Product guides

How IdeaFast Scores Pain Points

The exact methodology IdeaFast uses to turn raw Reddit and Hacker News discussions into scored, evidence-linked pain points founders can trust.

Last updated July 6, 2026

Quick answer

IdeaFast scans public Reddit and Hacker News discussions, groups posts into pain themes using text embeddings so the math decides what belongs together, and scores each theme from 0 to 100 based on how acute, frequent, and recent it is. Every theme's evidence is literally the posts it was built from, so a quote can never contradict the theme it supports.

Step one: collect real public discussions

A scan pulls recent public posts and comments from the subreddits or communities you specify, or that we auto-discover from your stated interest. Nothing is invented; every discussion used is a real, public post that existed before the scan ran.

Step two: cluster by meaning, not by keyword

Posts are converted into embeddings, a numeric representation of meaning, and grouped by an algorithm rather than a single model call guessing at similarity. This matters because it means a pain theme's evidence is not a loose collection a model thought looked related afterward, it is literally the set of posts the clustering math grouped together. A quote can never drift from the theme it is supporting.

Step three: score each theme

Each cluster gets a score from 0 to 100 based on three factors: how acute the pain is (how strongly people express it), how frequently it recurs across independent posts, and how recent the discussions are. Only clusters with real, repeated signal make it onto a page; one-off rants and self-promotion get filtered out.

Every pain point on IdeaFast links to the real post or comment it came from. This is a deliberate design choice, not an afterthought: research tools that summarize pain into paraphrased quotes ask you to trust a description of a complaint. Linking to the source lets you verify it yourself in one click.

Frequently asked questions

Does IdeaFast use AI to invent pain points?

No. The AI's role is narrow: it labels clusters that an embedding and clustering pipeline already grouped from real posts. Every pain point traces back to actual discussions, never an invented example.

What does the 0 to 100 score mean?

It reflects how acute, frequent, and recent the pain signal is across the discussions a scan found. A higher score means more people are expressing the same specific frustration more recently, not a guarantee anyone will pay for a fix.

Can two different pain themes share the same evidence post?

A single post can touch on more than one theme if it genuinely discusses multiple problems, but the clustering process assigns evidence based on meaning, not by copying the same quote across unrelated themes to pad results.

Where does the data come from?

Public Reddit and Hacker News discussions. IdeaFast does not use private data, paid review databases, or AI-summarized third-party reports.

Related

Skip the manual digging

IdeaFast scans Reddit for you and scores real pain points with evidence. Run your first scan free.