For most of this year, this site said in several places that email sending was not part of the product, that it was scheduled for the fourth quarter, and that it would stay marked as unbuilt everywhere it came up. It shipped in August. This is the post that says so, with the numbers, because the alternative was letting those paragraphs quietly go stale and hoping nobody compared them with the changelog.
01What we promised, and what changed
The promise was narrow on purpose. Every message would go through a contact form the company had published itself, because a form is a door the company opened for exactly this, and an inbox is not. We still believe the second half of that sentence. What we got wrong was how often the first half is available.
By the middle of August the pipeline had qualified more than two hundred companies and had a growing pile of messages a person had read and approved that could not be delivered. Not because the company was wrong or the message was wrong, but because the door was dead. The honest options were to leave those on the floor or to build the second door earlier than planned. We built it, and we updated the paragraphs that said we had not.
02Why the form is still tried first
A published contact form is a statement from the company: this is where we want to hear from people we do not know. Somebody is paid to read what comes through it, it is monitored during business hours, and nothing about using it requires knowing anything about any individual. That is why it is the door we try first, every time, and why the email path only exists for the case where there is no such door.
The earlier piece on why forms get read when email does not still stands, and it now carries a short update saying so. The argument was about attention, and attention has not changed. What changed is our count of how many companies actually publish a form that can carry a message.
03What a dead door looks like
We classified every parked message before building anything. The largest group, 38 of them, were parked for reasons that were facts about the door rather than the company. They fall into three shapes:
- No contact form at all. The site has a sales page, a booking widget, a chat bubble, or nothing. A booking widget is not a contact form: it asks a stranger to take a meeting before saying a word.
- A form behind a CAPTCHA or a bot challenge. We do not solve those, on principle and in practice, so the form exists and is unusable to us. The company has stated a preference and we treat it as one.
- A form with no message field. Name, email, company, and a submit button. It exists to start a sales sequence in the other direction, and there is nowhere in it to say anything.
Those 38 are the messages that had already been written for a form before the form turned out to be dead. They are not the whole count of companies without a usable form: when the pipeline finds no form at the moment it reads a company, it goes looking for a person straight away, and those messages are written for an inbox from the start and never sit in the parked pile. Of the 38, we found a named person at 23 companies. For the other 15 we found nobody, and those are now marked unreachable in the ledger rather than being retried every fifteen minutes forever. Nobody was written to at a guessed address, and nobody was written to twice.
04What the email path does
When a company has published no working form, we look up one person at that company who would own the decision, through a contact data provider, and only for that one company. We check that the address accepts mail before anything is sent. The message that was written for the form is given the greeting and sign-off a letter needs and goes back into the review queue as a draft, because the person who approved it approved it for a form, and a named inbox is a different act with a recipient nobody has looked at yet.
One more click, then it goes. It goes once. Every email carries an unsubscribe header that any mail client can act on, and a reply goes to a real person at the company doing the writing. There is no second message if the first one is not answered.
05What it refuses to do
- No guessed addresses. No first.last permutation, no pattern inferred from a domain. If the only route we could construct is a guess, the company is turned down for want of a door and stays turned down.
- No sequences. One message per company per period. You can raise the ceiling on how often; you cannot remove it.
- No bulk path. There is no list upload and no send-to-all, and there will not be.
- Nothing sends itself. Every rescued message went back to a person before it went to a person.
- No profiling. The record we hold about a contact is name, title, work address and a public profile link, kept to deliver one message and honour a request to stop, and deleted on request. The privacy policy now says exactly that.
- No country gate. This site used to say some countries were excluded by default. Nothing in the pipeline enforced it, so the promise was withdrawn rather than the code written. Where a recipient sits is recorded on every company and shown at review, and the decision is made by a person, one message at a time.
06The numbers so far
Since the first form knock on 30 July and the first email on 3 August, against our own business:
- 994 companies read and written down in the ledger, with a reason attached to each. 204 were judged worth writing to.
- 63 messages delivered. 12 went through a published contact form. 51 went by email.
- Of the 51 emails, 23 were the rescued messages from the parked 38, written for a form that could not carry them. The other 28 went to companies where no form was found at the outset, so the message was written for a person from the start and never parked.
- 15 companies had no form and no findable person, and were marked unreachable rather than approximated.
- The first reply we have had arrived through a form, the same morning the message went in, from a company whose form had rejected our first attempt and only accepted the second, made through a real browser.
Read that middle line again, because it is the one that surprised us. Four in five delivered messages went through the second door. The published contact form, the thing this whole product was designed around, turned out to be the exception rather than the rule among the companies we most wanted to reach. We do not think that makes the form the wrong door. It makes it a rarer one than we assumed, and it means a product that could only knock on forms was quietly declining to contact most of its own best prospects.
So the short version is this. The form is still first, for the same reason it was always first. When the form is a dead door, one person is found, one message is approved again, and it goes out once with a way to say no attached. And the paragraphs on this site that described the roadmap in the present tense have been corrected in place, dated, and left where they were.