How Demand4It measures demand

Each Demand Check searches public conversations on Reddit, Quora, Community Forums, Hacker News, Bluesky, YouTube, Stack Exchange, GitHub and Mastodon. Every accepted signal links back to its original post so you can read it yourself.

Search terms bring back many candidates. An AI review then checks each one in three steps: is there real, unmet demand that this specific product could address; which category does it fit; and how strong is it (1–10). Mentioning a keyword or competitor is never enough — the text must contain an exact phrase showing need, pain, switching, a workaround or willingness to pay. Praise for an existing tool without an unmet need is rejected.

Accepted signals are weighted by category and strength:

• Actively Looking — Someone is openly searching for a solution right now.
• Recommendation Request — Asking peers what tool or service to use.
• Seeking Alternatives — Looking to replace something they already use.
• Competitor Frustration — Complaints about an existing provider.
• Problem / Pain — Describing the underlying problem without naming a solution.
• Money Signal — Budget, pricing or willingness-to-pay evidence.
• DIY / Failed Solution — Hand-rolled workarounds that are not working well.
• Feature Need — A specific capability people are asking for.

You can change these weights for your own project. Your Custom Score is shown beside the standard D4I Demand Score; the standard score is never overwritten.

Confidence describes how much you can trust the evidence, not how big demand is. It rises with more accepted signals, more independent sources, less reliance on a single source, more signals with known dates and more signals in the trend windows. Lots of similar posts from one source can't reach High on its own. Fewer than 15 signals is flagged as limited evidence; one source providing half or more is flagged as concentrated.

Signals from the most recent 30 days are compared with the 30 days before. +20% or more is Rising, −20% or less is Falling, anything between is Stable. With too few dated signals the trend shows Insufficient data. Posts without a reliable date are left out rather than guessed.

Money-related signals are split into demonstrated spending (already paying, paying and looking to switch, or paid for something that failed) and stated or hypothetical interest (willingness to pay, price frustration, hypothetical pricing). Demonstrated spending is stronger evidence.

Separate searches look for objections — “free tools are enough”, “wouldn't pay”, “easy workaround”, “happy with what I use”, “not a real problem”. Each is reviewed to confirm it genuinely challenges the idea. Counter-signals are shown next to supporting ones but never change the score.

• It measures observed online conversation, not market size or revenue.
• People who talk online are not a random sample of all buyers.
• Saying you'd pay is not the same as paying.
• A low score can mean the audience talks elsewhere, not that demand is absent.
• AI review can make mistakes — every quote links to its source so you can check.

Real commitment — signups, replies, payments — is the strongest evidence, and is kept separate from observed demand.