The fourteen days of an outbound pilot, written down before you pay

Most pilots are sold as a wager: give us money, give us a month, and we will see what happens. The wager is not the
problem. The problem is that nothing in the arrangement lets you check the work while it is being done, and nothing in
the read-out lets you tell a bad segment from bad copy. So here is the pilot I run, day by day, before anyone pays for
it — and the numbers each day actually produces.

Days 1–3: where the domain comes from

A segment is only as honest as the field the domain was read out of. That is the rule I hold to, and it is stricter
than it sounds: the domain has to sit in a field its owner filled in, not be extracted from prose.

The best version of that rule is a package registry. A library's projectUrl on NuGet or its homepage_uri on
RubyGems is written by the vendor about itself, and it is machine-readable, which means it can be audited rather than
trusted. The last pass over two registries looked at 483 candidate domains and found 138 live sites that passed a
product test (pricing, a signup path, a demo request, integrations) — a 29% survival rate.

Compare that with a job feed, where the company domain has to be dug out of the body of a vacancy. On hiring threads
the same pipeline yields about 11%. The rows are not worse because job posts are dishonest; they are worse because the
domain is inferred rather than declared. Inference is where invented addresses come from.

The working file now holds 7,641 company domains from 48 public sources, and 1,355 of them publish an address
(18%). That last number is the one I would want if I were buying: the majority of companies with a live product site do
not publish an email address on it, so a list sold as "100% verified" is either smaller than claimed or has been
completed by guessing.

Days 3–5: verification, and what is handed back

Verification is not a single check. It is a sequence, and each step drops rows:

1. Domains folded to their root, so a blog or an app subdomain does not become a second company.
2. Duplicates collapsed on the root domain, not on the string.
3. Rows without a domain dropped outright — no domain, no row.
4. Each host fetched live. A 404 is a dead company, not an unverified one.
5. Each address matched to the domain it claims to sit on. hello@takt.com on the page of taktstudios.ca is not an address for the company on the page.
6. An SMTP check before anything is sent.

What does not survive is not deleted quietly. It is handed over as a separate file with the reason attached: which step
dropped the row, and what the row looked like when it went in. An agency reselling this needs that file more than it
needs the good rows, because it is the only way to explain to a client why the count changed.

Days 5–7: the domain that is not the main domain

Sending happens from a separate domain, warmed in the open, with SPF, DKIM and DMARC in place. This is the part of the
pilot that is deliberately not clever: if deliverability is going to be damaged, it gets damaged in a domain built for
outbound, never in the one the company answers email on.

The infrastructure is the deliverable here, not a service you rent forever. Records, warm-up history and the sending
domain stay with the client at the end of the fourteen days.

Days 7–14: one batch and a read-out that does not flatter

One segment, one batch, replies worked as they arrive. The read-out counts replies and conversations. It does not count
opens, and it does not report an "engagement rate" assembled from pixels.

It also does not promise meetings. A pilot is a measured test: if the reply rate is zero, the read-out says zero and
then says which family the segment came from, how many addresses survived the SMTP check, what was delivered and what
bounced. A zero with a reason beats a promised pipeline with none, and it costs less to find out.

What the pilot is not

• Not a guarantee of meetings. Reply rate depends on the segment and the offer, and it is measured rather than assumed.
• Not sending from your main domain.
• Not a scraped list with invented addresses.
• Not a retainer in disguise: fourteen days, one batch, one read-out, then a separate decision.
• Not a channel that reads your inbox — replies arrive at an address you can see at any time.

What you hold at the end is a verified segment with its source column intact, a warmed domain in your name, the copy
that went out, and the method. That is the deliverable. Meetings are the outcome, and they are earned rather than
promised.

The pilot is published at $900, written down day by day, here:
the $900 outbound pilot.
The three scopes and prices are on pricing,
and the sourcing rule behind the numbers is described in B2B lead list building.
The Telegram link on any of those pages is prefilled if you want the first segment priced.