Community playbooks
Hacker News B2B leads: a practical guide
Hacker News B2B leads are real and rare. A handful of threads a month deserve a human reply, on a site whose own moderators will ban your domain for treating it as a channel.
I wanted to know how few. Not a vibe, an actual count.
So we queried the Hacker News Algolia search API for calendar July 2026 and counted the whole site: 314,410 comments. Five of the phrases every monitoring tool tells you to watch appeared in 418 of them, 0.13%. Exactly one comment in July contained "looking for a tool."
That's our own measurement, taken August 5, 2026, and reproducible by anyone with curl. Most writing about Hacker News as a lead source promises volume. There is none. What there is, a few times a month, is a technical buyer describing your problem in their own words.
Does Hacker News allow lead generation?
Reading the site is unrestricted, free, and needs no key. Promotion is where HN draws hard lines. The guidelines ask you not to use the site primarily for promotion, the FAQ bans soliciting upvotes, and moderators enforce both by hand. Listening breaks nothing. Campaigning is what gets removed.
The HN guidelines sentence is short:
Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.
Note the shape of that. It's a ratio rule with no ratio, and the judgment belongs to a moderator rather than a formula. The HN FAQ is blunter on vote manipulation, answering "Can I ask people to upvote my submission?" with:
No. Users should vote for a story because they personally find it intellectually interesting, not because someone has content to promote. We penalize or ban submissions, accounts, and sites that break this rule, so please don't.
Read the sanction unit in that last sentence. Submissions, accounts, and sites. Reddit suspends accounts. HN can put your domain on a list, and a new account posting the same domain doesn't fix it.
Enforcement is public too, sitting in the same searchable index as everything else. In November 2021 dang wrote:
Single-purpose accounts aren't allowed here, and neither are trollish usernames (which celebrity names usually are), and neither is using the site primarily for promotion, so I've banned this account.
In August 2024 he told a different user "The other 98% or so of your posts have been promotional in some way. This is far over the line at which we ban accounts on HN," and then declined to ban that one. Both are moderator opinion rather than written policy, and that's the point. The policy here is thin, the moderation is thick, and all of it is searchable by anyone evaluating your company.
How many Hacker News B2B leads does a month produce?
Fewer than any tool's marketing implies. Calendar July 2026, measured against the Algolia index on August 5: 314,410 comments and 31,299 stories across every topic on the site. Five common intent phrases appeared in 418 of those comments combined. Not 418 in your category. 418 across the entire site.
| Exact phrase, in comment text | July 2026 comments | Share of the month |
|---|---|---|
| "alternative to" | 297 | 0.094% |
| "switched from" | 82 | 0.026% |
| "any recommendations" | 22 | 0.007% |
| "migrating from" | 16 | 0.005% |
| "looking for a tool" | 1 | 0.0003% |
| All five combined | 418 | 0.13% |
The method matters as much as the numbers. These are exact-phrase queries (advancedSyntax=true with the phrase quoted), and the denominator is a sum of day-scoped queries: ask Algolia for a whole month at once and it returns an estimate rather than a count, flagged exhaustiveNbHits: false. Day-scoped queries come back exhaustive. Drop the quotes and Algolia matches the tokens separately, returning anywhere from twice to thirty times more. Over the trailing 12 months to July 30, 2026, the quoted queries returned 95 comments for "migrating away from", 60 for "we switched from", and 26 for "looking for an alternative to", all exhaustive per the index. The unquoted versions of those same three returned 455, 1,041, and 590. Unlike the quoted counts, those three are estimates the index does not certify. If a vendor publishes a Hacker News mention count without publishing the query, assume the larger number.
That's dozens per year for a given phrase. Put it next to the platform everyone starts with:
| Hacker News | ||
|---|---|---|
| Public volume | 31,299 stories and 314,410 comments in July 2026, our own count | Roughly 500,000 posts and 4,000,000 comments per day, per F5Bot's author in 2018 |
| Cost to read | Official API documented as "There is currently no rate limit", no key | Free Data API tier of 100 queries per minute per OAuth client, 10 without |
Reddit is where volume lives, and finding leads there is a different job. Hacker News is a precision surface.
Which Hacker News threads carry the signal?
Not the front page. Ask HN questions where someone lays out a stack problem, and the comment layer under a Show HN, where readers volunteer what they use today and what it costs them. Comments outnumber stories on HN by roughly ten to one, and the counted intent phrases live almost entirely in that layer.
Ask HN is the site's solution-request feed. Someone describes a constraint, names what they tried, and asks the room. The class of buying signal is the same one you'd look for anywhere, with a higher technical floor and less patience for a pitch.
Show HN is more interesting and almost nobody works it. When a competitor launches, the comment thread fills with people saying which product they run now, what broke, and what they would need before switching. A thread about someone else's product is often a better list than a thread about yours.
A representative Ask HN comment, constructed rather than quoted. No handle, no real thread:
"We're on a self-hosted setup that's been fine at 20 engineers and is falling over at 60. The per-seat pricing on the obvious commercial option is more than our staging environment costs. Has anyone migrated off it recently and been happy? Mostly need SSO and audit logs that a compliance reviewer will accept."
Read what the buyer handed over. Current tool, a growth trigger with a number attached, the objection (per-seat pricing), two hard requirements (SSO, audit logs), and a preference for hearing from someone who's done the migration. Nothing in a firmographic database gets you that. A reply from someone who has run that migration is welcome. A reply opening with a product link is flagged.
What converts on Hacker News, and what gets you flagged?
A specific, technical answer from someone who has solved that exact problem converts, with the affiliation disclosed and the link last or not at all. The failure mode is not a warning. It's a quiet penalty: a flagged submission, a killed account, or a domain that stops working, applied by a moderator whose reasoning is searchable.
Three things separate the reply that works from the one that gets removed.
Answer the question that was asked, including the parts your product doesn't solve. HN readers are the buyers, and they can tell within a sentence whether you read the thread or matched a keyword.
Disclose early and plainly. "I work on X" costs you nothing here and buys you the right to be specific about your own product. Omitting it turns a helpful comment into an astroturfing accusation. That's the accusation HN escalates hardest.
Accept that there is no volume version of this. HN has no direct-message feature, so there is no inbox to work. The only contact channel is whatever a user published in their profile, and the public reply is the outreach. That constraint is the source's best property. A spam strategy here is structurally impossible.
Then there's the thing that ends companies here. Don't solicit votes, don't seed comments from accounts you control, and don't ask your team to upvote a submission. That's the rule carrying the domain-level sanction, and the downside isn't a lost thread. It's being unpostable on the site where your buyers read.
If you want a promotional surface, HN gives you exactly one. Show HN "is for something you've made that other people can play with," and its rules rule out as "[o]ff topic: blog posts, sign-up pages, newsletters, lists, and other reading material." It also asks you to remove barriers "such as signups or emails," and bans asking friends to upvote. A landing page is not a Show HN.
How do you monitor Hacker News without drowning in it?
Two saved exact-phrase queries plus your competitors' names covers most of what exists, and the data is free. The Hacker News API needs no key and its README states "There is currently no rate limit." The hard parts are not coverage. They're triage, and remembering which thread you already answered.
Free alerting is where this usually breaks. F5Bot is genuinely good, free, monitors HN, and its own FAQ tells you it's the wrong tool for this job: "F5Bot is designed to alert you about fairly uncommon keywords or phrases." Intent phrases are common phrases. Its tiers page documents a per-keyword and a whole-account daily cap, 10 and 20 on the free tier. A cap that fills produces silence, not an error, so you can't tell "nothing happened this week" from "the cap ate it."
Paid coverage exists and is cheap. Syften monitors Hacker News alongside Reddit, GitHub, Stack Exchange, Lobste.rs, and a dozen more, from $29.95 a month. Anyone claiming these sources are unwatched is selling you something. Breadth isn't the scarce thing.
Memory is. Nothing in the standard toolchain records which thread you already replied to, so the same conversation resurfaces on every run and someone on your team eventually messages a person a colleague already answered. That's the gap LeadSurface fills. It watches Hacker News, Reddit, GitHub, Stack Exchange, and Discourse forums, classifies each post instead of pattern-matching it, deduplicates by URL per account, and hands you the few worth a reply with the source thread attached. A human writes the reply and sends it. On this site that's the only approach that survives the moderators.
You can test the idea for free this afternoon. Search the Algolia index for your closest competitor's name, quoted, over the last 90 days, and read every hit. There won't be many. That is the finding.
Hacker News won't fill a pipeline. It will hand you a few threads a month where someone described the problem you solve, in public, in their own words, and it will remove you for arriving as a campaign instead of a person.
A handful of threads, answered like a human. That's the whole channel.