Anatomy of a decision card

What a decision card is

A decision card is Verinode's unit of work. Every pattern IQ finds in your data, a vendor cycle time drifting past benchmark, a carrier's payments sliding later, a certification about to lapse, gets compressed into one card: one number (when there is a clean dollar figure), one title, one recommended move, and three controls. The design goal, stated in the component itself, is that a card should be readable by "a 10 year old": one number, one color, one button to press, with the proof hidden behind a "Why?" toggle instead of cluttering the surface.

Verinode does not decide anything on your behalf. A card surfaces a pattern and a recommendation; you choose to Act, park it for later, or let the agent explain its reasoning before you decide.

Where to find it

Open Decisions from the left sidebar at iq.verinode.ai/decisions. The sidebar badge on Decisions shows how many decisions are waiting on you.

The Decisions page opens to a gallery of compact tiles, one per decision, each showing its dollar figure (or risk line), title, and a status color. Click a tile and it expands into a full-width overlay: the decision card described in this article, with the swipe affordance "← / → or swipe to navigate decisions · Esc to go back" so you can page through every open decision without closing the view. The same card also appears inline in the Feed and inside a section's Findings tab (vendors, clients, equipment, and so on), so a Vendors decision reads identically whether you meet it in the Feed, in Decisions, or drilled into from the Vendors page.

Reading the card top to bottom

1. Header row

The top line carries three pieces of metadata:

• A colored dot and a status word. A pulsing red-orange dot with the label Declining means the trend behind this decision is getting worse. A plain dot with Needs context means IQ doesn't have enough confidence yet to commit to a number. Otherwise the label reads Decision, a fresh finding with a stable or improving trend.
• The entity name, when the decision is about a specific vendor, client, carrier, or piece of equipment, shown after a middle dot (for example "· Acme Restoration Supply").
• A confidence pill on the right: high, medium, or low. This is the same confidence the impact number below is built on, not a separate score.

2. The big impact number

This is the headline of the card, styled as the largest text on the surface. Two variants exist, depending on whether the underlying finding carries a clean dollar figure.

When there's a dollar figure, the card shows:

$42,300 /yr at stake

The dollar amount is IQ's estimate of what this pattern is costing you (or what it would recover), rounded to the nearest dollar. The unit suffix that follows it depends on the time base of the number, not a fixed assumption: /yr for an annual run rate (the default when no period is set), /mo or /wk for a monthly or weekly figure, /job for a per-job rate, and, distinctly, one-time (no slash) for a one-time pool like a stale-WIP catch-up or a backlog that clears once and doesn't repeat. Verinode is deliberate about this: labeling a one-time cleanup as an annual number would overstate the opportunity, so the two are never conflated.

Directly under the dollar figure is a small tag that tells you how solid the number is:

• Calculated, in green, when confidence is high, meaning the figure comes directly from your own operator data.
• Estimated, in amber, for medium or low confidence, meaning the figure is a working assumption IQ believes in but hasn't fully confirmed against your numbers yet.

Next to that tag you may see a second qualifier: Peer cohort (the estimate leans on what similar operators experience) or Industry baseline (the estimate leans on a published research benchmark rather than your peers). This only appears when the underlying finding is classified that way; most cards show just Calculated or Estimated with no second qualifier.

3. Impact basis

Right after the Calculated/Estimated tag, separated by a middle dot, is a plain-language sentence explaining how IQ arrived at the number, the impact basis. This is not boilerplate, it's the actual calculation trace for that specific decision. For an equipment cycle-time finding it might read something like:

Dehumidification cycle 12.4d vs benchmark 8.2d (4.2d drag) across 14 jobs

Hovering the dollar figure shows the same basis text as a tooltip. Not every decision has one, older or simpler findings may show just the dollar figure with no basis sentence.

4. When there's no dollar figure

Some risks don't reduce to a clean dollar amount, lapsed insurance, thin certification coverage, a safety gap. For these, the card swaps the dollar block for a plain consequence line instead:

• A small chip reading a risk kind (Compliance risk for safety, cert, and insurance domains; Cash flow risk for billing; Revenue risk for client and job domains; Business risk as the fallback for anything else) followed by an urgency word: Urgent for critical severity, Act soon for warning severity, or When you can otherwise.
• Below the chip, a plain-language sentence stating the consequence directly (for example, a lapsed certificate and what it exposes you to).

You'll only ever see one of these two blocks on a given card, never both.

5. The title

Below the impact block sits the card's title, the finding itself in plain language, for example "Acme Restoration Supply: response time drifting from peers." This is what the decision is actually about.

6. Lifecycle dots

A four-stage progress indicator: Flagged → Planned → Acting → Resolved. On a fresh card this sits at the first stage; it advances automatically as you act, as the agent builds a plan, and as plan steps get checked off. It's the same lifecycle-dot pattern used elsewhere on the platform (jobs use Assigned/Started/Billed/Paid), so you can scan a list of cards and see at a glance where each one stands without opening it.

7. The recommendation tile

One flat tile sits below the lifecycle dots, and it's always one of two things:

A recommended action, when IQ has a concrete move and confidence isn't low. It shows:

• Recommended action as a small copper label, then the action itself in bold, for example "Renegotiate your rate with this vendor."
• "If you don't:" followed by the consequence of inaction, when IQ has one to show.
• For medium or low confidence cards only, a "Refine this estimate:" line naming the specific thing that would tighten the number, drawn straight from what IQ needs from you (for example, forwarding a category of documents). This is the same thread the agent picks up in chat after you click Act, so you're never asked to answer the same question twice.

"Needs your context", when IQ doesn't yet have enough to commit to a specific recommendation (this also covers the low-confidence case). It reads the consequence headline if one exists, or falls back to "We see a pattern but can't recommend specifics without more from you," and below that: "Click Act and the agent will ask 2-3 quick questions to shape a recommendation you can commit to."

8. Peer decision path

Below the recommendation tile, separated by a hairline, is the peer path preview, Verinode's answer to "what did operators like me do here, and did it work?" It only appears when there's enough cohort coverage behind this specific signal to report on safely; when there isn't, the block simply doesn't render, there's no "not enough data" message to clutter the card.

When it does show, it reads as two or three short lines, something like:

18 similar operators faced this in the last 90 days. 11 acted (61%). Of those, 7 saw improvement (64%). Most chose a two-step plan.

The cohort description ("similar operators," "operators nationally," and so on) never names a specific peer or reveals which operators are in it, only a scope description. The share of operators who acted, and of those, the share whose outcome measurably improved, are read straight off aggregated peer data, never your own guess. The optional third line calls out the most common shape of plan peers ran (for example a two-step or a multi-step plan), when one path clearly dominates.

The three controls

Every open card carries the same three buttons at the bottom:

1. 1Act. Commits you to this decision. Verinode records that you acted, and routes you straight into the full workspace at /decisions/[id], where the agent starts building (or continues building) a step-by-step plan. There's no intermediate form on the card itself, one click hands you to the workspace and the agent.
2. 2Not now. Opens a small reason picker in place of the button row: Too busy, Need more info, Not convinced, or Other. Choosing one parks the decision, it fades to half-opacity and collapses to a single line: "Parked, this will resurface later." Parking isn't a permanent dismissal, IQ will bring it back around.
3. 3Why? Only shown when the card has proof to back it up. Toggles open the proof drawer directly beneath the buttons, without navigating away.

Before you ever open a card, the compact tile you clicked from (in the Decisions gallery, the Feed, or a section's Findings tab) carries a third option alongside Act and Not now: Ignore. Ignoring a decision is the more permanent dismissal, distinct from parking it; ignored decisions move to the Ignored status filter on the Decisions page rather than resurfacing.

The proof drawer

Clicking Why? opens a short, flat panel with up to two sections:

• Proof points. A row of small stat blocks, your own value for the metric behind this decision, the peer average when one exists, and a research or top-performer benchmark when one exists. Each stat has a label, a value, and sometimes a short context line underneath.
• Root cause. A short causal chain, one line per step, each one an arrow-prefixed sentence describing how one metric feeds into the next. This is IQ's explanation of why the pattern is happening, not just that it is.

Both sections only render when there's content for them; a card with no causal chain simply shows proof points, and vice versa.

What the card looks like once you've acted

Once you click Act, the card you see afterward on Decisions, the Feed, or a section's Findings tab is not the same card, it becomes a compact task list:

• A pulsing status label reading Plan loading… while the agent is still drafting, or In progress once a real plan exists, plus the entity name and a N/M done step counter.
• The decision's title, the same lifecycle dots, now advanced past Flagged.
• Either the plan's steps as a checklist (check one off directly from the card, without opening the workspace, and the next unchecked step is subtly highlighted with a due-date label like "Today," "Tomorrow," "In 3d," or "2d overdue"), or, if the agent hasn't produced steps yet, a placeholder line: "Tap Act and Verinode IQ will draft a step-by-step plan to tackle this," or, if you've already acted and the plan is still generating, "Verinode IQ is preparing your action plan, open it to watch it come together."
• A footer naming the next step to do, or, once every step is checked off, "All done, close this out," alongside a Go to Action Plan → link into the full workspace.

Clicking anywhere on this task-list card (other than a checkbox) takes you to the same /decisions/[id] workspace as clicking Act the first time.

What the card looks like once a decision resolves

When a decision closes out with a recorded outcome, the card collapses to a short green summary: a checkmark and "You acted on this Nd ago," the outcome's label, an estimated savings figure so far when one exists ("Estimated savings so far: ~$X"), and, when there's a before/after metric to show, the two values with an arrow between them.

What you'll see with little or no data yet

A brand-new operator, or one whose data hasn't reached a particular pattern yet, won't see empty decision cards, they simply won't see a card for that pattern until IQ has enough signal to raise it. Individual pieces inside a card degrade gracefully rather than showing placeholder errors: no impact basis sentence if IQ hasn't built one for that finding type, no peer decision path if your cohort doesn't have coverage on that specific signal yet, no "Refine this estimate" line once confidence is high enough that there's nothing left to tighten. As more of your data flows in, through connected tools, forwarded documents, or answers you give the agent, these pieces fill in on their own.

Related reading

• The Decision Workspace covers what happens after you click Act: the full /decisions/[id] page, the agent thread, and how plans get built and tracked.
• Acting on decisions walks through the Act / Park / Ignore lifecycle in more depth, including how parked and ignored decisions resurface or stay dismissed.
• The Feed explains where decisions first surface as you sign in, before you ever visit the Decisions page.
• Understanding your margin explains net income, the "what you keep" number many decision cards' dollar figures ultimately roll up into.
• How benchmarks work and Reading a benchmark explain the peer and industry comparisons behind the Peer cohort and Industry baseline qualifiers on a card's impact tag.