Score the Signal and Gate the Spend

Outcome: a hiring score built from free columns, a threshold set from capacity, and a tested gate above the person-level work.

Surface: App
Level: Intermediate
Uses: Formula columns · conditional runs
Credits: 0 to build; saves most of the weekly run
Prerequisite: Lessons 02–05

Everything the score reads is already free

By this point you have role match, opening counts, growth rates, department headcount, hiring intensity and the extracted post details. All of it is either already paid for at the cheap tier or computed as a formula.

The score costs nothing and decides where the expensive credits go.

The model

Hiring score — max 15

Criterion	Points
3+ open roles in your function, last 30 days	5
2 open roles	3
1 open role	1
”First X hire” language in the post	4
Leadership role in your function	3
Named competitor or complement tool in the post	3
Department size just above your minimum viable threshold	3
Openings growth above 25% quarter on quarter	2
Department below minimum viable size	−4
Post older than 30 days	−3
Company is a recruiting or staffing firm	−15

Whole numbers, and every input is a column you already have.

The “first X hire” criterion is worth its weight. A newly-created role means no incumbent tool, no previous decision to defend, and a person arriving with a mandate to choose. It is the single most predictive detail in a job post for most B2B products.

Decay

A hiring signal is worth less as the post ages, because the role gets filled and the moment passes.


age <= 7 days     ×1.0
age <= 14 days    ×1.0
age <= 30 days    ×0.7
age > 30 days     ×0   — drop

Note the flat top: unlike funding, a job post is not worth less in week one. If anything the ideal moment is week two to three, once the pressure the post implies has started to be felt.

Set the threshold from capacity

Compute weekly capacity

How many new accounts can your team actually work in a week? That is the number.

Score the whole surviving set

Free.

Sort and read off the threshold

At your capacity row.

Read ten above and ten below

If they are indistinguishable, add a criterion — usually the post-detail ones, since they carry the most information.

Place and test the gate

The gate sits directly above find_people_within_company — the first expensive step.

Testing it is not optional:

Run on 20 rows and count survivors by hand.
Confirm both branches fire — some pass, some fail.
Test the null case deliberately: what does the gate do when department_headcount is empty?

A gate reading a column that has not populated will pass everything or fail everything depending on how the comparison was written, and both look like a working pipeline. If your gate has never excluded a row, it has not been tested.

The cost difference

500 companies matched by lesson 02, gated to 120.

	Ungated	Gated
Growth + department columns	1,500	1,500
Find people (2 each)	300	72
Email + verify	1,100	264
AI copy	1,000	240
Total	3,900	2,076

Nearly half, and the 120 accounts contacted are the ones with the strongest signal. At weekly cadence that difference compounds across the year.

Do this now

Build the score as a formula column

Every input already exists.

Add decay by post age

Compute weekly capacity

Score, sort, read off the threshold

Read twenty accounts around the line

Place the gate above the people step

Test it on 20 rows, counting by hand

Including the null case.

Recompute the weekly cost

Gated and ungated. Write both down.

Check your work

Every scoring input is a free column
Decay drops posts over 30 days
The threshold came from a capacity number
The gate was tested with survivors counted by hand
You have both weekly cost figures

Where this breaks

Scoring on posting volume alone puts large companies at the top of every run, because they always have more open roles. A 5,000-person company with eight relevant postings is not a stronger signal than a 60-person company with three — it is a bigger company. Normalize by department size using the hiring-intensity ratio, or your weekly queue will be the same set of enterprises every week.

Further automation

Track reply rate by score band. After a quarter you will know whether the top band genuinely outperforms the middle, and can move the threshold on evidence instead of capacity guesswork. If they perform identically, the score has no resolution and needs a sharper criterion.

Next lesson

08 — Schedule the weekly run, turning this into something that happens without you.

Reference for this lesson: Actions, Tables, Credits, GTM Engineering — scoring and gating.