Skip to main content

Prospect Qualification Standard

What this is: the one place that says which qualification gate is answered at which point in the funnel, by whom, and from what evidence. It is a routing document. It defines almost no new thresholds, because the thresholds already exist — scattered across fourteen documents, several of which disagree.

What this is not: another screen. The Results-First offer §5 anticipated this exactly — "This is not a second checklist and must not become one — a separate pass re-reading the same Google Business Profile is how two documents start disagreeing." A fifteenth list would be the worst available move. Where a number lives somewhere else, this document points at it and does not restate it.

Why it exists. A sweep on 2026-10-09 found fourteen places that filter, screen or disqualify a prospect, and seventeen contradictions between them. Only one of the fourteen runs automatically — the Prospect AI agent's filters — and it points the opposite way from most of the others.


1. The problem in one paragraph​

Qualification grew up in three eras: an ICP written for a referral business, a bucket taxonomy written for cold email, and an offer written for a results-contingent engagement. Each is internally coherent. Together they ask a human to apply criteria at points where the inputs cannot be known, and they let a machine select a supply that the offer cannot accept. Nothing here is wrong because someone was careless. It is wrong because three correct documents were never reconciled.


2. The layer model​

A gate is answerable only where its evidence exists. Placing one earlier than that does not make qualification stricter; it makes it fictional.

LayerApplied byCost per prospectWhat can genuinely be known
0 — SupplyMachine: agent filters, n8n qualification stageZeroTrade, geography, has-website, review count, rating
1 — Batch triageOperator, from the record aloneSecondsRating, review count, conversion score, trade
2 — SlottingOperator, from public pages~5 minutesGBP match, website, decision-maker, SMS path, bucket
3 — ConversationOn the call or the demoA callRevenue, headcount, deal value, intent, authority

The rule this document exists to enforce: no gate is applied at a layer above the one where its evidence first exists. Everything below follows from it.


3. The seven Results-First gates, placed​

Thresholds are canon in the offer §5. This table adds only the layer and the evidence.

#GateLayerEvidence
1Google Business Profile exists and resolves2The live profile, against the record
2Working website, theirs, not a Facebook page2Open it
3Rating ≥4.0; 10–150 reviews lifetime; ≥1 in the last 90 days; fewer than 3 in the last 601 for the count and rating, 2 for recency and paceCount and rating are on the record; recency and pace need the profile, sorted by newest
4Reachable decision-maker, confirmed current2Slotting step 8
5$100K–$2M revenue, 2–50 staff, one customer ≥$3,0003The demo. Nothing earlier; the cold call books the demo and does not ask it. Layer-0 proxy below
6Any lead bucket except review-response on its own2The decision tree
7Takes our number on Google, and lets the site sit with us3The call or the walkthrough. Nothing earlier can answer it

Gate 5 was being guessed, and now is not​

Gate 5 was checked at slotting — a step whose own rule is "Everything here is observable from public pages." Revenue, headcount and per-customer value are none of them observable. The gate was therefore being waved through on every prospect, which is worse than not having it, because the offer doc then reports seven gates passing when six were checked.

Gate 5 moves to layer 3. It is asked on the demo. The cold call scripts only book the demo and ask no qualifying question.

Its layer-0 proxy is the trade. "One new customer worth ≥$3,000" is invisible on a listing and strongly implied by what the business does:

Clears ≥$3,000 comfortablyDoes not, or only sometimes
Roofing, HVAC replacement, construction and remodeling, masonry, excavation, flooringLawn mowing and maintenance, cleaning, handyman, pest control, most single-visit repair

This is the lever that makes the niche choice a qualification decision rather than a marketing one. It is a proxy and it is wrong sometimes — a design-build landscaper clears $3,000 easily. A proxy decides what to source, never what to disqualify. A prospect who reaches a call is screened on the real question.


4. Layer 1 — batch triage​

New, and the cheapest point in the funnel. Given a weekly batch, decide which records are worth five minutes of slotting. It reads only fields the record already carries, so it adds no pass over the Google Business Profile and does not become the second checklist §5 warns about.

Run in the slotting checklist step 1. Any one failing parks the prospect.

  1. Conversion score above 60% — existing gate, unchanged
  2. Rating and review count — read both; they decide the bucket and gate 3's first half
  3. Trade — the gate-5 proxy above

⚠️ Triage must work from either supply route. The external route produces contacts with no prospect card and no AI conversion score, so a triage written around the card is inapplicable to the larger of the two supplies. Where the score is absent, check 1 is skipped and recorded as skipped — not failed, and not guessed.


5. Contradictions resolved​

Each of these was two documents disagreeing. The resolution is stated once, here.

5.1 Deal value: $3,000 or $500?​

The ICP screen and offer gate 5 say $3,000 or more. Older versions of the call scripts disqualified under $500; the scripts no longer ask, and the demo applies both.

Both stand, because they answer different questions. $500 is the floor below which no Symphony Core engagement makes arithmetic sense. $3,000 is gate 5, and gate 5 governs only Results-First eligibility. A prospect worth $1,500 per customer is a perfectly good client on standard terms and is not Results-First eligible. Neither number changes.

5.2 Review count: four floors that cannot all be right​

FloorWhereWhat it governs
10–400External sourcing bandWhich businesses get imported at all
Fewer than 15B1Which finding leads
10–150 lifetime, ≥1 in 90 days, fewer than 3 in 60Offer gate 3Whether Results-First can be offered. The ceiling exists because a shop already reviewing fast would meet the review trigger on its own pace
15 or moreB2 / Review Response PlayWhether the play applies

Bucket choice and offer eligibility are different questions, and a prospect can pass one and fail the other. The taxonomy's own worked example proves it: Talarico, 5.0 stars and one Google review, is called "a first-rate prospect" and slots B1 — and fails gate 3 outright. Both are correct. Talarico is sendable, B1, and not Results-First eligible.

The one that is a genuine mismatch is the sourcing band: a 10-review floor at import means the strongest B1 archetype in canon can never be sourced externally. Recorded in §7, not resolved here.

5.3 Headcount floor​

The ICP says 2–50 and explains why: "Below 2 = owner-only, often DIY-mindset." The Audit Accuracy SOP quality checklist — the thing prospecting-audit-verified actually attests — says 1–50. The external sourcing target is a "2-15 truck band" whose own compliance note says much of the real ICP is one-truck owner-operators.

The operational floor is 1. It matches what is attested today and what the sourced list contains. The ICP's caution about owner-only businesses stands as a caution, applied on the call at layer 3, not as a sourcing exclusion nobody enforces.

5.4 Which supply the offer can actually accept​

Gate 6 admits B1, B2 or B3 only. The two supply routes produce almost opposite things:

RouteProducesOffer-eligible?
Prospect AI agentGBP ratings Below 3.0 blocks B1; three appointment-shaped niches mean B3 rarely fires. Concentrates in B4 and B5Almost never
External sourcingHVAC and plumbing, emergency-shaped, 10–400 reviews — B3Yes — but these records have no prospect card and could not be slotted until §4 above

Both halves are now addressed: the agent's filter and niches change at layer 0 (§6), and layer-1 triage works without a prospect card.


6. Layer 0 — what the machines can and cannot filter​

⚠️ The platform cannot express the filter this offer needs. The Prospect AI agent's GBP rating filter offers only Below 1.5 · Below 2.0 · Below 3.0 · Below 4.0 · Below 5.0 · Any GBP ratings. There is no "above 4.0". The agent can exclude well-rated businesses; it can never select for them.

The consequence is structural, not a configuration mistake: rating and review screening cannot happen at layer 0 and must happen at layer 1. That is why layer 1 exists.

Target configuration for agent Noah:

SettingFromToWhy
GBP ratingsBelow 3.0Any GBP ratingsRemoves a filter that selects against gate 3 and blocks B1. The right filter does not exist
NichesRoofing, Landscaper, ConstructionRoofing, HVAC contractor, ConstructionGate-5 proxy: HVAC replacement clears $3,000; the landscaper category skews to maintenance. Also emergency-shaped, so B3 can fire

Configuration of record lives in the agent configuration reference, and is updated after the change, not before.


7. Known conflicts this document does not resolve​

Named here so they are in one place. Each is tracked elsewhere; none is re-proposed.

ConflictWhere it is tracked
sc-suppressed is described as overriding everything and does not exist in the sub-accountOffer sprint, rows at :66 and :67
Imported records carry no record-type:, which the hygiene standard requiresLead engine overview, L-1
Email cap: cadence SOP says 4, send checklist says 3, both dated 2026-10-09Unticketed — raise it
Audit freshness: 7–14 days is "refresh" in one SOP, "fail" in a checklist boxUnticketed — raise it
Phone-only, no-name prospects are complete, importable and callable, yet a fast-no for the email operator — roughly half the batchUnticketed — raise it
The sourcing band's 10-review floor excludes the canonical B1 archetype§5.2 above
The ICP ranks legal first; no supply route targets it, and the bucket taxonomy does not cover itA positioning decision, not a qualification one
No send-side check reads record-type:, the one mechanism that keeps clients out of cold outreachUnticketed — raise it


Document History​

VersionDateChanges
1.22026-11-10Gate 5 is asked on the demo, not the cold call, following the call scripts v1.19
1.12026-11-04The offer is final. Gate 3 gains a ceiling (150 lifetime, fewer than 3 in 60 days) and the 4.0 rating floor, because the result can now be 3 new reviews and a fast-reviewing shop would meet it unaided. Gate 6 admits any bucket except review-response alone. Gate 7 becomes taking our number and hosting the site with us, which only the call can answer, so it moves to layer 3
1.02026-10-09Initial version, written after a sweep found fourteen qualification surfaces and seventeen contradictions between them. Introduces the layer model and its single rule: no gate is applied at a layer above the one where its evidence first exists. Gate 5 moves from slotting to the call, because revenue, headcount and per-customer value are not observable from public pages and the gate was being waved through on every prospect; a trade-category proxy replaces it at sourcing. Adds layer-1 batch triage, defined to work without a prospect card so that the external supply route can be slotted at all. Resolves the deal-value, review-floor and headcount conflicts by separating bucket choice from offer eligibility, and records that the agent's GBP filter offers no "above" option, so rating screening cannot happen at source