A two-person company has two sending profiles and one attention budget. That is the whole input. List size, tooling and copy all sit downstream of those two facts, and the arithmetic that runs from them to a booked call is short enough to fit on one page.
We run the engine on our own pipeline and on client workspaces, so every number below comes out of a send log. Between 29 July and 5 November 2026 it produced 4,889 cold LinkedIn invitations from 35 sending profiles across 15 workspaces, 1,251 acceptances and 234 first replies from people we invited. This article walks that log backwards: capacity, conversion, the honest size of the result, and the human time it costs.
The one number you control is capacity
A profile has a sustained weekly ceiling, and it is lower than the daily limit suggests. Counting every profile-week with 20 or more invitations since 15 August gives 83 profile-weeks across 21 profiles: median 41 invitations, mean 40.2, range 21 to 80. The daily limits set on our 22 active profiles span 1 to 30, so the median week is well inside what the settings allow. Pacing, working hours and the platform's own cooldowns eat the difference.
Call it 177 invitations per profile per month. Two profiles hold roughly 354. The last 30 days confirm the shape at the top: 1,266 invitations across 22 profiles, and the two busiest profiles sent 199 and 152, which is 351 between them. That is the ceiling a two-person company works against, and it moves only by adding a person, adding a channel, or renting an account. Our comparison of what a rented profile buys and charges for covers the third option.
Month one is smaller than 354. Across 25 profiles connected since 1 August, the median gap from connecting an account to its first invitation was 4.63 days, and the worst was 18.9. Add the ramp itself and the first calendar month of a new profile runs at roughly half capacity.
The conversion chain, measured on matured invitations
An invitation sent yesterday has not finished happening. Our cohort rule is to count only invitations at least 14 days old, which for this log means everything sent before 22 October: 4,189 invitations, 1,084 acceptances, 208 first replies.
That gives three rates worth keeping:
- 25.9% of invitations get accepted.
- 5.0% of everyone invited sends a first reply.
- 19.2% of everyone who accepts sends a first reply.
Push 354 invitations a month through those rates and you get about 92 acceptances and about 18 first replies. The reply number is the one a two-person company should plan on, because it rests on 208 events and moves slowly. At $3,000 a month for a managed engine that works out to roughly $8.50 per invitation, $33 per acceptance and $170 per first reply. The fuller version of that calculation, including what a hire costs for the same output, is in our breakdown of cost per meeting from LinkedIn outreach.
Follow-ups sit inside that 5.0%, and they stop paying early. In our log the first three cold touches carry 99% of first replies, and the fourth and fifth together added 0.2 percentage points across 89 attempts. The detail is in why the second follow-up outperforms the fifth. For pipeline math, plan a three-touch sequence and read the 5.0% as its full output.
One qualified conversation a month, and the band around it
Of everyone who has ever replied to us, 21 people reached a qualified conversation or a booking, and 13 of those 21 came from someone we invited. Thirteen events against 4,889 invitations is one qualified conversation per 376 invitations. This is the step where plans usually promise the most and our log supports the least.
Two profiles at 354 invitations a month land on 0.94 of those per month. Thirteen is a small count, so the honest version carries its band: the 95% interval around 13 events runs from about 7 to 22, which puts the monthly figure between one every two months and one and a half per month. Six people reached a booked call, four of them invited by us, which is one booked call per 1,222 invitations, or about one per three and a half months of two-profile capacity.
The close rate lives outside this log: deals that follow a booked call are tracked elsewhere, and six events are too few to divide. So the chain stops at the qualified conversation. Treat the first reply as the unit you measure and the conversation as the unit you forecast loosely.
A third of the result arrives without an invitation
Of 344 people who have ever replied to us, 110 arrived with no invitation from us: existing connections, people who accepted back before the engine started, and people who wrote first. That is 32% of all replies. The pattern is stronger at the valuable end: 8 of the 21 qualified conversations, 38%, came from that group.
For a two-person company this is the most useful line in the log. Over a third of the result comes from a surface that volume math leaves out, and it sits on profiles belonging to real stakeholders: on our rented accounts, all 29 replies came from people we invited first. What a founder's own profile does is laid out in our piece on founder-led outbound. So posting, commenting and answering your own inbox are pipeline work with a measured share of the output, and they deserve a line in the same spreadsheet.
The lag makes a monthly review lie
Acceptances come back fast. On 1,244 matured invitations the median gap from invitation to acceptance was 0.25 of a day, roughly six hours. Of those, 827 landed the same day and 1,089 (87.5%) inside seven days, with the 90th percentile at 8.89 days.
Replies are slower and much wider. On 232 people the median gap from invitation to first reply was 3.13 days, the 75th percentile 11.95 days and the 90th percentile 20.99 days. Counted another way: 149 of those 232 replies arrived inside a week, 185 inside two weeks, 208 inside three. So a month of sending is readable in the month after it, and a cohort read on the last day of the month is still missing a fifth of its replies, with the newest ten days of sending missing much more than that. A two-person company that reads this month's sends against this month's replies sees a thinner number than the one it will finally have, concludes the market went quiet, and rewrites copy that was working. The cohort discipline for this is written out in how to read an outbound weekly report, and the per-step acceptance and reply benchmarks we compare against are collected on our outbound benchmarks page.
The second person's real workload
Capacity planning usually stops at sends. The attention cost sits in the inbox, and it is measurable.
In the last 30 days the engine saw 1,360 inbound messages from 138 people and built 474 reply drafts. Of those drafts, 190, which is 40%, were addressed to machines: out-of-office notices, ticket confirmations, people who left the company. Triage dismissed 328 drafts without sending. Of the 111 that went out, 48 left on the 48-hour timer and 63 left because a person clicked send. Across the whole log, 118 of 360 sent drafts were edited before they went.
So the human load for two profiles at full capacity is roughly 63 send decisions a month, three per working day, plus about 18 conversations to actually hold. That load grows with every added profile, which is why the third profile pays only when a third pair of hands comes with it.
The preparation side is heavier and hides well. In the same 30 days the engine sourced 3,077 people to produce 1,266 invitations, which is 2.4 people sourced per invitation, rejected 887 candidates over 60 days, split 302 batches, and ran 61,220 text checks across 13 rules and 6 surfaces that returned 2,759 findings. Email ran alongside at 310 letters from two mailboxes across seven workspaces. Of 359 decisions logged in those 30 days, 10 were taken by the owner, one human decision per 36 machine ones.
What this gives you on Monday
Write your own version in five lines. Capacity: profiles times 177 invitations. Acceptance: times 0.26. First replies: times 0.05 of invitations. Qualified conversations: divide invitations by 376, and carry the band. Fifth line: the inbound share, kept separate, since sending volume leaves it where it is.
If the bottom line comes out under what the business needs, three levers move it: one more sending profile, a second channel on the same list, or sharper targeting that raises the reply rate at the same send count. Then set the review window two weeks behind the sending window, so the number you read is a number that finished happening.
Questions buyers ask
How many cold messages do I need to send for one meeting?
In our log one qualified conversation came per 376 invitations, and one booked call per 1,222. With two sending profiles holding about 354 invitations a month, that is roughly one qualified conversation a month and one booked call per three and a half months from cold sending alone. The counts behind those ratios are 13 and 4 events, so plan with the band rather than the point.
How many LinkedIn invitations can one profile send per month?
About 177. Our 83 profile-weeks with 20 or more invitations since 15 August give a median of 41 a week, a mean of 40.2 and a range of 21 to 80, on profiles whose daily limits span 1 to 30. A newly connected profile runs lower: median 4.63 days from connection to first invitation, so month one lands near half.
How long before outbound shows results?
Acceptances arrive in hours, with a median of 0.25 of a day and 87.5% inside a week. First replies take a median of 3.13 days, with a 75th percentile of 11.95 days and a 90th percentile of 20.99: 149 of 232 landed inside a week, 185 inside two weeks, 208 inside three. Read any month of sending about two weeks after it ends, and the number will be close to final.
Does outbound work when the company is two people?
It produces about 18 first replies a month on two profiles, and 110 of our 344 replies came from people who arrived with no invitation from us, which is a surface a two-person company is unusually good at. The binding cost is attention: 63 send decisions a month plus the conversations themselves, so the next hour of inbox time is worth more to the pipeline than the next sending profile.