How a Solo Founder Recruits 30 Beta Users from LinkedIn in 10 Days
Beta users are the single highest-leverage input in a pre-launch product. Get 30 of the right ones and the feedback loop tightens to days. Get 30 of the wrong ones and you spend three months building features that no one will pay for. Most solo founders run a six-week "DM everyone I vaguely know" routine and end up with 8 lukewarm testers. This is the 10-day version that gets you to 30.
The routine assumes you have a working MVP, a sharp ICP definition, and a LinkedIn account you're already signed into. It does not assume a network, a list of warm intros, or a budget for ads. The total cost is roughly 420 LinkFetch credits, which at flat per-call pricing comes in well under most teams' monthly Sales Navigator seat cost (snov.io, 2026).
Why "DM everyone I know" is the wrong shape
The default solo-founder beta-recruit motion is to message every LinkedIn 1st-degree connection with a short pitch and a Calendly link. It gets a 4 to 8% reply rate on a generous day, and the replies skew toward friends who feel obligated rather than users in the actual ICP. You'll close 8 testers, half of whom never log in after the first call (rachelandreago.com, 2026).
The shape problem is that the 1st-degree network is the wrong universe. It optimizes for relationship strength when you need ICP fit. The 10-day routine flips the universe: search the entire LinkedIn graph for people who match the ICP, then use relationship strength as a tie-breaker rather than the primary filter. With LinkFetch (a compliance-first LinkedIn data API) plus Claude wired through MCP, that flip takes one afternoon to set up and four hours a day to execute.
A second shape problem: most founders message a candidate once, don't hear back, and write them off. The reply rate on a single DM to a cold LinkedIn contact in 2026 sits around 6%. Three touches across LinkedIn, a comment on a public post, and an email if you can find one pushes that to 18 to 24% (snov.io, 2026). The routine below builds in the second and third touch from day one.
Day 1 to 3: build the candidate universe
The goal of the first three days is a candidate list of 250 named people who match the ICP, with public LinkedIn URLs and enough signal to write a non-generic first message. Three hundred is the upper bound; fewer than 150 and the funnel won't produce 30 conversions.
Run this prompt in Claude Desktop with the LinkFetch MCP server connected:
Apply our ICP rubric to a fresh search of LinkedIn profiles in [target geography] with current title containing [primary title] or [secondary title], current employer headcount between [N1] and [N2], and tenure in current role under 24 months. Use linkfetch.profiles and linkfetch.companies. For each match, return: name, public LinkedIn URL, current title, current company, headcount band, and one specific recent signal (a post, a job change, a hiring action). Skip anyone in our exclusion list (paste below).
Expect 250 to 400 candidates back depending on how tight the ICP is. Credit cost lands around 180 credits for a list of this size at LinkFetch's flat per-request rate. Keep the raw output in a Google Sheet, you'll filter twice more before you message anyone.
Day 2 is enrichment. Re-run linkfetch.profiles on the top 150 by manually-judged ICP fit. The second pass pulls deeper signal: who they worked with at their last job, public talks they've given, the last three companies they followed. That's the fodder for personalization in day 4 to 6 and it's the lifeblood of the 38% reply rate.
Day 3 is the rank. Take the enriched 150, paste into Claude with a ranking prompt, and ask it to sort by three weighted criteria: problem-pain density (do they show signs of feeling the problem you're solving), buying-window proximity (are they in the 30-to-90-day window where new hires evaluate new tools), and shareability (do they post publicly, run a podcast, write a newsletter). The top 80 names become the message queue.
Day 4 to 6: write the first 80 messages
Eighty messages over three days is 27 a day. Two and a half hours at four minutes per message, which is the realistic per-message cost when each one references a specific signal. Anything faster reads as templated and dies. Anything slower kills your day.
The prompt that's worked on every beta-recruit cycle we've seen:
For each person in the ranked list (paste below), write a 60-word LinkedIn DM that names one specific recent signal from their profile in the first sentence, names the single hardest problem my product solves in plain English, and ends with a 15-minute ask. Do not flatter, do not mention "synergy", do not name-drop, do not use em dashes. If the signal is thin, say so honestly and ask if they're feeling the underlying problem.
Claude will draft 80 in maybe 12 minutes of generation time. Plan to keep about 50 as written, edit 25, and discard 5 that read wrong. Don't send them as a batch, LinkedIn throttles anything over roughly 20 connection-request-plus-DMs from a fresh-ish account in a 24-hour window (dev.to, 2026). Spread the 80 across three days at 25 to 30 a day.
Day 7 to 8: the second touch
Half of the 80 won't reply on the first DM. The single biggest lift in beta-recruit reply rate is the second touch, and the second touch is not "just checking in". It's a thoughtful comment on a public post the candidate made in the last two weeks. LinkFetch surfaces those posts in the profile enrichment from day 2.
The day 7 prompt:
For every candidate from the message queue who hasn't replied to the day-4 DM, find their most recent public post (within the last 14 days) using linkfetch.profiles. Draft a 30-word comment that engages substantively with the content, not the author. Avoid praise, avoid emojis, avoid the word "great". The comment should ask one specific question.
That comment gets you back into the candidate's feed without re-pinging the DM inbox. About a third of the non-repliers will respond to the comment, and a small subset of those will follow up on the original DM unprompted.
Day 8 is the email backup. For candidates whose work email you can find through your normal stack (Hunter, Apollo, or even a quick LinkedIn-bio scan), send a short follow-up email that references the LinkedIn touch but doesn't apologize for it. Email reply rates on a warm-from-LinkedIn first send run roughly 22 to 28% in 2026 (rachelandreago.com, 2026), which is meaningfully higher than the LinkedIn DM-only channel by itself.
Day 9 to 10: book and confirm
By day 9 you should have somewhere between 35 and 55 positive replies across the channels. The funnel goal is 30 confirmed testers, so plan to convert about 60 to 75% of positive replies into a calendar booking. The remaining gap closes naturally over the next two weeks as second-touch replies trickle in.
The booking-conversion prompt:
For each positive reply in the response sheet, write a 40-word Calendly-link message that confirms the time commitment (15 minutes), names what we'll cover (their current workflow, the pain points, no demo), and offers two specific time windows in the next 72 hours. Do not write "looking forward to it" and do not write "let me know what works best".
The 15-minute discovery format is load-bearing. A 30-minute slot gets half the booking rate at this stage because the candidate hasn't committed yet, and a "let me know when works" close-line gets a third of the rate of two specific windows. The asymmetric upside is huge: every confirmed call at this stage is an hour-saving versus chasing time-window negotiations over DM.
| Channel | Reply rate | Of replies, % that convert to booked call |
|---|---|---|
| LinkedIn DM (first touch) | 14–18% | 55–65% |
| LinkedIn post comment (second touch) | 28–34% on the comment | 18–24% of those into a DM |
| Email follow-up | 22–28% | 60–70% |
| Combined funnel through day 10 | n/a | 30 confirmed testers from 80 messaged |
What goes wrong and how to fix it on the fly
Three things break this routine in practice. The first is an ICP that's too broad. If you can't write a one-sentence buyer profile ("a head of RevOps at a 200 to 600 employee B2B SaaS company who just inherited the data stack from a predecessor"), the day 1 pull returns 800 candidates and the rank prompt can't differentiate. Tighten the ICP before day 1, not during.
The second is reusing a generic message template even for the top ten candidates. The whole reply-rate gain comes from the signal-anchored first sentence, and the signal-anchored first sentence requires actually reading the day-2 enrichment output for each top candidate. Founders who skip that reading skim hit the same 4 to 8% reply rate as the "DM everyone I know" baseline, which defeats the point.
The third is treating beta-user calls as demos. A discovery call is about their pain, their current workflow, their last failed attempt to solve the problem. If you find yourself opening with a screen share inside the first 90 seconds, the candidate will give polite feedback and disappear. Save the demo for call two.
FAQ
How is this different from a typical "cold outbound" routine?
The shape is similar (filter, enrich, rank, message, follow up) but the target output is different. Cold outbound is optimizing for a booked sales call against a price-tagged product. Beta-user recruitment is optimizing for a research conversation, with no ask beyond 15 minutes of their attention. The lower commitment lets you target a wider universe, which is why the candidate count starts at 250 rather than the 80 to 100 that works for sales outbound.
What if I don't have a working MVP yet?
Run the same routine, but pitch the conversation as customer discovery rather than beta. The 15-minute ask becomes "I'm researching how teams currently handle [problem]" rather than "I'd love your feedback on the prototype". Reply rate goes up by roughly 5 percentage points at this stage because the ask is purely about their expertise, but the conversion to a long-term tester goes down. Use the no-MVP version to refine the ICP, then run the full version once the prototype exists.
Is messaging 80 strangers on LinkedIn allowed under their terms?
Sending a connection request with a short, personalized note to people who plausibly share a professional interest is well within LinkedIn's terms. The rate ceiling matters more than the volume ceiling: 25 to 30 connection requests in a 24-hour window from a mature account is generally fine, but a fresh account sending 80 in one day will trigger a verification step. LinkFetch reads only what you can already see logged in, so the data-access side is clean by construction.
How do I keep the 30 testers engaged after the first call?
Send a single, named-channel asynchronous update every two weeks (loom video, email digest, or a private Notion page they have read-only access to). Resist the temptation to build a Slack community in the first six weeks: the management overhead is higher than the signal you get back. The first 30 want to feel heard, not managed, and a regular update plus a quick reply to any DM is enough.
What if my ICP is consumers rather than B2B?
This routine assumes the buyer has a public LinkedIn profile with work signal on it. For consumer-product beta recruitment, the channels shift: Reddit, Discord, and topic-specific newsletters will outperform LinkedIn by 3 to 5x on reply rate, and the LinkFetch surface is less useful since consumer profiles have less work signal to rank on. The filter-rank-message-follow-up shape still applies, but swap the data source.
Last updated: 2026-05-25 · Author: LinkFetch team