What Verinode can detect, the signal catalog

What a "signal" is

Every night, and again whenever fresh data lands, Verinode runs a long list of detectors against your book of work. Each detector asks one narrow question: is a vendor's price drifting above market, has a carrier's approval rate dropped, is a certification about to lapse, has a job gone quiet with no activity for too long. When the answer is yes, it writes a signal: a single, dated finding with a severity, a domain, and (where the math supports it) a dollar figure.

Verinode does not decide anything on your behalf. A signal is a fact plus context, not an instruction. It reads the invoices, jobs, vendor relationships, certifications, and other records you have already shared, compares them against your own history and, where enough peers have contributed, against the anonymized network, and lays out what it found. You choose what to do with it, act, snooze, or dismiss.

The detectors that produce signals are the same ones whether you are on the free Contributor tier or a paid membership. What differs is the layer on top: clustering related signals into one weighed decision, rewriting a raw finding into plain English, and sanity-checking that a dollar figure is realistic is a deeper AI pass that runs for paid, reasoning-eligible memberships. A Contributor still sees the underlying signal; a paid membership additionally gets it clustered, framed, and reviewed before it reaches the Feed.

This article is the map: every domain a detector can fire in, roughly what each one watches for, and where in the product you actually run into it.

Where you actually run into a signal

Signals do not live on one page, they surface everywhere the thing they describe already lives:

• The Feed. New signals that clear the bar for your attention appear as cards on your home feed, most consequential first. See the-feed for how the feed prioritizes and filters.
• The Take Action tile at the top of a section page. Jobs, Vendors, Clients & carriers, Team, Equipment, Safety, Compliance, Recruiting, Certifications, Reputation, and Margin each show a row of tiles at the top of the page. A tile with a small glyph in its header is a signal (usually clustered into a decision) surfaced right where the entity it concerns lives. Clicking it opens the decision detail as an overlay on top of whatever section you were already on.
• The entity's own Signals panel. Open any vendor, job, client, or carrier profile and you will find a "Signals" section. It lists the open signals tied to that specific entity: a severity dot, the title, the status, and a dollar figure with its lever tag when there is one. When there are more than the panel shows, a "View all N signals →" link takes you to the Feed's Decisions filter.
• The Decisions workspace (/decisions). Once a signal has been clustered into a decision, or acted on, this is the working log. See the-decision-workspace and acting-on-decisions.
• The Triage Log. Signals Verinode auto-archived because they went stale or low-priority, with a one-click Restore. Covered below under Empty states.

Empty state (entity panel). A vendor, job, client, or carrier with nothing open reads: "No signals detected for this entity." That is not a broken widget, it means nothing about that specific record has crossed a detection threshold yet.

Anatomy of a signal

Whatever surface you meet it on, a signal carries the same handful of facts:

• Severity: Critical, Warning, or Info. Critical and Warning use the same Ember Red / Hard Hat Yellow tokens as the rest of the platform's signal palette; Info is neutral.
• Status: a signal starts new, becomes seen once you have viewed it, actioned once you have done something about it, or parked if you have asked to revisit it later. All four count as "open." Dismissed and resolved close it out.
• Confidence: High, Medium, or Low, shown as a small pill labeled "Confidence" wherever a signal card has room for it. This reflects how much evidence backs the finding, not how big the dollar figure is.
• A dollar figure and its period, when the underlying math supports one: an annual, monthly, weekly, per-job, or one-time estimate, shown with the matching suffix (/yr, /mo, /wk, /job, or "one-time"). Some carry a small lever tag underneath, cost, time, revenue, or risk, naming which kind of margin lever the finding pulls.
• A benchmark comparison, when peer data is behind it: how many standard deviations your number sits from the peer median. Verinode never shows you the number of peers behind that comparison, only the fact that a comparison exists and how far off you are.
• A recommendation, and on some signal types a pre-drafted email or an agent-authored plain-English rewrite (the card title and impact statement) that replaces the raw internal title once the framing pass has run.

The 16 detection domains, and the glyph each one wears

Every signal belongs to one of sixteen domains. Wherever a signal shows up as a tile, its header carries a small icon, and that icon almost always matches the section its domain naturally lives in, so you can tell what kind of finding you are looking at before you even read the title.

| Domain | What it watches | Where you'll see it | |---|---|---| | Vendor | Price changes against your own history, price above the peer market rate, negotiation opportunities, peer-preference shifts, contract renewals coming due, a state-required certification you don't hold for a category you buy, a carrier program that requires or blocks a vendor you use | Vendors (/vendors) | | Field / Job | Job-level margin, cycle-time, and payment anomalies; jobs that have gone quiet with no activity on record; an open job at risk of or past its cycle-time target; the job type (peril) lagging the rest of your own book | Jobs (/jobs) | | Carrier | A carrier missing its own SLA, denial-reason concentration (fixable vs. structural), a line item cut on your job but paid elsewhere in the network, your own denial-rate trend, per-program compliance posture | Clients & carriers (/clients), shown with the Risk glyph | | Billing | Billed, unpaid work with no written follow-up on record past the carrier's usual payment window | Margin's Cash Flow tools, shown with the Costs glyph. See cash-flow-and-runway | | Team | Single points of failure, a certification lapse compounding a coverage gap, tenure milestones, thin bench depth, unfilled custom roles, succession risk, declining cert coverage | Team (/team) | | Reputation | Composite or per-platform review score drops, slow response time to reviews, negative topic clusters, an unclaimed listing | Reputation (/reputation) | | Equipment | Aging assets, calibration coming due, capacity limits, rent-vs-buy math, peer equipment preference | Equipment (/equipment) | | Safety | Incident rate against peers, certification expiry, an insurance gap, EMR trending the wrong way, a certification wall closing in | Safety (/safety) | | Weather | A weather alert overlapping a service line where your crew coverage is already thin | No dedicated section; surfaces in the Feed and on Team, shown with the Risk glyph | | Competitive | Peer gaps, research you haven't reviewed yet, TPA program thresholds, momentum shifts, the peril that lags the peer cohort most on collection rate or cycle time | Benchmarks (/benchmarks). See benchmarks-overview | | Client | Revenue concentration risk in one client or carrier relationship, stance-pressure patterns | Clients & carriers (/clients). See clients-and-carriers | | Cert | Per-technician, per-firm, or per-sub certification expiry, a single point of failure on a credential, a certification wall closing in, a program lockout coming, a depth gap against peers | Certifications (/certifications) | | Compliance | Exposure findings, audit windows, audit-finding remediation, lagging adoption of a new regulation | Compliance (/compliance), shown with the Audits glyph | | Margin | Labor burden creeping up relative to revenue, crew utilization slipping, which cut line items concentrate your pushback, which job type is dragging your margin, adjuster-level supplement patterns, whether a slower or faster job stage actually correlates with better margin or satisfaction in your own book | Margin (/margin). See understanding-your-margin | | Process | Service lines you offer without a documented SOP, carrier-program SOP gaps, an adopted SOP missing the certifications it implies, whether an adopted SOP actually paid back, your step sequence diverging from the peer cohort's consensus, a milestone stage drifting slower than its own recent history, and data-quality flags when dated records are out of order | Processes (/processes) | | Setup | Onboarding actions still open: branding, email forwarding, pulling your books, emailing your insurance broker, exporting jobs from your job-management system, pulling the paper documents that live in a filing cabinet | Settings (/settings) |

A few adjacent detector families run the same way but sit outside this list of sixteen because they are scoped to their own section: Fleet, Facilities, Recruiting, and the growth/sales-and-marketing detectors. Each writes its own domain tag and wears its own section's glyph (Fleet, Facilities, Recruiting, Benchmarks) so it looks and behaves identically to the sixteen above, just narrower in scope.

From a raw signal to a decision you can act on

Most signals do not stay atomic for long. A handful of orchestrators run after the detectors: one clusters related recruiting signals for the same requisition, one clusters adjuster-level patterns for the same adjuster, one clusters team signals like a single point of failure plus a weather alert plus a bench-thin warning into one hiring, reallocation, or sub-out decision. Clustering only fires when there is more than one signal worth weighing together, an isolated finding stays a single card.

After clustering, a decision framer pass rewrites the newly-fired signals that have not been worded yet into a self-describing card title and, when there is no dollar figure to lead with, a plain-English consequence line. Last, a feasibility reviewer, the same business-analyst pass that reviews every decision, clamps the dollar figure to a realistic share of your revenue, rewrites the basis in plain language, and can suppress a decision entirely if its premise does not hold up once reviewed. None of this changes what the underlying signal found, it changes how legibly that finding reaches you.

1. 1Open the section page for the domain you care about (Margin, Vendors, Certifications, and so on).
2. 2Look at the Take Action tile row at the top. The glyph in a tile's header tells you which domain the finding is really about, even if it surfaced here.
3. 3Click the tile to open the decision detail as an overlay.
4. 4Read the evidence and the recommendation, and use "Discuss with agent" where it appears to work the specifics in chat.
5. 5Act, snooze, or dismiss. Nothing happens automatically; the choice and the follow-through are yours.

Self-resolving signals and the Triage Log

Many detectors are built to clean up after themselves. A stalled job stops flagging once activity resumes. A labor-burden or crew-utilization drift closes once the number recovers. A process data-quality flag clears once the share of out-of-order records drops back to a healthy level. You do not need to manually dismiss a signal whose underlying condition has already gone away, most check their own premise again on the next run and resolve themselves.

For everything else, Verinode runs a nightly triage pass that auto-archives low-priority signals so your Feed does not fill up with things that have gone stale: an Info-level finding idle 30 or more days, a Warning-level finding idle 60 or more days, a parked signal idle 90 or more days, anything past its own expiration, duplicates of a newer signal on the same subject, or ones the weekly agent judged low priority or a stalled plan. The Triage Log lists everything archived in the last 60 days with the reason, the domain, the entity, and a one-click Restore.

Triage Log empty state. "Nothing has been auto-archived recently. Triage runs nightly, check back tomorrow."

Coverage depends on data you have already sent

A detector cannot fire on data it has never seen. Coverage widens as you connect more sources, it does not require you to do anything with the signals themselves. See connecting-your-data and forwarding-documents for the fastest ways to get a section from quiet to populated: forwarding vendor invoices and carrier mail turns on Vendor and Carrier signals, exporting your job-management system turns on Field/Job and Process signals, emailing your insurance broker turns on Fleet, Equipment, Compliance, and Safety, and pulling your accounting reports turns on Margin, Billing, and the peer comparisons behind Competitive.

Peer-compared detectors are additionally gated by how many other operators in your category and size band have contributed. Rather than show you a comparison built on too thin a sample, those detectors stay quiet until Verinode has enough contributors behind the number, then they turn on with no action needed from you. See how-benchmarks-work and reading-a-benchmark for how that same cohort logic underlies every peer number on the platform.

Best-practice example

Say Margin's Take Action row shows a tile with the Process glyph and a dollar figure. That is a crew-utilization drift, margin domain, process-styled icon because the finding is about a slipping process, not a straightforward cost. Click it: the decision overlay shows your utilization dropped against your own trailing average, with a recommendation to check recent scheduling against job load. Meanwhile Certifications shows an unrelated tile with its own glyph flagging a certification wall closing in on a program you rely on. Neither of these is a coincidence Verinode is telling you to connect, they are two independent findings that happen to share a week. Work each on its own terms: the utilization drift against your scheduling, the cert wall against who on your team still needs to sit for the exam.

Data sources