The List Is the Campaign

A campaign goes out to four thousand contacts. The reply rate comes back at 0.4%. The meeting that follows is about subject lines, and the next version has a different opener, a shorter second paragraph and a new call to action. It goes out to the same four thousand contacts.

The copy was probably not the problem.

What a bad list actually costs you

The obvious cost is the one everybody counts. Messages to people who cannot buy produce no replies, and that is a wasted week.

The cost that matters more is slower and it compounds. Email sent to people with no reason to hear from you generates deletions without opens, spam complaints and bounces, and every mailbox provider reads those signals as evidence about you rather than about the list.

Google publishes the thresholds in its sender guidelines. Bulk senders must authenticate with SPF and DKIM, publish a DMARC record, offer one click unsubscribe, and keep spam complaints below 0.1%, never exceeding 0.3%. Those numbers are small. On four thousand sends, four complaints is the comfortable ceiling. A list assembled by filtering a database on company size and country can clear that in a morning.

Once the domain is marked, your good email stops arriving too. The genuinely interested prospect, the proposal, the invoice. A bad list does not simply fail. It takes the channel down with it.

The part nobody checks until it is too late

There is a quieter version of this that surfaces months later. The domain you prospect from is usually the domain you invoice from, and sending reputation attaches to the domain rather than to the campaign. Teams discover this when a client says they never received the proposal, and the proposal is sitting in a spam folder because of a campaign that ran in March.

The usual response is to buy more domains and spread the sending across them. That buys time and changes nothing, because the same list produces the same complaints on the new domains, only more slowly. You end up maintaining nine domains in order to keep running one bad list.

The cheapest check available is to look at what share of your sends went to addresses that have never opened anything, ever, across the last three campaigns. Then compare that share with your reply rate. If the first number dwarfs the second, the list is the campaign, and no subject line is going to rescue it.

Why everybody rewrites the copy instead

Because the copy is visible and the list is not. Anybody can read an email and have an opinion about it. Nobody can read four thousand rows and tell you they are wrong.

There is a reporting reason too. Reply rate is the number on the dashboard, and reply rate responds to copy in a visible, satisfying way inside a week. Deliverability damage shows up over months, in a metric nobody is watching, attributed to nothing in particular.

And building a list properly used to be expensive in a way that writing was not. Deciding whether one specific company is worth contacting means looking at what they do, how they are organised and whether the thing you fix is a thing they have. At thirty seconds a company, four thousand companies is thirty three hours. So the list gets built the only way that scales without people: pick an industry code, a headcount band and a country, then export.

That filter produces companies matching a description of your market. It does not produce companies with a reason to answer this month, and those are different sets.

What AI changed about this

The qualifying step is now affordable. A system can visit each company's own site, read how they describe themselves, look at what they are hiring for, and judge all of it against a written description of what makes a good fit. Four thousand companies become six hundred with a stated reason each, and the three thousand four hundred you dropped are the ones that were going to generate the complaints.

It also changes what can go in the message. When the system has already read the company, the first line can refer to something specific and true. That is not a personalisation trick. It is the reason the message stops being a complaint waiting to happen, because a message that is obviously about the reader does not get marked as spam.

And the list can be maintained rather than built. People change jobs, companies change shape, and the signal that made somebody a good fit in March is gone by September. Re-reading the list on a schedule is now a background job instead of a project.

What AI will not do for you

It will not fix your authentication. SPF, DKIM, DMARC, unsubscribe handling and sending volume are infrastructure, and no amount of qualification compensates for a domain that was never set up properly.

It will not make a weak offer work. A precisely targeted message about something nobody wants earns a precisely targeted silence.

And it will not tell you what a good fit is. It applies your definition at scale. If your definition is companies between twenty and two hundred staff in Quebec, you will get a faster version of the list you already have.

What to build instead

Write the fit definition in sentences rather than filters. "Firms running field crews who still collect job information on paper" is something a system can check against a website. An industry code and a headcount band is not.

Put a qualification step between the source and the send, and make it produce a reason per company. Any company it cannot give a reason for does not get contacted. That one rule does more for your deliverability than any copy change.

Keep the prospecting domain separate from the one you invoice and contract from, and treat that separation as permanent rather than as a trick for buying headroom.

Measure the campaign on qualified replies and complaint rate together. A campaign that lifts replies while pushing complaints toward 0.3% is borrowing from next quarter.

And rebuild the list on a schedule. A list is a perishable asset, not an item of inventory.

We build the qualification step rather than just the sending, so the system reads each company, decides whether there is a reason to write and tells you which ones to drop. See Cold Email Outreach.

Sources

Google, Email sender guidelines, requirements in force from February 2024. support.google.com