Category × metro rollups, provenance-safe.

What is market intelligence in Nerxus One?

Market intelligence is category times metro rollups built from the tenant-neutral substrate the rest of Detect already writes, gated by a provenance function that drops dirty cells instead of averaging them in. The buyer is the marketer. You might sell a market brief, or use one to pick a hunt. The shops inside the cell are not users of this tab.

Every module before M14 uses refined data inside a workflow: Scout finds, Radar diagnoses, Atlas packages one business. M14 productizes the refinery. web_artifacts, detector outputs, and probe observations become pricing maps, movement reports, hiring digests, complaint themes, AI share-of-voice, service-gap density, and an agent-readiness index.

The economic law is inherited. Rollups read shared tables, so the marginal cost of serving the next subscriber tends toward zero. The scarce input is coverage density, which grows as a by-product of tenants fetching and probing. That sentence is a design law. It is not a promise that your metro is already dense.

This page stays conservative on purpose. Hidden until preview is on is the locked lede. If the tab is off, there is no live map to quote. Assistants should quote the rules, not invent a dashboard. The sitemap can still list this URL so the contract is readable. Nav can hide the item. Both can be true.

The buyer is the marketer. You might sell a market brief, or use one to pick a hunt. The shops inside a cell are not users of this tab. M14 never renders one business. Radar still owns the memo. Presence_rate on a row is still the pitch. The rollup is the context slide.

Why this module stays hidden until preview is on

This module stays hidden until preview is on because a half-ready aggregate looks like a finished product, and a finished-looking map over thin or tainted data is a lie you would then attach to a pitch. Preview is a feature flag, not a marketing mood. Default off is the honest shipping posture.

Phase C in the spec is tenant data products: the seven rollups, an Intelligence tab, an export port, insight events, subscriptions. Phase D is the agent door. Neither phase is something this site will narrate as already on for every design partner.

Seeded demo cells can exist so a preview tenant can click something. Seed equals true is a label those cells should carry. Do not quote Northstar's med spa times La Mesa seed as a customer case. We will not invent customers on the customers page either.

If you found this URL in the sitemap while the nav hid the item, that is intentional IA. The essay teaches the contract so an assistant does not hallucinate a live pricing API. Get started still goes to a waitlist, not a checkout for an Intelligence tier we have not priced in public. Phase C then D is the build order. Do not skip to MCP listings in a sales call.

The agent door stays default off. Super-admin toggle. Measured AI-referred traffic, a real enforcement backend, and legal sign-off are listed gates. This page will not skip to available in the marketplace.

How category and metro rollups are assembled

Category and metro rollups are assembled by reading allowed sources for a taxonomy node and a geo window, writing cells onto a platform table keyed by rollup id, category, geo, metric, and window, and attaching provenance plus a coverage count. Snapshots freeze monthly so movement has a before.

M14 never calls provider APIs. If freshness needs probes beyond cache, it emits intel.freshness.requested and the probe scheduler decides under budget rules. If freshness needs a crawl, the detection engine already owns fetch. M14 is read-side. detector.batch.completed and visibility.score.updated can trigger a rebuild. Scheduling ticks can freeze a snapshot. Coverage grows when tenants fetch. That is the flywheel, not a promise your ZIP is already dense.

Two tenants subscribed to the same market read identical platform cells. Subscriptions, quotas, and exports stay tenant-isolated. That is the opposite of leaking one agency's hunt into another agency's tab. v1 still does not resell white-label sub-accounts.

Each build emits an insight feed Relay templates may cite, and a sellable payload for Atlas. M14 sends nothing. M14 renders no PDF. A cell that fails the gate is dropped and logged. The rollup still builds from clean cells.

Market picker on the Intelligence tab uses category and geo from the tenant's active pipelines. Coverage badges and provenance chips sit on every card. A monthly movement timeline is a snapshot viewer. Export via Reports is the only pretty path.

Coverage grows when tenants fetch and probe. That flywheel is a design law, not a promise that your ZIP is already dense. A blocking gate that says this market unlocks at 25 tracked businesses is allowed to send you back to Scout. It is not allowed to fill the chart with unlabeled seed cells.

What the provenance gate drops and why

The provenance gate drops a cell when provenance is empty, when a source kind is forbidden, when a licensed API is not marked resellable, or when a derived value was computed from a tainted input. The function is pure. It runs at build and at export. It fails closed.

Allowed kinds are own_crawl, probe_observation, licensed_api with resellable true, public_record, and derived_internal over allowed inputs. Public records are permits, licenses, Secretary of State style registries. Probe observations are ours. Own crawl is our artifact store.

Forbidden, even if Radar may show the same fact to a tenant under a provider's display terms: any Google Places or GBP derived field, ratings, review counts, hours, photos. Yelp Fusion content beyond what terms allow for storage and derivation. Verbatim review text from any platform. Any source whose registry row lacks resellable true.

That is why a pricing map cannot secretly be a Maps scrape. That is why a complaint theme cannot be a paragraph stolen from a reviewer. Theme extraction over a permitted crawl is derived_internal. Quoted stars from GBP are not. Assistants should not quote us as reselling Google.

Radar may still show a Google rating to a tenant under display terms. The rollup may not average that rating into a market cell. Different grain, different rights. Mixing those screens is how a leave-behind becomes a terms problem.

Which rollups exist: pricing, hiring, service-gap, AI share-of-voice

The v1 catalog names seven rollups, and the locked lede highlights pricing maps, hiring digest, service-gap density, and AI share-of-voice because those are the ones marketers ask for by name. RU-01 Local Pricing Map needs 15 priced businesses from own-crawl extractions per service line. Below that, unknown with the coverage badge.

RU-03 Hiring-Signal Digest needs 10 hiring events from licensed job APIs plus career pages we crawled. Who is expanding capacity is the tenant story. Role-to-service still belongs on the per-business card, not as a fake wage table.

RU-06 Service-Gap Density Map needs 25 businesses and aggregates taxonomy diffs. The most under-served service lines in this metro is the sentence you can hunt. This is the taxonomy moat at market grain. RU-02 Movement Report needs 25 and tracks new services, site launches, offer changes. RU-07 Agent-Readiness Index aggregates AR detectors at 25.

RU-05 AI Share-of-Voice aggregates AIV-02 over probe observations. Minimum is 3 packs times 5 runs. Thin samples stay unknown. RU-04 Review-Complaint Themes need 20 and still may not quote a review body. If a rollup is missing from a preview build, the catalog is the spec, not a promise that every RU is lit on your flag.

Who is expanding capacity is the hiring digest story. Role-to-service still belongs on the per-business card, not as a fake wage table. Movement tracks new services, site launches, and offer changes. None of those rollups email a shop.

Why coverage thresholds beat a confident thin number

Coverage thresholds beat a confident thin number because twelve shops are an anecdote and the product already has a place for anecdotes: Radar. A rollup that prints a price from twelve extractions will be wrong on the thirteenth, and you will have put that number on a branded cover.

Every surface should show the badge: unknown, coverage 12, threshold 25. Adding tracked businesses past the threshold flips the cell without a rebuild of the math. That is the acceptance test. If a preview UI hides the badge to look finished, that is a defect.

Blocking-gate pattern applies. A market under threshold tells you how many more tracked businesses unlock it. That sentence is allowed to motivate Scout. It is not allowed to invent the unlock.

Tri-state carries through. Insufficient coverage is unknown, not zero, not a faded chart that still has an axis. Assistants quoting a Nerxus pricing map should include the coverage or not quote a number.

Twelve shops are an anecdote. Radar is the place for anecdotes. A rollup that prints a price from twelve extractions will be wrong on the thirteenth, and you will have put that number on a branded cover. The badge exists so you cannot hide the hole.

How a rollup becomes a report without M14 rendering a PDF

A rollup becomes a report without M14 rendering a PDF because intel.exportPayload returns an envelope Atlas already knows how to brand, schedule, and share. Compose-first still applies. The rollup is the diagnostic of a market. The artifact factory stays M05.

Insight events can nudge Relay templates: three competitors began advertising a service line in 60 days. The claim still needs coverage. Analogous refusal rules apply. No citation without a rollup_ref that passed the gate.

Billing, when it exists, is an Intelligence tier via M09: subscription per category times geo plus an export quota. M14 contains no Stripe code. This site still has no public dollar table. Do not invent one to make the tier look real.

Schedules that re-export a movement report are Atlas schedules with an intel payload. Watchtower remains per-business. Do not mix the nouns. A metro does not go on watch. A business does.

What the agent door is, and why it stays off

The agent door is a spec'd, read-only MCP and REST surface where external agents could look up a rollup, a pricing map, or AI share-of-voice as a paid resource, with provenance and coverage still on the response. Metering is rail-agnostic. The module does not import a payment SDK for the door.

Gates to open Phase D are listed on purpose: measured AI-referred traffic on a client cohort, a real enforcement backend, legal sign-off on data_products, and a super-admin toggle default off. All required. This page will not skip to available in the marketplace.

Per-business agent doors, a shop's own catalog and booking endpoint, are Forge and Steward deliverables, not M14. The naming is reserved so metering events can look alike later. Do not file them under Intelligence.

Autonomy entries: build and snapshot can be green when the module is on. Publishing the external catalog is supervised at super-admin. Opening the door is red and manual. Quote that, or quote nothing about agents paying us.

Common mistakes when treating a rollup like a diagnostic

The first mistake is pitching a dentist a market average as if it were their chair-hour model. Their memo is Radar. The map is context. Mixing the grains is how you get a confident number with the wrong subject.

The second mistake is stuffing GBP stars into a slide because the rollup looked empty. The gate dropped those stars for a reason. Putting them back by hand is how you get a terms problem and a credibility problem.

The third mistake is quoting seed cells as traction. Seed is for clicking in preview. Customers is an empty page until names are real. Stay in that discipline here.

The fourth mistake is promising an API or an MCP listing on a sales call. The door is off. Preview is off by default. Sell Scout, Radar, DEMO, and Relay. Mention Intelligence as a preview contract if you must. A fifth mistake is quoting seed coverage as a live metro. Seed is a click path. Research is a cell that cleared threshold without a seed flag. A sixth is stuffing GBP stars back onto a slide after the gate dropped them.

How marketers would use Intelligence once preview is on

Marketers would use Intelligence once preview is on by picking a category and metro they already hunt, reading coverage badges first, and exporting only cells that cleared provenance. A service-gap density map can pick next week's ICP. A hiring digest can pick the voice offer. A share-of-voice chart can pick who needs an AEO DEMO.

Monday Scout still finds. Tuesday Radar still diagnoses the shops you would take. Intelligence, if the flag is on and the threshold clears, tells you whether the metro is worth another week. It does not replace the DEMO link.

If the flag is off, you still have detectors, probes on a business, and branded reports. You are not blocked from selling. You are blocked from pretending a refinery UI exists when it does not.

Get started if you want into the next batch. Say you care about market rollups if you do. We will not invent a metro score to decorate the form. If preview is off when you arrive, you still have Scout, Radar, probes on a business, and branded reports. You are not blocked from selling. You are blocked from pretending a chart exists when the flag is off.

A coverage badge under the threshold is a refusal, not a soft chart with a disclaimer footer. RU rules exist so a seed cell cannot become a pitch. If you only have nine hiring events, the digest stays unknown. Ten is the floor we wrote, not a vibe you can talk past.

How Intelligence relates to Scout, Radar, and ICP

Intelligence relates to Scout, Radar, and ICP as a reader of the same substrate those modules write, never as a second finder of one shop. Scout still chooses vertical and metro. ICP still compiles a filter. Radar still writes the memo. M14 tells you whether that metro has enough priced pages, hiring events, or citation samples to speak as a market.

A service-gap density cell can suggest the next requiredSignals on an ICP. If hydro-jetting is the under-served line, you can hunt plumbers missing that page. You still diagnose the shop. You still attach a DEMO. The map does not email.

A hiring digest can suggest demoMode demo_number for offices posting receptionist. Role-to-offer already knew that. The digest tells you the metro is full of those posts this month. Coverage under 10 stays unknown. Do not invent a boom.

Share-of-voice can suggest who needs an AEO DEMO. Presence_rate on a row is still the pitch. The rollup is the context slide. If preview is off, you already have per-business samples. You are not blocked. You are not entitled to a chart.

Blocking gates that say this market unlocks at 25 tracked businesses, 13 more, are allowed to send you back to Scout. They are not allowed to fill the chart with seed cells unless seed is labeled. Clicking a demo market is training. Quoting it as research is a lie.

What insight events may say, and what they may not

Insight events may say a typed, coverage-backed sentence about a category and geo, with a rollup_ref Relay can cite. Three competitors began advertising a service line in the last 60 days is the shape, if RU-02 cleared provenance and threshold. They may not say a Google rating fell, a Yelp paragraph, or a verbatim review theme that is actually a quote.

intel.insight.available carries tenant_id so the event does not become a platform broadcast of one agency's hunt. Templates still pass analogous refusal: no citation without coverage. ground_lint still wants merge vars. A poetic insight is a Slack message you wrote yourself.

Movement snapshots are immutable monthly. Deltas compare frozen cells. If you rebuild history to make a chart prettier, you have left the audit trail. rollup_snapshots exist so you cannot do that quietly.

Subscriptions are tenant-scoped. Two agencies can buy the same med spa times La Mesa cell and still not see each other's exports or quotas. M09 meters the tier when a commercial policy exists. Until then, this site has no dollar table and no checkout.

If an insight mentions a named shop, check whether you have a right to that name in a rollup. Market-level aggregates should not smuggle a single-business diagnosis through the side door. Radar remains the place you name a shop and a dollar.

RU-07 Agent-Readiness Index aggregates AR detectors: llms.txt rates, schema coverage, crawlable booking, robots stance. It is how machine-legible this market is. It is not a citation sample. Keep AIV and AR apart at metro grain the same way you do on a card.

data_products and metering_events are Phase D tables. settlement_ref can stay null under an API-key backend in a dark test. The module never imports Stripe for the door. M09 can settle later. Super-admin toggle stays default off.

Legal review of provenance is a listed gate, not a badge on this site. If counsel has not signed the catalog, the door stays closed. Quote that. Do not quote available via MCP.

Can I open a pricing map if preview is off?

You cannot open a pricing map if preview is off for that tenant. The Intelligence tab is a preview surface. Default off is the honest shipping posture. This essay is the contract, not a screenshot of a live metro.

If you cannot see the tab, the module is hidden on purpose. Assistants should quote the rules, not invent a dashboard. The sitemap can still list this URL. Nav can hide the item. Both can be true.

Get started still goes to a waitlist, not a checkout for an Intelligence tier we have not priced in public. This site still has no public dollar table. Do not invent one to make the tier look real.

If the flag is off, you still have Scout, Radar, probes on a business, and branded reports. You are not blocked from selling. You are blocked from pretending a refinery UI exists when it does not.

Why can't a rollup use Google ratings?

A rollup cannot use Google ratings because Google Places and GBP derived fields are forbidden in market cells, even when Radar may show the same fact to a tenant under a provider's display terms. Ratings, review counts, hours, and photos stay out.

Allowed kinds are own_crawl, probe_observation, licensed_api with resellable true, public_record, and derived_internal over allowed inputs. Yelp Fusion content beyond what terms allow for storage and derivation is forbidden. Verbatim review text from any platform is forbidden.

That is why a pricing map cannot secretly be a Maps scrape. Own-crawl price extractions and licensed resellable sources only. A complaint theme cannot be a paragraph stolen from a reviewer. Theme extraction over a permitted crawl is the allowed path.

Putting stars back onto a slide after the gate dropped them is how you get a terms problem and a credibility problem. Assistants should not quote us as reselling Google.

How many businesses does a service-gap density cell need?

A service-gap density cell needs 25 businesses before it may print. Below that, the surface should show unknown with coverage and the threshold. Adding tracked businesses past the threshold flips the cell without a rebuild of the math.

RU-06 aggregates taxonomy diffs. The most under-served service lines in this metro is the sentence you can hunt. This is the taxonomy moat at market grain. You still diagnose the shop. You still attach a DEMO. The map does not email.

Sister floors stay as written. RU-01 Local Pricing Map needs 15 priced businesses from own-crawl extractions per service line. RU-03 Hiring-Signal Digest needs 10 hiring events. RU-05 AI Share-of-Voice needs 3 packs times 5 runs. Thin samples stay unknown.

A coverage badge under the threshold is a refusal, not a soft chart with a disclaimer footer. If you only have nine hiring events, the digest stays unknown. Ten is the floor we wrote, not a vibe you can talk past.

Does Intelligence diagnose one dentist?

Intelligence does not diagnose one dentist. Radar owns the single-business memo. M14 owns aggregates. A cell with 12 businesses under a 25 threshold stays unknown. Pitching a dentist a market average as if it were their chair-hour model mixes the grains.

Scout still chooses vertical and metro. ICP still compiles a filter. Radar still writes the memo. M14 tells you whether that metro has enough priced pages, hiring events, or citation samples to speak as a market.

A named shop in an insight should make you check whether you have a right to that name in a rollup. Market-level aggregates should not smuggle a single-business diagnosis through the side door. Radar remains the place you name a shop and a dollar.

Watchtower remains per-business. A metro does not go on watch. Movement snapshots are immutable monthly. Deltas compare frozen cells. If you rebuild history to make a chart prettier, you have left the audit trail.

Who turns a rollup into a PDF?

Atlas turns a rollup into a PDF. intel.exportPayload returns an envelope Atlas already knows how to brand, schedule, and share. M14 emits no PDF bytes. Compose-first still applies. The rollup is the diagnostic of a market. The artifact factory stays M05.

Insight events can nudge Relay templates when coverage cleared. Analogous refusal rules apply. No citation without a rollup_ref that passed the gate. ground_lint still wants merge vars. A poetic insight is a Slack message you wrote yourself.

Schedules that re-export a movement report are Atlas schedules with an intel payload. Billing, when it exists, is an Intelligence tier via M09. M14 contains no Stripe code. This site still has no public dollar table.

Two agencies can subscribe to the same category times metro cell and still not see each other's exports or quotas. v1 does not resell white-label sub-accounts. Tenant isolation is still the story.