A year of outbound read through the mailboxes: 117 campaign folders, 3,895 prospects, 2,097 emails, 29 reply digests whose 37 flagged messages mostly repeat on follow-ups, 53 bounces and 42 auto-replies. Which numbers we keep, which ones lie, and the four ratios we act on.
Every outbound campaign produces two kinds of number: the ones that feel like progress and the ones that pay for the renewal bill. After a year of sending domain offers, the second list is shorter than the first and much less flattering, and the work of the last twelve months was mostly learning to tell them apart.
The archive behind this article is not a dashboard. It is 117 campaign folders, a consolidated prospect file of 3,895 addresses, a sending log, and 29 daily reply digests that a script has been writing since the first days of October. Reading it back is the only way to answer the question every domain seller asks after a quiet month: is the campaign dead, or was it never running?
The honest answer in our case is that it was running, that it moved real mail, and that the constraints were never the ones the numbers on top of the sheet suggested.
The five counters worth keeping
A campaign can be measured a hundred ways. These five are the ones that survived a year of use, because each one changes a decision on its own.
- Sends — the count of messages that left the mailbox. It is the denominator for everything else, and the only counter that stops being ambiguous the moment you filter out bounces.
- Deliveries — sends minus the failures that came back. This is the real reach of a campaign, and it is always lower than the sending log pretends.
- Human replies — messages from a person who read the offer and typed back. Not an out-of-office, not a delivery report, not a marketing email that happened to mention the domain.
- Negotiations — a human reply that contains a number, a question about terms, or a request for time. This is the only counter that predicts revenue.
- Closes — money received. Rare enough that a single one distorts the whole month's ratio, which is exactly why it should never be the metric you optimise daily.
Everything else — opens, clicks, positive sentiment, "engagement" — is either unavailable in plain SMTP mail or too easily inflated to survive a second look. Opens require a tracking pixel, and a tracking pixel in a cold domain offer is the fastest way to land in a promotions tab or a spam complaint.
What the year actually sent
The consolidated sending log holds 2,097 messages. Against 112 campaigns that is roughly nineteen messages per campaign, which tells you most campaigns were small, targeted and short rather than long blasts across a bought list.
That ratio matters more than the total. A campaign of nineteen messages aimed at nineteen businesses that already use a variation of the name will outperform a campaign of two thousand messages aimed at anyone who owns a competing domain in the same industry. Volume is the least interesting variable in this business, because the buyer for a specific name is a specific operator, not a market segment.
The prospect file shows the same discipline from the other side: 3,895 addresses collected and screened, of which 2,097 were actually written to. The gap is deliberate. Names that came back as undeliverable, domains parked behind a catch-all that swallows everything, and contacts whose role was clearly not the decision maker were dropped before sending rather than after. A list that is written to before it is cleaned shows up later as a bounce problem in the middle of a campaign, which is the worst possible moment to discover it.
The reply digests and the 37 that needed attention
Since the first week of October a job has been writing a digest of replies across all campaigns every afternoon. Twenty-nine digests now cover 4 October to 7 November. They flagged 37 messages that needed attention, 36 other messages that mentioned a campaign, 53 bounces and 42 auto-replies.
Those four numbers get misread in four different ways, and the misreadings are the useful part.
The 37 is not 37 buyers. Several of those threads are the same person replying twice — a "not interested" on the first touch, followed by a sharper "not interested again" after a follow-up. A follow-up that produces a second negative reply looks like engagement in a daily count and is actually a stop signal. The digest flags it because a human should see it; the campaign metrics should never count it twice.
The 36 "other" messages are mostly not replies. A confirmation email from a contact form, a marketing blast that happened to include our domain name, a LinkedIn notification — all of them contain the campaign name because the campaign name is the domain name. They are noise that a keyword filter picks up and a human eye discards in under a second.
The 53 bounces are the cheapest lesson in the file. Every one of them is an address that should have been screened. At the volumes we run they cost almost nothing; at ten times the volume they would decide the reputation of the sending domain. The correct reading of 53 bounces across two months is not "deliverability problem" but "screening gap", and the fix is upstream of the send.
The 42 auto-replies are the counter you must never add to the reply rate. Out-of-office messages look identical to a reply in a mailbox view, arrive within seconds, and can double a believable reply rate into an imaginary one. We keep them in their own column, and the column is only ever read at the end of a month to check that a campaign is reaching live mailboxes at all.
The denominator problem
Here is the arithmetic that makes most domain outreach reports worthless. Suppose a campaign sends 200 messages, and every single one of them is a bounce. A naive count says 200 sent, five replies, a 2.5% reply rate. A correct count says zero delivered, and a reply rate that cannot be computed, because there was nothing to reply to.
The sending log always reports more than the campaign delivered. It records attempts. On our own numbers the difference between "written to" and "written to an inbox that exists" was never zero, and on the badly screened campaigns it was large. Every ratio in a campaign report should be divided by deliveries, not by sends, or the report will reward a list that was never cleaned.
The same discipline applies to the reply rate. "Forty-two auto-replies and thirty-seven flagged messages" reads like a healthy campaign with 1.9% engagement across 4,000 messages. Reduce it to humans who are not on holiday, remove the duplicate negatives, and the number that remains is small — small enough that a single reply changes it by a visible amount. That is the true shape of this business at our size, and it is better to know it than to be flattered by it.
Four ratios we act on
Numbers that only get reported do nothing. These four are the ones that produce a change in the next campaign.
Delivered per written address. If a screened list loses more than a small fraction of its addresses to bounces, the screening pipeline is the problem, not the copy. The fix is to refuse the address before the first touch.
Human replies per hundred deliveries. This is the health check on targeting. A campaign with a low rate on a clean list is aimed at the wrong operators, and no rewrite of the subject line fixes that.
Negotiations per human reply. This separates a reply from a conversation. Someone who writes "not interested" engaged with the mail and rejected the offer; someone who writes "what would you take" has moved the sale forward. If this ratio is zero across many replies, the offer itself is wrong — the price band is not where the buyer's budget lives.
Net per close. A close is not revenue until the commission, the escrow fee and the renewal already spent on the name are subtracted. One sale at a headline price with a 30% commission ladder can net less than a smaller sale on a standard listing. The gross number is the one that gets shared; the net number is the one that decides whether the campaign is worth running again in January.
What the numbers refused to tell us
Three things a year of campaigns cannot answer, and it is worth writing them down so future reports stop pretending otherwise.
Volume does not predict replies. The largest campaigns were not the most productive, and the small nineteen-message campaigns that targeted businesses already operating near the name produced the conversations. If a report ranks campaigns by size, it ranks them backwards.
A reply is not a buying signal. The 6 November article in this log covers this in detail: out of every reply our campaigns received in 2026, exactly two came from a human who was actually a candidate buyer of the name. The rest were either declines, service providers selling something to us, or automated messages wearing a human name.
Time-to-reply is not a quality signal. Our fastest replies were the declines. The slowest were the ones that turned into a number. A campaign winner is not the thread that answered in fifteen minutes; it is the one that answered in eight days with a question.
The sheet we keep now
The final artefact of the year is a small sheet with one row per campaign and eight columns: domain, date launched, addresses written to, deliveries, bounces, human replies, negotiations, net received. Nothing else. No opens, no clicks, no sentiment score, no "leads generated" field.
Eight columns is small enough that the numbers can be trusted, which is the only property that matters in a log kept by a person running campaigns in the margins of a working week. The 117 folder archive is the raw material; the sheet is the reading.
A year of sending also produced an unglamorous conclusion that no report column can hold: the campaigns that worked were the ones aimed at operators who already had a name-shaped problem, and the sending mechanics — the volume ramp, the follow-up sequence, the bounce discipline — only determined whether the message arrived to be rejected or ignored. The list was the campaign. The mail was just the delivery method.
If you run this yourself, start the log on the first day, not the hundredth. The numbers you cannot reconstruct are the ones that would have told you why.
Buying a name this week?
193 exact-match, brandable and geo domains — buy-now price, escrow-protected, same-day transfer. See the portfolio · the contact page
Top comments (0)