How Nova's Quality Algorithm Works: Craft, Not Clicks
By Product Team 2025-11-09
algorithmdiscoveryhow it worksquality scoringwriter resources
Most platforms rank stories by clicks. We rank them by craft—then validate with real readers.
You've probably noticed that on other platforms, the loudest stories win. The ones with the flashiest covers, the most aggressive marketing, the biggest fanbases. Quality gets buried.
Not here.
Nova's algorithm asks three simple questions:
If all three answers are yes, your story rises. You can't fake any of them.
We hear the criticism often: "You can't quantify art. Algorithms force a generic style. You're trying to turn writing into math."
We agree. You cannot quantify art. But you can quantify the signals that art emits.
Nova isn't a robotic editor forcing you to write like Hemingway. It's a mirror.
Nova's approach is different. We provide a "mirror" to help writers understand the signals their work emits, allowing them to refine their craft with intent.
If you're writing a numb, dissociated scene, a flat emotional line is perfect. If you're writing a climax, it's a problem. Nova gives you the data; you decide if it matches your intent.
We measure resonance, not rules.
Think of Nova's algorithm like a literary awards committee with perfect recall. It doesn't just count votes—it reads the room.
We analyze your actual prose using 768-dimensional semantic embeddings (the same tech behind advanced LLMs, but tuned for analysis):
Genre matters here. A thriller's pacing matters more than a literary novel's. Poetry is scored completely differently than prose. We don't apply a one-size-fits-all rubric.
You can't fake craft. You either have vocabulary depth or you don't. You either understand pacing or you don't. Bots can't write their way past this.
This is where readers vote with their attention:
This matters more than you'd think. A story with 10,000 views but 20% completion rate doesn't rank higher than a story with 1,000 views and 80% completion. Finishing = real quality signal.
Recent engagement counts more. If your story was hot three months ago, that doesn't guarantee it stays on top. Sustained reader love is what Nova rewards.
Nova uses semantic analysis to understand your story's DNA:
This can't be gamed with keywords. Semantic analysis reads the actual content, not just your tags.
Here's where Nova gets clever:
Your Score = 40% Craft + 40% Reader Love + 20% Newness Boost
Why split it this way?
What's the newness boost? New stories automatically get extra visibility for the first month. Then the boost fades. This prevents old stories from blocking all new work, but doesn't let any story coast on newness alone.
The key insight: You need to be good AND retain readers. You can't buy your way up. You can't trick the algorithm. You have to write something people actually want to finish.
New story, no reader data yet. How does Nova launch you?
Your story is scored on writing quality alone. Nova samples it to real readers—a mix of genre fans and adventurous readers exploring new work.
You get guaranteed initial readers. About 100 people across different reading styles and backgrounds, so the first data isn't skewed.
As real readers engage, their behavior starts mattering more. Your craft score doesn't disappear, but it shares weight with engagement data as it arrives.
Once you have enough reader data, the full composite score kicks in. Your story's ranking is now based on how well it's written AND how readers are responding.
Why this matters: New writers aren't invisible. Your first story gets a fair shot based on quality. But you also can't coast on newness—you have to keep readers reading.
Nova doesn't just score stories. It watches for manipulation:
Suspicious patterns we detect:
What happens? Suspicious accounts get lower weight in scoring. Their interactions count for less. Bad actors don't disappear, but they can't game the system.
Hidden metrics. We don't tell authors every metric we track. This prevents people from optimizing specifically for our formula. You can't hack what you can't see.
As a writer on Nova, you get transparent data:
Your Story Score: 73/100
Craft (82/100)
├─ Vocabulary richness: 85th percentile
├─ Pacing rhythm: 60th percentile ← Room to improve
├─ Dialogue quality: 75th percentile
└─ World-building: 70th percentile
Engagement (64/100)
├─ Reader completion rate: 80% ✓
├─ Return visits: 60%
└─ Drop-off point: Around paragraph 8 ⚠️
Suggestions
└─ 40% of readers stop around paragraph 8.
Common causes: pacing shift, info dump, or weak momentum.
Try: Earlier conflict escalation or tighter pacing in that section.
This isn't just a score. It's feedback. You see where you're strong and where you can improve. You see exactly where readers disengage and why.
You probably noticed: Nova's recommendations are different.
No algorithm designed to keep you scrolling forever. We're not maximizing your screen time. We're maximizing your reading satisfaction.
When you search for "cozy fantasy," you get actual cozy fantasy—not grimdark with a friendly-faced cover. When you filter by "completed stories," you're not risking abandonment. When you save a story to your reading list, the algorithm understands you're serious about it.
Recent, highly-rated completions bubble up. Forgotten gems surface when relevant. Spam doesn't exist because it doesn't pass the craft filter.
Three commitments:
Wave 2 improvements (coming soon):
Wave 3 future:
Nova's algorithm doesn't reward gaming. It rewards writing.
Write something genuinely good. Finish it. Make readers care. That's it.
The platform handles the rest.
Not really. You can try buying fake engagement, but it gets weighted down. You can try spamming keywords, but semantic analysis sees through it. You can try rapid-publishing trash, but craft scoring tanks immediately. Your best strategy? Write well.
Nova doesn't penalize niche work. A 500-person cult following a dark academic fantasy with deep prose matters more than 50,000 casual clicks on something mediocre. The algorithm finds your audience.
New stories get guaranteed initial readers in week one. If they're well-written and readers finish them, visibility grows. Most good stories see meaningful traction by week 3-4. If nothing's happening by week 6, the story probably needs revision.
Usually because recency matters—newer stories get the newness boost, so older work naturally slides. Or engagement might have plateaued. Check your drop-off points in your dashboard. That's often where improvement happens.
Yes. We see aggregate metrics (80% completion rate) but never who specifically is reading. Authors get dashboard data. Readers remain anonymous. Everyone's privacy is protected.
Read the technical deep-dive → if you want the math behind these scores.
Questions? Email us → We're happy to explain more.
Note: These statistics are bound to change as our algorithm evolves. The current state, although viable, should not be your only source of analysis.
Quality rises. Everything else fades away.