LinkedIn InMail vs Connection Request vs Cold Email: Which Gets Replies

At a glance

What the 2026 benchmark data actually shows

I'm Hlib Storchak. I build outbound systems for B2B founders and sales teams, and I've run all three of these channels for clients, often on the same list in the same quarter. The question "which one gets replies" comes up constantly, and most answers online cherry-pick whichever number makes their preferred channel look best. So here is what the actual 2026 data says, with the source for each number so you can check it yourself.

The cleanest LinkedIn-side dataset available is Expandi's 2026 outreach benchmark, built from 13,218,869 connection requests across 13,302 accounts between May 2025 and April 2026. It puts platform-wide connection acceptance at 28.5%, connection-note reply at 3.0%, and reply to the message sent after someone accepts at 10.4%. That last figure, the 10.4% message reply rate, is a blended number across every type of outbound message LinkedIn tracks in that category, including InMail, follow-ups, and group and event messages, not an isolated InMail figure. I want to be upfront about that limitation rather than presenting a blended number as an InMail-specific one.

On the email side, Instantly's 2026 Cold Email Benchmark Report, built from billions of tracked sends, puts the average cold email reply rate at 3.43%, with the top 25% of campaigns clearing 5.5% and the top 10% clearing 10.7% or higher. That is the number I use as the baseline for any cold email projection I build for a client, because it is the largest, most current dataset that discloses its own sample.

InMail: a good response rate on a small, capped pool

InMail's real advantage is reach: it lets you message someone you are not connected to, without the acceptance step a connection request needs first. That is genuinely useful for a named list of senior, hard-to-reach targets who you know will not accept a cold connection request from a stranger.

The constraint is volume. LinkedIn's own help pages confirm Sales Navigator Core, Advanced, and Advanced Plus all receive 50 InMail credits a month per seat, with unused credits rolling over up to a cap of 150 total, and each credit expiring 90 days after it is issued. Fifty messages a month is roughly two a working day. That is not a channel you scale into your primary pipeline source; it is a channel you point at your highest-value accounts and nowhere else. If a rep is burning InMail credits on a broad, low-fit list, that is the first thing I fix when I take over an account, because the credits are too scarce to spend on anyone outside the actual ICP.

Connection requests: cheap, personal, risk-limited not credit-limited

Connection requests cost nothing to send and, unlike InMail, are not metered by a credit system. What limits them is risk: LinkedIn's own Help Center confirms it never publishes an exact daily or weekly cap, and pushing volume too far is what gets an account restricted, not a hard published number. Expandi's 2026 data (the same 13.2 million request sample above) puts average acceptance at 28.5%, ranging from roughly 17.5% up to 40.1% depending on industry, and shows the reply rate to your first connection note itself has fallen from 3.5% in May 2025 to 2.2% by April 2026, while the reply rate to the message you send after someone accepts held much higher. The practical read: spend less effort perfecting the connection note itself and more on what you send once someone says yes.

A separate, independent dataset worth knowing here: Belkins' 2026 LinkedIn Outreach Study, built from 11.5 million connection requests, found personalized notes get a 55% higher reply rate than blank ones (8.2% vs 5.3%) at a small cost to acceptance rate. I covered that trade-off and the message frameworks that make the most of it in a separate article, so I will not repeat the full breakdown here.

Cold email: the only channel built to scale past a few dozen touches a day

Cold email replies at the lowest rate of the three per touch, 3.43% on average per Instantly's 2026 report. What it has that neither LinkedIn channel does is volume with no seat-based credit ceiling: a handful of properly warmed mailboxes across a few sending domains can carry hundreds of sends a day without touching a platform-imposed cap the way InMail does. That volume is what makes cold email the channel you can actually A/B test at a pace that produces a statistically meaningful answer within weeks, not months.

This is the sending stack I run for clients by default, mostly because I already know its quirks: Salesforge for sequences and sending, paired with Mailforge and Infraforge for the domains and mailbox infrastructure behind it. That is a personal preference built on familiarity, not a claim that it out-performs Instantly, Smartlead, or Lemlist on the merits; any of those is a reasonable choice if your infrastructure and warmup process behind it are solid, and the platform itself matters far less than domain health and list quality.

What 100 meetings a month costs by channel: the maths

Cost models like this are only useful if you can see and swap every input, so here are the assumptions, stated plainly, for a rough monthly cost-per-meeting by channel. Treat every number below as an assumption you should replace with your own before you act on it.

Run the arithmetic and the picture is consistent with what the volume caps already tell you: InMail cannot get you to 100 meetings a month on its own math, not because the response rate is bad but because the pool is 50 messages. Connection requests get closer but are still bounded by acceptance-rate risk if you push volume. Cold email is the only channel in this table whose monthly touch count you can multiply by adding mailboxes and domains, which is why it tends to carry the bulk of a 100-meetings-a-month target, with LinkedIn as the channel that adds warmth and a second touchpoint rather than carrying volume on its own. This is the same model I build with clients before we split a monthly meetings target across channels: put the assumptions on the table first, then argue about the numbers, not about which channel "feels" better.

Restriction risk: what actually gets an account flagged

InMail cannot really get you restricted for volume, since the credit system caps it before you could send enough to trigger a pattern flag. Connection requests are the opposite: because there is no published cap, the actual risk is sending too many too fast, especially with a low acceptance rate, which LinkedIn's own systems read as spam-like behavior. Cold email carries a different risk entirely, deliverability rather than account restriction: send too fast on a cold domain, get too many spam complaints, or skip authentication, and your messages stop reaching the inbox long before any human decides not to reply. All three risks are manageable. None of them are eliminated by a bigger subscription tier; they are eliminated by pacing and by warming whatever you are sending from, whether that is a mailbox or a LinkedIn seat.

Which channel fits which ICP and deal size

InMail earns its scarcity when the account list is short and the deal size is large enough that two or three thoughtfully placed messages a day to named enterprise targets beats broad volume. Connection requests fit best in the middle: mid-market and SMB targets where a real relationship, even a shallow one, meaningfully changes whether the eventual pitch lands. Cold email fits best wherever the addressable market is large enough that you need volume to find the fraction of it that is in-market right now, which describes most SMB and mid-market motions and a meaningful share of enterprise top-of-funnel too.

Sequencing them together beats picking one

None of this is really a "pick one" decision in practice. The setups that produce the most consistent numbers for clients run connection requests or InMail first to establish a name a prospect recognizes, then follow with cold email carrying the bulk of the volume and the follow-up cadence, sometimes closing with a second LinkedIn touch once the email thread has gone quiet. Each channel does the job it is actually good at instead of being asked to do all three jobs at once.

The mistake I see most often when a team leans on one channel

The mistake I see most often when I take over an account is a team that picked one channel because a rep liked it, then tried to force that one channel to carry a volume target it was never built for, usually InMail stretched across a list far broader than 50 credits a month can responsibly cover, or a single cold email domain pushed past what a fresh domain can survive. The fix is almost never a better message. It is matching each channel's actual capacity to the part of the funnel it is suited for, then building the volume you need from the channel, cold email, that is actually built to scale.

Which I reach for, and when

For a named enterprise list under a few hundred accounts, I reach for InMail first, spent deliberately since there are only 50 shots a month. For a mid-market or SMB list where a warm-ish relationship changes the outcome, I open with a personalized connection request and let the post-accept message carry the actual pitch. For anything that needs real volume, and that is most outbound programs I run, cold email carries the load, with LinkedIn layered in as a second channel rather than the primary one. I would not build a 100-meetings-a-month program on InMail or connection requests alone; the caps and the risk profile do not support it. I would also not skip LinkedIn entirely in favor of email-only, since the data above shows the post-connect message reply rate is meaningfully higher than a cold email's average once someone has already said yes to knowing who you are.

FAQ