We call last, on purpose

Revealing a phone number costs money and email does not, so the call waits until day three. Everything that can be checked without the number is checked before the number is bought.

A phone number is something you buy

Revealing a candidate’s phone number costs money. Sending that same candidate an email does not. Those two sentences read like a billing detail, and they are really a design constraint, because between them they decide the order in which a sequence is allowed to do things.

On one candidate the difference is trivial. On a role where a few hundred people are being approached, the order you spend in is most of what you spend. A sequence that reaches for the number first pays for everybody on the list. A sequence that reaches for it last pays only for the people the free channel could not reach.

So the call sits on day three

On EmployLabs the call comes late in the outreach sequence, on day three, once email has had a real chance to work. Not because the call is a weak channel — it is often the one that finally lands — but because it is the paid one, and a paid step placed early buys numbers it turns out not to have needed.

That was not settled by taste. We looked at what happened when a dial went out to somebody who had received exactly one email. Across a small run of dials it reached nobody at all. Every one of those numbers had been paid for first.

A number bought on day one is bought before the free channel has finished its work.

The run was small and we are not going to dress it up as a statistic. It did not need to be large to settle the question, because the question was never whether calling works. It was whether calling works better than an email that has not yet been given time to arrive, be opened, and sit in somebody’s inbox over a lunch break.

Everything knowable for free is checked before you pay

There is a second and quieter way to waste a reveal, and it is worse than mistiming one: buying a number for a call that was never going to be placed.

The candidate may have opted out. The role may have closed. They may already have replied to the email, which makes the call not merely unnecessary but rude. None of those facts needs a phone number to establish, so all of them are checked before the reveal is paid for, never after it.

Written down, the ordering sounds obvious. It is easy to get wrong in practice, because the natural shape of the work is fetch the contact details, then decide what to do with the candidate — and that shape pays for every refusal it makes.

A dial that never happens still records why

If the answer to “why was this person not called?” lives in a log line, nobody can ask it about forty people at once. So every dial that does not happen writes down its reason in a field you can group by, rather than in prose somebody has to read one row at a time.

That turns an unanswerable question into a table. Forty people were not called: how many because they had already replied, how many because they had opted out, how many because the role had closed, how many simply because the sequence had not reached day three yet. Those are four different situations, and only some of them are a problem.

It also means an absence never has to be explained by guesswork. A recruiter who notices that a batch looks quiet can see what the quiet is made of, in one view, without asking anyone to go and look.

The principle is older than the feature

Order your spending so the expensive step runs only once the cheap one is genuinely exhausted. That rule reaches well past phone numbers. Any pipeline mixing free work with paid work has a right order and a wrong one, and the wrong one is usually the order that feels more thorough.

• Cheap first, and for long enough to count. A free channel that is cut off after a day has not been tried; it has been sampled, and the paid step inherits work the free one was about to do.
• Refusals should be free. Every reason to stop that can be established without the purchase is established without the purchase, so deciding not to act costs nothing.
• An absence needs a reason, not a silence. The thing that did not happen is as much a part of the record as the thing that did, and it is only useful if you can count it.

How to use it

1. Let the sequence reach day three before you judge the call rung. A campaign read on day one is a reading of the email, not of the calling.
2. When a batch produces fewer dials than you expected, group the not-called reasons before changing anything. Replies and opt-outs are the sequence working, not failing.
3. If you want the call sooner on a particular role, move it deliberately and watch what it costs. The default is late because late is cheaper, not because early is forbidden.