Targeting is the part your competitors have already done as well as you have; timing is where the variance is. Only about 5% of your market can buy in a given quarter, and 94% of buying groups rank their preferred vendor before they ever speak to a seller — so the shortlist forms before the conversation you are trying to book. Our analysis of 429 public buying questions found the median first recommendation arrives in 21 minutes and the median asker gets just three answers. Big companies answer this with brand advertising to stay in memory. If you cannot afford memory, buy detection instead: write a trigger list of the public events that move someone from the 95 into the 5, and check it daily rather than monthly.
We collected every public “which tool should I use?” question we could find on Hacker News over the last three years — 429 of them — and timed how long it took for the first real recommendation to arrive. The median was 21 minutes. Two thirds had an answer inside the hour. By the time you read one of those threads over coffee the next morning, somebody else has already written the shortlist.
This is the part of finding customers that most advice gets wrong. Nearly all of it is about targeting: define your ideal customer, build the list, write the sequence. Targeting is worth doing. It is also the part your competitors have already done about as well as you have. The variance is not in who you picked. It is in when you arrived.
The window is about twenty minutes wide
A public request for a recommendation is the purest buying signal there is. Someone has decided to spend money, admitted they do not know what to buy, and asked strangers in writing. So we measured what actually happens to those requests.
Median time to first recommendation21 minutes
Answered within 15 minutes39% of questions
Answered within 1 hour69%
Answered within 24 hours97%
Replies arriving in the first 24 hours79% of all 6,008 replies
Median number of answers the asker receives3
Questions getting 3 answers or fewer61%
The first number is the one people expect. The last two are the ones worth sitting with. The median person who publicly asks what to buy gets three answers, and six times out of ten they get three or fewer. This is not a crowded auction where your product fights for attention against forty others. It is a short conversation that starts within minutes, fills up within a few hours, and then goes quiet for good.
Being one of three named options is enormous leverage. The cost of entry is showing up the same day. The cost of missing it is total, because a thread that is two days old is an archive, not an opportunity — 98.6% of all replies had landed within a week.
- Sample
- 429 “Ask HN” threads published between September 2023 and September 2026, collected via the public Hacker News search API and filtered to questions asking which product, tool, or service to use.
- Measure
- Minutes between the question being posted and the first top-level reply. In 86% of threads that first reply named or linked a specific product, which is what makes it a fair proxy for “the first recommendation”.
- Caveats
- Hacker News is fast, technical, and self-selecting, so treat these as a best case for response speed. A quieter industry forum will be slower — but the shape holds, and so does the reason for it.
Why the window is so short everywhere else too
Twenty-one minutes is a Hacker News number. The underlying mechanism is not. It shows up in the research on how businesses buy, and it has a shape worth understanding.
Start with the constraint everyone ignores. Work by John Dawes at the Ehrenberg-Bass Institute, published through the LinkedIn B2B Institute, points out that companies replace major services — software, banking, legal, telecoms — roughly every five years. That means about 20% of your market is in play in a given year, and only about 5% in a given quarter. Ninety-five percent of the people on your carefully built list are not ignoring you. They cannot buy from you today at any price, however good your email is.
Now look at what the 5% actually do. 6sense surveyed more than 4,000 B2B buyers for its 2025 Buyer Experience Report and found that 94% of buying groups rank their preferred vendors before they ever contact a seller, and that the vendor they preferred before that first conversation goes on to win the deal 77% of the time. Gartner’s work points the same way: buying groups of six to ten people spend just 17% of their total purchase time with all potential suppliers combined, which leaves roughly 5% for any one of them.
The meeting you are trying to book is not where the decision gets made. It is where a decision that was already made gets confirmed.
Put those together and the Hacker News thread stops looking like an internet curiosity and starts looking like a time-lapse of the whole process. Somebody enters the market. A shortlist forms fast, from whatever is in front of them. Almost everything that follows ratifies it. The only difference online is that you can watch it happen and read the timestamps.
Big companies buy memory. You have to buy detection
Here is where the standard conclusion and the useful one part company.
The 95:5 rule is nearly always cited as an argument for brand advertising. If most buyers cannot act now, advertise to them anyway so that your name is the one they recall when their window opens. That is sound, and if you run marketing at a company with a real budget you should do it. Memory is an asset that pays out years later.
It is also the single most expensive asset a small company can try to buy, and the slowest to arrive. So take the same data and draw the opposite tactical conclusion. If you cannot afford to be remembered when the window opens, you have to be there when it opens. You cannot buy memory. You can buy detection, and detection is close to free.
Practically, that means writing the document nobody writes. Most founders have some version of an ideal customer profile. Almost nobody has its other half — a trigger list. The profile tells you who to talk to. The trigger list tells you when they became reachable.
Write down the five to eight public, observable events that move someone in your market from the 95 into the 5. Not attributes — events, with timestamps. Then score each one honestly on three things: how reliably you can detect it, how fast it decays, and how often it happens.
Highest intent, shortest life. A same-day answer puts you on a three-item list; a week later the thread is closed.
The problem has just become someone’s job. Budget and authority are in motion but requirements are not fixed yet.
Lowest urgency, highest volume, and easy to plan around. This is where a small company can be genuinely early.
The scoring matters more than the list. A trigger you cannot detect is a wish. A trigger that decays in hours is worthless unless you are actually watching daily. A trigger that almost never fires will not fill a pipeline no matter how well it converts. The triggers worth building a routine around are the ones that score at least adequately on all three.
Speed to signal, not speed to lead
There is a well-known study about response time that almost everyone applies too narrowly. In 2011, researchers audited 2,241 US companies for Harvard Business Review and found that firms contacting a web lead within an hour were about seven times more likely to qualify it than those that waited just sixty minutes longer, and roughly sixty times more likely than those who waited a day. The average response time was 42 hours. Twenty-three percent never replied at all.
It is an old study and it is about inbound forms, so it usually gets filed under “answer your leads faster”. That is the small version of the idea. The useful version is that the clock is not started by the form. It is started by the signal — and unlike a form submission, a public signal is visible to your competitors too.
This is also what makes the arithmetic survive contact with reality. Instantly’s 2026 benchmark report, drawn from a year of cold email data, puts the average reply rate at 3.43% — down from around 8.5% in 2019. At that rate, thirty untargeted messages buy you one reply. Fifty messages sent because something specific and recent happened do not need to be ten times better to win; they need to beat a baseline that is quietly collapsing under the weight of everyone else’s automation.
Two practical notes from the same report. Only 58% of replies come from the first message, so a single send discards nearly half your yield; four to seven touches is the range where follow-up still adds something. And when you do follow up, the trigger gives you the one thing a sequence normally lacks — a reason to write again that is about them.
- The claim
- “Emails referencing buying signals get 15–25% reply rates, a 5x improvement” — widely repeated, and usually attributed to the benchmark report cited above.
- The problem
- We read the report. That figure is not in it. It would have been the most useful statistic in this article, which is exactly why it is worth checking.
- The lesson
- The signal-based selling case is strong enough without it. Be suspicious of any number that conveniently proves the thing its publisher sells — including, in fairness, the 6sense figures above.
Make it a routine rather than a project. Two or three times a week, run your saved searches, look at what fired, and write to the handful of people whose timing just changed. Ten minutes a day beats a monthly blitz, because the triggers at the top of your ladder do not survive until the blitz.
What detection cannot do
One honest limit, because the argument is weaker without it. Detection only wins deals you can see. It does nothing for the buyer whose window opens quietly, who never posts a question, and who writes their shortlist from whatever names they already trust. That is most of the market most of the time, and no amount of monitoring reaches them.
The brand people are right about that. What they are wrong about is the price. The affordable version of being remembered is not advertising — it is being visibly, publicly useful in the places your buyers already read, so that the name in their head when the window opens happens to be yours. That is the same community work as always, done for a different reason and on a longer clock.
So run both, and be clear about which is which. Detection is this quarter’s pipeline and you should measure it weekly. Presence is next year’s shortlist and you cannot measure it at all. The mistake is spending another week refining an ideal customer profile — the one document that does neither — and calling it progress.
Sources
- Original analysis by FindCustomers.org: 429 “Ask HN” product-selection threads, September 2023 – September 2026, collected via the public Hacker News search API. Collection script, raw data, and full output are available on request.
- John Dawes, Ehrenberg-Bass Institute for the LinkedIn B2B Institute — the 95:5 rule. marketingscience.info
- 6sense, 2025 B2B Buyer Experience Report (4,000+ buyers). 6sense.com
- Gartner, B2B buying journey research — buying group size and time allocation. gartner.com
- Oldroyd, McElheran & Elkington, “The Short Life of Online Sales Leads”, Harvard Business Review, March 2011. hbr.org
- Instantly, Cold Email Benchmark Report 2026 (data from 1 January – 18 December 2025). instantly.ai