports from the front?
When you’re looking at any characteristics which you hear ‘agents and editors don’t like’ but books you love are thick with, don’t forget to compare like with like: because the culture evolves, looking at authors and books which garnered readers more than twenty or so years ago can’t tell you much about how publishers think readers now approach the work of a new writer who they know nothing about.
If that careful inspection of recent books whose readers will like the book you’re writing, tells you something about length, plot, blurb or anything else, then that’s useful information about what that kind of reader likes: what they, with their purses open, are saying Yes to.
But even recognising the need of agents and editors to concentrate on trying to sell books which don’t present obvious-to-the-industry unsaleabilities, I do think that there’s a risk for our individual creativity in adopting the ‘removing reasons to say no’ mindset too completely. And in the end, your individual creativity is all you have.
3) Editing a book into something that no publisher has any reason to object to is an essentially negative process
It says nothing about what’s good, only about what isn’t bad. So if you adopt that mindset too completely you will be using your writing mind as ChatGPT and the others use deep language learning: to come up with the highest common factor, or perhaps the lowest common denominator, for your story-points.
At best, you will get more industry people saying a provisional ‘Maybe’ - but how much will the book be the product of your writing self? Don’t forget that in a competitive market, the only writing you can do better than any other writer is your writing - so losing sight of that is disastrous, if only in the longer term. At worst your book will be so bland and seen-it-all-before that no one will want the bother of reading it at all.
Of course your readers will be likely to have many things in common, but it’s the sting of the specific, the individual, the particular and above all the new-to-the-reader (£), which will make your story come alive - and it’s vital that your writing has that quality, if an agent or an editor is going to say Yes.
In other words - the real work is to create all those reasons to say Yes.
All this, of course, is equally true if you’re aiming to reach readers directly by self-publishing - except that finding that sweet spot between giving readers what they want, and not giving them mere lowest-common-multiple storytelling, rests wholly on your own shoulders, as does the print bill.
What’s more, just because 90,000 is the peak of the fiction wordcount bell-curve doesn’t mean that there are no other possible points on the curve for a novel to sit at. In the 2023 HWA Gold Crown there were some real doorstoppers at every point of the literary-commercial spectrum, and some of them were very good: I could completely see why a publisher thought they were worth publishing, and some didn't even get my editorial pencil twitching to tighten them up.
4) Some ‘No thanks, it’s too long’ rejections are just the rejector’s experience talking.
Of course, the other reason editors say No to longer books is because writers whose books are too bony are relatively rare, while writers whose books are carrying too much weight are extremely common. So many of the not-bad 130,000 novels agents and editors have read contain a potentially good 85,000 worder trying to get out, but such novels will need to be very good indeed in other ways, for the editorial work they need to be worth the agent’s or publisher’s time.
If you’ve had the ‘too long’ feedback, try some of the suggestions in the Fiction Writer’s Pharmacopoeia to see what the real, underlying problem might be. If you already sense it really is about too much plot and story, try this post on structural edits, and if really is about too many words per event, try this post on line-edits.
But once you’ve been really ruthless with surplus words and story-elements, so that there’s no word, sentence or scene which isn’t earning its keep in at least two different ways, but it’s still on the long side in merely arithmetical terms, then it could be well worth sticking to your guns.5 If an agent or an editor loves the voice, the story, the premise and the writing, and thinks those aspects would make it saleable, then if the length isn’t ideal that won’t stop them at least saying ‘Let’s talk’.
Finally, it’s always worth remembering that agents will often pick on a relatively objective reason for rejecting a book, such as its length, when a more accurate explanation would be, ‘It just doesn’t float my boat in a way that I don’t have time to analyse in an email (and besides, I have no evidence that you are one of the minority of rejectees who will argue back, stalk or even threaten me, but I’ve had enough of those guys for a lifetime).’
In other words, bear in mind that some of those ‘too long’ rejections are not data you can use, or even add to your writers’ group data bank, because they shouldn’t really be filed under Reasons To Cut at all.
TLDR: Wordcounts for different forms and genres
These are my general guidelines, but definitions vary a lot, so make sure you check the submissions guidelines for the specific publication, agent, publisher or competition, and stick to them.
Flash fiction: definitions vary wildly: some I’ve seen say ‘under 400 words’, others ‘750-1,500’. For myself, I’d say ‘under 1000,’ but what do I know?
Short stories: magazines and competitions mostly state 2k or even less; more than 5k seems very rare.
Novellas: anything below around 15k is a long short story, anything 20-50k probably counts as a novella, though in category fiction such as Harlequin Mills & Boon novels may well be only around 50k. But the novella/novel distinction can be as much about the ‘feel’, the structural complexity, the number of characters, scenes and settings and so on, as the arithmetic.
Commercial fiction seems to fall into the bracket of around 80-100k words, with the peak of the bell-curve at 90k.
Literary fiction has a bit more leeway: say 70-130,000? However, publishers do find that it’s harder to generate as much noise and excitement around a debut as they need to, if the novel is very slim.
Memoir and creative non-fiction is general sold in the same way and through the same channels as fiction, but my sense is that it’s almost invariably at the slimmer end of things: say 75-85,000?
Different genres vary: speculative fiction and historical fiction seem happy to be at the longer end of these ranges, since readers love to feel they’re buying into a great, fat ‘other’ world. The readers of contemporary-set crime and women’s fiction are maybe less keen.
Children’s fiction is a different ecosystem which I only know at second-hand, so I can do no better than to send you to the brilliant Kit Lit blog for the full breakdown of lengths from Board Book to New Adult.
It’s perhaps worth mentioning that wordcounts are the currency of agents and editors. No one thinks in terms of number of pages until they’re actually talking to printers. Talking about your own story-in-progress as a certain number of pages tells none of us anything useful, because it varies wildly depending on the typeface, point-size, line-spacing, margins, and a dozen other factors.
Perhaps the greater resistance to length is because these days, in some genres, half or fewer of all sales are in print, so the print bill is shared across fewer units. In a low-margin, high volume industry like the book trade, unit cost is the central conundrum to be solved.
Apparently, the average graduate or other educated person has a usable vocabulary of about 60,000 words. Obviously multiplying that by 80,000 as I have isn’t quite representative, as some words can’t be followed by some other words. But it’s still a lot of decisions.
And no, the fact that Orbital won the Booker, while Ducks, Newburyport was only shortlisted, is not a useful conclusion in our sense. It just tells you something about which book that set of judges, on that day, with that shortlist, could agree should win.
For what it’s worth, when my editor at Headline Review said that something in A Secret Alchemy was taking up too much space, and anyway at 137,000 words the book was a good bit longer than the usual, she was astonished when I pointed out that The Mathematics of Love was 141,000 words: she’d completely forgotten. She was right about A Secret Alchemy, and after I put it on a diet, it ended up at 127,000. However, when it was published, everything about the book design was set up to make it look as plump as The Mathematics of Love, even thought it wasn’t in wordcount. We readers may feel we have shorter attention spans in the 21st Century, but many readers still want to feel they’re buying a whole, rich world where they’ll be staying for a nice long time.