Running deep research, end to end
A deeper pass runs where it could change a decision and not where it would only add detail. It is triggered by an undecided qualification or by a person asking for one prospect, it produces claims that each carry a source, and everything it finds attaches to that prospect's record — not to the workspace's Knowledge, which is a different thing entirely.
The chain, stage by stage#
- Trigger — a qualification that came back undecided because the evidence was thin, or a person choosing one prospect and asking for more.
- External event — none. Everything read is already published; the organisation is not contacted and does not learn that it is being researched.
- Authentication and workspace resolution — the prospect, the spend and the output all belong to one workspace, and the allowance being drawn on is that workspace's.
- Request — the cost router decides. A deeper pass is authorised where the extra evidence could move the decision, and declined where the decision is already settled either way.
- Canonical record — the existing prospect record is the target. Nothing new is created to hold the results, which is why a second pass deepens a record rather than forking it.
- Reasoning — public sources are read more widely and more carefully than in the first pass, against the same criteria.
- Knowledge, memory and rules — a claim that cannot be evidenced does not reach the record. A deeper pass produces fewer, better-sourced statements rather than more assertions.
- Autonomy and approval — nothing is sent, so no autonomy mode applies. Research is inward work and stays that way.
- Action — no provider is contacted on the prospect's behalf and no message is composed.
- Result — the qualification is re-decided with the new evidence, and may come back the same. A pass that confirms an existing judgement is not a wasted one if it was genuinely undecided before.
- Relationship, timeline and memory — nothing joins a person's timeline. A prospect is still not a relationship.
- Audit, usage and Needs You — the spend is metered against the workspace's allowance and recorded, and a prospect needing a human judgement is ranked into Needs You.
What decides that it is worth running#
Research spend is routed rather than spread evenly. The question the router asks is not "is this organisation interesting" but "could more evidence change what happens next" — and for most prospects the answer is no, in both directions.
| Situation | Deeper pass? | Why |
|---|---|---|
| Clearly outside the criteria | No | More detail about an organisation you would not approach changes nothing |
| Clearly inside the criteria, contactable | No | The decision is already made; the next step is outreach, not more reading |
| Undecided on thin evidence | Yes | This is the case the budget exists for |
| Qualified but the research is stale | Re-research rather than deepen | Freshness governs this; the problem is age, not depth |
| A person asks for one prospect | Yes | A human judgement that this one matters outranks the router |
Where the output goes#
Everything found attaches to the prospect record as evidenced claims. That is worth stating precisely because of the thing it is not: research output does not enter the workspace's Knowledge. Knowledge is material a workspace deliberately supplies for grounding answers to its own customers, and quietly filling it with third-party research would change what Connect states as true about your business.
- An evidenced claim
- Reaches the record, with the source that stated it. This is the only shape research output takes.
- An unevidenced observation
- Does not reach the record at all. It cannot influence a score from somewhere off to the side.
- A contact address
- Attaches with its provenance, if a source states one. A deeper pass finds addresses more often, and still constructs none.
- A fit reason
- Is rewritten against the new evidence, so the explanation on the record matches what is now known rather than what was known first.
When it disappoints, and what that tells you#
Some organisations publish almost nothing. A deeper pass over a business with one page and no filings returns very little, and the honest result is a prospect that stays undecided with a note that the evidence is not there to be found.
That outcome is information about your criteria. If a whole segment of your list behaves this way, the segment is one that will be judged on a conversation rather than on research — which argues for the phone, for an introduction, or for a shorter first message that asks rather than asserts.
Check the allowance before assuming a pass did not run.
Result Prospecting work stops when the allowance is spent. It is a refusal with a reason, not silence, and it is visible rather than inferred.
Read what the pass actually added, not just the new score.
Result Two or three well-sourced claims that move a decision are a good result. A longer record that changes nothing is the case the router exists to avoid buying.
If research is old rather than thin, re-research instead of deepening.
Result Freshness is the right control for age. Deepening stale material produces a more detailed picture of the past.
Questions#
Can I force a deeper pass on a prospect the router declined?
Yes. A person choosing one prospect outranks the routing decision, which is the sensible order: the router is optimising spend across a list, and you may know something about this organisation that no public source carries.
Does research output become part of my Knowledge?
No, and the separation is deliberate. Research attaches to the prospect record as evidenced claims. Knowledge is what your workspace supplies for grounding answers about your own business, and third-party research does not belong in it.
Why did a deeper pass return almost nothing?
Because the organisation publishes almost nothing. That is a real finding rather than a failure, and if it happens across a segment of your list it is telling you those prospects will be judged in a conversation instead.
What happens when the allowance runs out mid-list?
Work stops with a refusal that says so rather than continuing silently on guesswork. The prospects already researched keep everything they gained; the rest wait for the allowance to reset or for the plan to change.