Lesson 2 of 5

Lesson 2.2.2

Pick the signals worth scoring on

6 min readIntermediate

What this costs

The signals cost what their columns cost. The scoring on top is free.

Before you start

Lesson 2.2.1.

The scoring agent is not a research tool. It reasons over the data already in the row, which means a row with three thin columns scores near zero no matter how good your prompt is. Choosing which columns to fill before you score is most of the work.

Score on what has worked for you

There is no standard set of signals. The right ones are the things you already know separate your buyers from everyone else: their size, the markets they sell into, a certification, the kind of distributor that has bought from you before. Start from the customers you have won and write down what they had in common that you could see from the outside. Those are your signals, and each one becomes a column the Scoring Agent can read.

15 second explainer. Names and figures on screen are illustrative.

Some teams also score on something that changed recently, such as a new hire, a new location or a funding round. Add one if it has mattered in your own deals. It is an option, not a rule, and a profile built only on who the company is can be exactly right for your market.

Building a people list from an intent signal in Kuration
Sourcing on a signal rather than on a firmographic filter. When the signal is what put the row in the list, the evidence for that criterion is already in the row.

How many columns before you score

  • Four to six filled columns is the working minimum for a score you can trust.
  • At least two of them should be things the source did not give you for free, otherwise you are scoring the source rather than the company.
  • Every criterion in your prompt must map to a column that exists. If it does not, either add the column or remove the criterion.

Scored a freshly sourced table

What you see
Almost every row scores 2 or 3
The fix
There is nothing to reason over yet. Fill four to six columns first.

A criterion with no matching column

What you see
The rationale mentions the criterion but never has evidence for it
The fix
Add the column or drop the criterion. A criterion the row cannot answer is noise in every score.

Check yourself

0 of 2 answered

1.You score a freshly sourced table and almost every row lands at 2 or 3. Why?

2.Where should your scoring signals come from?

Answer all 2, then open Weights, deal breakers, and the scoring formula to tick this lesson off.

Try it yourself

List every criterion in your ICP and write the column name that answers it next to each one. Circle the ones with no column.

You have done it when
Every criterion has a column, or has been deliberately dropped.
Credit budget
Free.