Guide

Six Things I Learned Watching Thousands of LinkedIn Campaigns Run (and Got Wrong First)

Six counterintuitive lessons from watching thousands of real LinkedIn outbound campaigns: blank invites, fast replies, bad lists, and where AI actually earns its keep.

Tomer Tarsky By Tomer Tarsky ยท ยท 4 min read
Share

I build Linkdify, an AI tool that runs LinkedIn outbound for founders and small B2B teams. That means I spend a strange amount of my week watching other people's campaigns run. Not the pitch decks. The actual sends, the accepts, the silence, the one reply that turns into a call. After thousands of them, a few things I believed about outbound turned out to be wrong, and a few things I had dismissed turned out to be most of the game. Here are six.

1. The list beats the copy. Every time.

When a campaign stalls, the founder rewrites the message. Almost nobody rewrites the list. From what I see, that is backwards. The campaigns that get replies are rarely the ones with the cleverest opener. They are the ones where the 200 people on the list all have the problem right now and are senior enough to do something about it. A dull message to the right 200 people beats a sharp message to a mixed bag of 2,000.

I have watched the same template run on two different lists in the same month, get completely different results, and the founder blame the template both times. If you are tempted to A/B test copy, test the list first. It is the bigger variable and the cheaper one to change.

2. The blank invite wins, and it isn't close.

This is the one I argued with. Every sales course says to personalize the connection note. Across 8,400 real connection requests sent through Linkdify in August and September, requests with a note were accepted 8.6% of the time. Requests with no note were accepted 48.6% of the time.

My read: a note turns a two-second decision into a judgment call. With no note, the person glances at your headline, decides you seem relevant, and clicks. With a note, they are now evaluating a pitch from a stranger, and most pitches lose. Save the words for after they accept. That is where the conversation actually begins.

Want LinkedIn outreach that runs itself?

Linkdify finds your leads, writes personalized messages, and sends them at safe, human-paced limits.

Try Linkdify free โ†’

3. Replies come fast or they don't come.

Across 2,511 messages, 14.6% got a reply, and the median reply arrived about 14 hours after the send. Most landed inside the first day. So the three-week, seven-touch cadence most tools ship as a default is mostly talking to people who already decided not to answer.

A second message two or three days later is reasonable. Message five on day 18 is noise, and it is the kind of noise that gets you reported. Run shorter sequences. Then spend the time you saved on the people who did answer.

4. "More volume" is the most requested feature and the most damaging one.

Almost every founder who signs up asks the same thing in week one: can I send more? I understand it. Twenty invites a day feels slow when you have a quarter to hit. But the accounts that push volume are the accounts I end up writing restriction post-mortems for. LinkedIn does not care how fast your tool is. It cares how many strangers ignore you, and volume grows the ignored pile faster than it grows the accepts.

So Linkdify keeps the cap and paces sends the way a person would, and I have said no to the loudest request in my inbox more times than I can count. It is the right no. A restricted account sends zero.

5. Half of "the tool is broken" is a spreadsheet problem.

This one humbled me. When someone writes in saying a campaign reached 140 of 200 leads and stopped, I used to go hunting for a bug. Now I look at the other 60 first. A blank LinkedIn URL. A profile that was deleted last year. A CSV exported from Excel that quietly dropped the first column, so every lead is missing a name and the opener reads Hi ,. Garbage in, silence out. The software did exactly what it was told. It was told nothing useful.

If you run your own outbound, open the leads that did not go out before you blame the sequence. The answer is usually sitting right there in the row.

6. AI is a better reader than writer.

I built an AI writer into the product and people use it constantly. But the place the model earns its keep is not the opener. It is the unglamorous work on either side of the send: deciding which 20 of your 400 leads should go out today because they just changed jobs or posted about the exact problem you solve, then reading the replies that come back and telling you which four deserve your morning. Humans are slow at that and bored by it. The model is neither.

A generated message, on the other hand, is only as good as the signal you feed it. Give it nothing specific and it writes something generic, and a generic message reads generic no matter which model wrote it. Use the AI to pick and to read. Write the one sentence that matters yourself if you have to.

Want LinkedIn outreach that runs itself?

Linkdify finds your leads, writes personalized messages, and sends them at safe, human-paced limits.

Try Linkdify free โ†’

What I'd tell a founder starting tomorrow

The campaign is the small part. Who is on the list, what you leave out of the invite, how quickly you show up when someone answers, and whether you keep your account alive long enough to learn anything at all. That is the job. The sending is the piece a machine can do, and it should, so you can spend your hours on the four people who wrote back.

Tomer Tarsky
Tomer Tarsky

Founder of Linkdify. Writing about LinkedIn outreach, GTM, and building pipeline without the ban risk.

Your LinkedIn GTM team, without the ban risk

Linkdify runs safe, human-paced LinkedIn outreach that books conversations for you.

Try Linkdify free