Almost nobody publishes their own outreach numbers, and the ones who do publish the good campaign. So here is all of it, including the angles that produced a zero, pulled from my own tables on the day of writing.
If you own an established business and you are weighing up whether outbound belongs in the way you attract new work, this is the honest shape of it at small scale.
LinkedIn, 899 messages
Seven campaigns between 5 and 26 August 2026. Every figure below is from a single table that records the send itself, not from a table of logged conversations, which matters for a reason I will come to.
- speed_ask, 666 sent, 30 replies, 4.50%
- rebuild, 119 sent, 4 replies, 3.36%
- ssl_cert, 44 sent, 0 replies
- ssl_down, 44 sent, 0 replies
- speed_ask_report, 22 sent, 0 replies
- speed_ask_report_v2, 3 sent, 0 replies
- mini_story, 1 sent, 0 replies
Total: 899 sent, 34 replies, 3.78%.
Read the bottom four lines before the top two. Three of those campaigns have send counts of 22, 3 and 1. Those are not results. They are attempts that were never given enough volume to mean anything, and if I reported them as failed angles I would be inventing a conclusion out of noise. They are on the list because leaving them off would make the account look tidier than it was.
The two ssl campaigns are different and they are the real finding. 88 messages across two variations of the same idea, and not one reply. That angle told a business its security certificate was broken, which was true, checkable and genuinely their problem. It produced nothing, twice.
Meta ads, $372.12
Seven campaigns, 29,148 impressions, three weeks. Meta reported two leads.
Both were my own test submissions. My event log holds 14 Lead events across 6 sessions, and not one of them carries the click identifier that Meta attaches when somebody arrives from an ad. Zero out of 14. Meanwhile 639 of 1,184 tracked page views did carry one, so the ads delivered clicks perfectly well. They just never delivered a conversion that could be traced to a click. The full breakdown is in what $372 of Meta ads actually bought.
At that budget this is close to structurally guaranteed. Conversion optimization needs a meaningful number of conversion events per ad set per week before the platform stops guessing, and $372 split across seven campaigns never gets there. The largest campaign took 15,078 impressions and the smallest took zero. That is seven separate nothings running at once.
258 outbound emails and 170 inbound logged in my CRM since 4 August 2026. I am deliberately not converting that into a reply rate, because the inbound figure includes ongoing conversations with people I already know, so dividing one by the other would produce a number that describes my address book rather than a cold channel. Email at FGN has not yet been run as a clean, measured cold campaign, and saying so is more useful than a fabricated percentage.
The number I got wrong first
The reason I insist on the send table is that I once did not use it.
I wanted a reply rate, so I queried the table of logged conversations and divided replies by contacts. It returned 17%, which is an excellent number, and I believed it for a while. It was meaningless. That table only contains people because they replied, so I had computed a reply rate from a dataset produced by replies.
The true figure at the time was 2.1%. Not 17%. The arithmetic was never wrong; the source was. Eight times off, and it did not present itself as uncertainty. It presented itself as knowledge, which is what makes this failure mode expensive.
What I would tell an owner
Three things, and none of them are encouraging in the way marketing content usually is.
Small paid budgets do not buy small results, they buy noise. You will get a dashboard that reports leads with confidence. Instrument your own site and define your own conversion, or the number you are reading is the platform grading its own work.
A real, verifiable problem is not a reason to buy. The ssl campaigns are the cleanest evidence I have of this. The problem was genuine and provable and nobody cared, because a problem you can measure from outside is a property of a business, and the willingness to fix it is a property of a person. Those two things are close to independent, which is most of what I got wrong about qualifying buyers.
Volume decides whether you learn anything. Four of my seven campaigns produced no usable information at any confidence, and three of them produced none because I never sent enough. If you are going to test an angle, send enough of it to be able to tell the difference between a bad idea and a quiet week.
Questions owners ask
Is a 3.78% reply rate good?
It is roughly normal for cold outbound at small scale and it is a bad way to think about the question. Thirty-four replies is the number that matters, because a service business does not need a percentage, it needs a handful of the right conversations. A 1% reply rate on a well-chosen list of two hundred people beats 5% on the wrong two thousand.
Should an established business do cold outreach at all?
It is the slowest and least pleasant way to fill a pipeline, and for most owners with a few years behind them, the people who already know you are a far better list to work first. That goes double for a coach or consultant whose work runs on trust. Cold outreach is worth it when you are deliberately entering a market where you have no history, and it should be judged over months rather than weeks.
Why publish the campaigns that failed?
Because a report containing only the successful campaign teaches nobody anything, including me. The zero on 88 ssl messages changed how I qualify prospects more than any of the 34 replies did.
Where did these numbers come from?
The LinkedIn figures are a single query against my own send log covering 5 to 26 August 2026. The ad figures come from the Meta API snapshot and my own site event log for 23 June to 27 August 2026. The email counts come from my CRM's email activity table. Every one of them is a table I can name, on a date I can state.