Between May 1 and July 10, 2026, roughly 10,000 outbound messages went out through SalesXMachina (9,925 to be exact): 7,834 emails, 729 LinkedIn direct messages and 1,362 LinkedIn connection requests. They reached close to 6,000 people at roughly 2,500 companies, mostly in the German-speaking market.
We did not just count these numbers, we ran the math: 14 hypotheses, each tested for statistical significance, plus a screening across 17 additional variables. Every single incoming reply was read and categorized until no leftover bucket remained. Here is what we found. No opinions, just data.
What comes back when you send 10,000 messages
- 441 replies in total, and 31.5 percent of them are out-of-office notes
- 209 replies came from actual humans. That is a reply rate of 3.5 percent of contacted people
- The most common human reply by far (40.7 percent): no need right now, check back later
- Roughly 9 percent of human replies show real interest, meaning a request for a meeting or an analysis
Sobering at first glance. But the spread between what works and what does not is enormous. The best segments in our data reply 13 times more often than the worst.
What works: the confirmed levers
1. LinkedIn beats email by a factor of 8
Reply rate per message: LinkedIn direct message 8.2 percent, LinkedIn connection request with a note 4.6 percent, email 0.97 percent. Email made up 79 percent of our volume but delivered only about a third of the replies. Channel choice is the single biggest lever in outbound.
2. An accepted connection request is gold
Contacts you are already connected with (1st degree) reply at 25.2 percent. Everyone else replies at 1.9 percent. That is a factor of 13. Anything that gets your connection requests accepted pays for itself many times over.
3. Researched hooks beat generic signals
Messages built on an individually researched hook hit a 10.6 percent reply rate. Generic triggers like expansion news (1.9 percent) or leadership changes (0.9 percent) fell far behind. It is not the signal that sells, it is the visible homework behind it.
4. Job changers reply almost twice as often
Campaigns targeting people who recently changed jobs reached 4.6 percent versus 3.0 percent for everything else. The effect also shows at the individual level: people less than 6 months into their role reply at 5.9 percent, everyone else at 3.0 percent. New role, new budget, open mind.
5. Fast follow-ups keep conversations alive
When the follow-up after a reply went out within 24 hours, a second reply came in 47 percent of cases. Waiting longer dropped that to 26 percent. Speed after the first reply is one of the easiest levers you can pull.
6. A maintained LinkedIn profile is a buying signal
Contacts with a filled-in LinkedIn bio replied on LinkedIn at 10.4 percent, contacts without one at 6.7 percent. People who maintain their profile are active on the platform, which makes them reachable.
7. The mid-market replies, the enterprise stays silent
Companies with 51 to 200 employees: 4.6 percent reply rate. Companies with more than 5,000 employees: 1.6 percent. If you target both, weight your volume accordingly.
8. Sunday is the underrated sending day
Messages sent on Sundays achieved 4.1 percent versus a weekly average of 2.0 percent, the only weekday with a statistically significant difference. All other weekdays are indistinguishable from each other.
9. The right message length depends on the channel
For email, longer won: 900 to 1,200 characters reached 2.6 percent versus roughly 1.0 percent for shorter mails. On LinkedIn it was the opposite: messages under 900 characters sat at around 10 percent, longer ones dropped to 5.3 percent. There is no universal rule for message length.
10. Guessed email addresses are wasted volume
Verified addresses: 2.1 percent reply rate. Constructed addresses, guessed from name patterns: 0.18 percent. One single reply out of 543 contacts. Without verification, email outbound is busywork.
What does not work: myths that died
Myth 1: Seniority drives replies
No measurable difference between C-level plus VP (3.2 percent) and manager level and below (3.0 percent). C-level alone actually sat below average at 2.7 percent. The idea that you always have to start at the very top does not survive contact with the data.
Myth 2: People reply in the morning and in the evening
Normalized per hour, at least as many replies arrive during the day (10 am to 5 pm) as in the morning, and clearly fewer in the evening. The magical morning reply slot does not exist in our data.
Myth 3: Connection requests get accepted within hours
Half of all accepted connection requests took more than 3.6 days. Only 12.6 percent were accepted within 6 hours. If your sequence assumes fast acceptance, it is planning against reality.
Myth 4: You are either good at outbound or you are not
The distribution of reply rates across senders is not two camps. Most sit in a broad middle, with a few outliers reaching 10 to 15 percent. Outbound is not a talent lottery, it is a craft with a learning curve.
Myth 5: The AI lead score predicts who will reply
The correlation between our own AI score and reply probability was statistically significant but tiny (r equals 0.03). For predicting who replies with genuine interest it was useless. Honesty where it is due: our own scoring does not prioritize well enough yet either.
Two results with a twist
The third sequence step barely generates replies (0.7 percent per message versus 1.9 to 2.8 percent for steps 1 and 2). But the replies it does generate are disproportionately often genuine interest. Cutting step 3 would be the wrong move. It filters out the unreachable and pulls in late but good replies.
Same with the call to action: directly asking for a meeting generates more replies than offering an analysis. But replies to an analysis offer contain real interest four times as often. If you count meetings, ask for the meeting. If you want pipeline quality, offer value first.
One observation on the side
Female contacts replied less often at 2.5 percent than male contacts at 3.6 percent, a statistically significant difference. The gender of the sender, on the other hand, made no provable difference.
Methodology
The basis is anonymized, aggregated data from campaigns that ran through SalesXMachina between May 1 and July 10, 2026: exactly 9,925 messages (7,834 emails, 729 LinkedIn direct messages, 1,362 connection requests) to 5,959 people at 2,518 companies, 441 incoming replies, 209 of them human. Reply rates were computed against the actual sending population, not as distributions across replies. Significance: two-proportion z-tests at the 5 percent level, Wilson confidence intervals, cells with too little data excluded. For the multi-variable screening we additionally applied Bonferroni correction and re-checked channel-confounded effects within each channel. Out-of-office notes, bounces and system mails were excluded from all rates. Correlation is not causation, and some effects (industry, for example) overlap with campaign types.
What this means for your outbound
- Shift volume from email to LinkedIn wherever you can
- Invest in getting connection requests accepted, they compound at a factor of 13
- Research the hook instead of sending generic triggers
- Prioritize contacts who are fresh in their role
- Follow up within 24 hours when someone replies
- Verify every email address before you write to it
- Test Sunday as a sending day
- Write long emails and short LinkedIn messages
- And build a clean resurfacing system for the 40 percent who want to buy later
These learnings are exactly what we built into SalesXMachina: finding signals, researching hooks, choosing the right channel, following up, resurfacing. Automatically, every week, without the energy ever dipping. If you want to see what that looks like for your product, get in touch.
Happy selling, Axel