Can you get customers on LinkedIn without an audience? One in five strangers I messaged wanted to talk

You built the product. You posted about it. Three people liked the post, nobody bought, and the advice you get is to build an audience first. This is what happened when I skipped that step: between September 15 and October 9, from my own LinkedIn account, I sent 133 connection requests to people I had never spoken to, each with a short note written from their profile. 72 accepted, 45 wrote back, and 25 wanted to talk: they wrote back again and did not say no. No audience, no ads, no waiting to be found.

Do you need an audience to get your first customers?

No: from a LinkedIn account with no audience, 133 connection requests to strangers between September 15 and October 9, 2026 produced 72 acceptances, 45 replies and 25 people who wanted to talk, as counted on October 14. You need a list of people who have the problem your product solves, and a reason to write to each of them.

The other 15 replied without a yes or a no yet, two of them in the last two days and not yet read. The 60 profiles that were read and got no request did not fit; they were never messaged.

An audience brings people who are curious about you. A list brings people who have the problem. The second group answers a direct message from a stranger at a rate the first group never will, because the message is about them. What those replies looked like, from the polite no to the ten-message thread, is its own post.

How many people do you have to message to get a real conversation?

About five: 25 of the 133 requests turned into someone who wanted to talk, one in five, as read on October 12, 2026, so a list of fifty people is ten conversations, and a list of a hundred is twenty. That is the arithmetic of the first customers: not a launch, a list.

The list is the work. Every one of the 133 was chosen after reading the profile, and every note said something that only applied to that person. The people who would have received a generic message are not in this table, because they were never sent one.

How fast does it add up?

In four weeks, 133 requests had produced 71 acceptances and 41 replies by October 11, 2026. Week by week, with the running total, as counted that day:

Two weeks in: 84 requests, 35 accepted, 18 replies. Three weeks in: 107, 57, 34. Acceptances and replies trail the requests by about a week, which is why the third week collected more replies than it sent requests, and why the fourth looks thin: it is not finished yet. The current week-by-week table, with the requests sent since, is in the benchmark.

How long does it take to hear back?

Of the 41 people who had replied by October 11, 2026, 26 did so within a day of my first message. The answer is days, not weeks.

The median was 13 hours; the slowest took ten days. Connection requests themselves are accepted within a day in two thirds of cases, which is its own post. From the first request to the first person saying "let's talk", the clock runs in days, not quarters.

How much of your time does it take?

Between September 16 and October 11, 2026, 21 of the 26 days had something for me to do, as counted on October 11. On a typical one of those days I approved seven drafts; on the busiest, 25. Over the whole period:

Reading a draft and deciding takes well under a minute, so a typical day was a few minutes and the busiest under half an hour. That is an estimate from the counts, not a stopwatch. The shape is the point: the work is deciding, not writing, and deciding is fast when the draft is about the person in front of you.

Is it the product or the marketing?

In this batch, 25 of the 72 strangers who accepted wanted to talk and 5 said no, as read on October 12, 2026, and none of them had heard of the product before the note. Before this, I could not have told you. "Nobody is buying" usually means nobody who has the problem has heard from you, and from the inside that looks exactly like "nobody wants this".

The list settles it. If one in five strangers with the problem wants to talk, the problem is real and the offer lands. If a hundred people with the problem get a personal note and almost none want to talk, you have learned something about the offer, in three weeks, for free. Either answer is worth more than another month of posting.

What to do with this

1. Write down who has the problem: a job title and the situation they are in the week they would buy.
2. Find a hundred of them on LinkedIn and read each profile before deciding to write.
3. Send each one a connection request with a note about them, not about you.
4. Count after a week, not after a day: acceptances and replies keep arriving for a week.
5. Judge the product after a hundred, not after ten.

Scout does the finding, the reading and the first draft, from your computer, on your account; approving each note and having the conversations stays with you.

How these numbers were measured

• Source: Scout running on my own LinkedIn account, prospecting for real, for the products I sell. No data from other users of Scout; the app keeps its data on each user's computer and sends nothing to us.
• What was counted: profile read, opened and read before any decision; connection request, sent with a note written from the profile, to someone who was not a connection; accepted, the request was accepted; replied, the person wrote back at least once after accepting; wanted to talk, the person wrote back at least twice and did not say no; said no, the person declined politely or asked not to be contacted; the other replies count as undecided.
• When: requests sent between September 15 and October 9, 2026; acceptances, replies and refusals counted on October 14; "wanted to talk" as read on October 12. The weekly table and the reply-time table are as counted on October 11; the weekly table uses seven-day blocks from September 15, and the current week-by-week numbers live in the benchmark.
• Reply time: from my first message after the acceptance to the person's first reply.
• My time: the app's record of what I did, from September 16 to October 11. The minutes are an estimate from those counts, not a measurement.
• Size and rounding: 133 requests. Rates are rounded to whole percentages. "One in five" is 25 of 133.
• Left out: people who were already connections. Nobody is named; only aggregates.
• The rules that apply to every post are in How the numbers on this blog are measured.

This is a living page: the numbers grow with each week of requests, and the changelog below says what changed.

Changelog

• 2026-10-11: first version. 193 profiles read and 133 requests sent between September 15 and October 9, with acceptances, replies, interest and my own actions counted on October 11.
• 2026-10-12: acceptances 71 to 72 and replies 41 to 43, after one more acceptance and two more replies. "Wanted to talk" and "said no" are now read by hand from the conversations (25 and 4) instead of the app's labels (27 and 11), with 14 undecided. The weekly and reply-time tables stay as of October 11.
• 2026-10-14: replies 43 to 45 and refusals 4 to 5 on the same 133 requests; acceptances unchanged at 72. The "wanted to talk" reading stays as of October 12. The weekly and reply-time tables now live in the benchmark, updated weekly.
• 2026-10-16: each question section now opens with its number, base and count date in one sentence; numbers unchanged.