Qualifying a prospect
Qualification turns what was read about an organisation into a judgement about whether it belongs on your list. It weighs signals drawn from sourced claims, produces a score and a written reason together, and never decides to contact anybody — that decision belongs to readiness, the compliance check and the channel's autonomy rule. A person can disagree, and the disagreement sticks.
What the score is made of#
Only sourced claims feed it. A signal with nothing behind it cannot contribute, because it never reached the record in the first place — that is the point of Evidence on a prospect and it is why the score cannot quietly encode a guess.
- Match to the trade and place you described — the filter part of your criteria, now checked against what research actually found rather than against a directory label.
- Match to what you said good looks like — the direction part, which is why a vague sentence there produces vague scores everywhere.
- Disqualifiers — properties you named as ruling a business out, applied once evidence supports them.
- Recency signals — a dated announcement is worth more than an undated page, and it decays.
- Contact reachability — whether a sourced address exists, kept separate from fit so the two never blur.
That last separation is the one people find surprising. A perfectly qualified business with no published address scores well on fit and cannot be written to, and Connect keeps those as two different facts rather than averaging them into one misleading number.
What the score may and may not do#
| Question | Decided by | Not decided by |
|---|---|---|
| Is this business worth pursuing? | Qualification | — |
| Which prospects get deeper research? | The cost router, informed by the score | The score alone |
| May a message be written? | Sender readiness | The score |
| May a message be sent? | Autonomy mode plus the compliance check | The score |
| Who is the right person to write to? | Decision-maker inference | The score |
So a high score is not permission. In ask_before_send a perfect prospect still waits for a human yes, and in any mode a do-not-contact entry ends it regardless of fit. Keeping scoring away from sending is what stops a confident model from turning a research opinion into an action.
Disagreeing with it#
Open the fit reason before the score. The reason names the claims it leaned on.
Result You are arguing with a sentence and its sources rather than with a number, which is the only kind of argument that can be settled.
If a claim is wrong, correct the claim.
Result The correction is recorded against the record with you as its origin, and it outranks research from a public page.
If the claim is right but your criteria mis-weighted it, edit the criteria rather than the prospect.
Result The next run judges the whole population your way instead of you re-judging one row at a time.
If you simply disagree, set the prospect's own state — qualified, or out.
Result A person's decision is recorded as a person's decision and the score does not overwrite it on the next pass.
The order matters. Correcting one row is the fastest fix and the least durable; changing the criteria is slower and fixes the next two hundred. Doing several rows by hand before changing the criteria is a reliable sign that a sentence in the criteria needs rewriting.
When the scores are all wrong together#
- Everything scores low
- The population does not match what you said good looks like. Either the trade wording is off or you are describing a customer you do not actually have.
- Everything scores high
- The criteria have no disqualifiers, so nothing can be ruled out. Naming what rules a business out is usually the single most useful edit.
- Scores look right, list is unusable
- Fit is fine and reachability is not. Check the contactable count rather than the qualified one — see Contactability.
- Scores changed without you
- Re-research updated an underlying claim. The record shows the newer claim and its source; a judgement built on a page that has changed should change with it.
Questions#
Does a low score delete a prospect?
No. It stays, ranked below the others, with everything researched about it intact. Deleting would mean paying again to learn the same thing the next time the same search runs across the same business.
Can qualification decide to email somebody?
It cannot. Qualification produces a judgement; sending requires sender readiness, a passed compliance check and the channel's autonomy rule. Those are separate on purpose, so a research opinion never becomes an action by itself.
Is the score comparable between two different searches?
Only loosely. Each score is a judgement against the criteria that produced it, so comparing a prospect from a plumbing search with one from a logistics search compares two different questions. Compare within a search, not across.